Pushed by YOCaml 2

2024-10-04 10:48:33 +00:00
15 changed files with 2261 additions and 0 deletions
--- a/articles/2024-02-03-python-str-repr.html
+++ b/articles/2024-02-03-python-str-repr.html
@ -0,0 +1,278 @@
 <!doctype html>
 <html lang="en">
  <head>
    <meta charset="utf-8">
    <meta http-equiv="x-ua-compatible" content="ie=edge">
    <meta name="viewport" content="width=device-width, initial-scale=1">
    <title>
        Robur's blogPython&apos;s `str.__repr__()`
    </title>
    <meta name="description" content="Reimplementing Python string escaping in OCaml">
    <link type="text/css" rel="stylesheet" href="https://blog.robur.coop/css/hl.css">
    <link type="text/css" rel="stylesheet" href="https://blog.robur.coop/css/style.css">
    <script src="https://blog.robur.coop/js/hl.js"></script>
    <link rel="alternate" type="application/rss+xml" href="https://blog.robur.coop/feed.xml" title="blog.robur.coop">
  </head>
  <body>
    <header>
      <h1>blog.robur.coop</h1>
      <blockquote>
        The <strong>Robur</strong> cooperative blog.
      </blockquote>
    </header>
    <main><a href="https://blog.robur.coop/index.html">Back to index</a>
 <article>
    <h1>Python&apos;s `str.__repr__()`</h1>
    <ul class="tags-list"><li>OCaml</li><li>Python</li><li>unicode</li></ul><p>Sometimes software is written using whatever built-ins you find in your programming language of choice.
 This is usually great!
 However, it can happen that you depend on the precise semantics of those built-ins.
 This can be a problem if those semantics become important to your software and you need to port it to another programming language.
 This story is about Python and its <code>str.__repr()__</code> function.</p>
 <p>The piece of software I was helping port to <a href="https://ocaml.org/">OCaml</a> was constructing a hash from the string representation of a tuple.
 The gist of it was basically this:</p>
 <pre><code class="language-python">def get_id(x):
    id = (x.get_unique_string(), x.path, x.name)
    return myhash(str(id))
 </code></pre>
 <p>In other words it's a Python tuple consisting of mostly strings but also a <code>PosixPath</code> object.
 The way <code>str()</code> works is it calls the <code>__str__()</code> method on the argument objects (or otherwise <code>repr(x)</code>).
 For Python tuples the <code>__str__()</code> method seems to print the result of <code>repr()</code> on each elemenet separated by a comma and a space and surrounded by parenthesis.
 So good so far.
 If we can precisely emulate <code>repr()</code> on strings and <code>PosixPath</code> it's easy.
 In the case of <code>PosixPath</code> it's really just <code>'PosixPath('+repr(str(path))+')'</code>;
 so in that case it's down to <code>repr()</code> on strings - which is <code>str.__repr__()</code>,</p>
 <p>There had been a previous attempt at this that would use OCaml's string escape functions and surround the string with single quotes (<code>'</code>).
 This works for some cases, but not if the string has a double quote (<code>&quot;</code>).
 In that case OCaml would escape the double quote with a backslash (<code>\&quot;</code>) while python would not escape it.
 So a regular expression substitution was added to replace the escape sequence with just a double quote.
 This pattern of finding small differences between Python and OCaml escaping had been repeated,
 and eventually I decided to take a more rigorous approach to it.</p>
 <h2 id="what-is-a-string"><a class="anchor" aria-hidden="true" href="#what-is-a-string"></a>What is a string?</h2>
 <p>First of all, what is a string? In Python? And in OCaml?
 In OCaml a string is just a sequence of bytes.
 Any bytes, even <code>NUL</code> bytes.
 There is no concept of unicode in OCaml strings.<br>
 In Python there is the <code>str</code> type which is a sequence of Unicode code points[^python-bytes].
 I can recommend reading Daniel Bünzli's <a href="https://ocaml.org/p/uucp/13.0.0/doc/unicode.html#minimal">minimal introduction to Unicode</a>.
 Already here there is a significant gap in semantics between Python and OCaml.
 For many practical purposes we can get away with using the OCaml <code>string</code> type and treating it as a UTF-8 encoded Unicode string.
 This is what I will do as in both the Python code and the OCaml code the data being read is a UTF-8 (or often only the US ASCII subset) encoded string.</p>
 <h2 id="what-does-a-string-literal-look-like"><a class="anchor" aria-hidden="true" href="#what-does-a-string-literal-look-like"></a>What does a string literal look like?</h2>
 <h3 id="ocaml"><a class="anchor" aria-hidden="true" href="#ocaml"></a>OCaml</h3>
 <p>I will not dive too deep into the details of OCaml string literals, and focus mostly on how they are escaped by the language built-ins (<code>String.escaped</code>, <code>Printf.printf &quot;%S&quot;</code>).
 Normal printable ASCII is printed as-is.
 That is, letters, numbers and other symbols except for backslash and double quote.
 There are the usual escape sequences <code>\n</code>, <code>\t</code>, <code>\r</code>, <code>\&quot;</code> and <code>\\</code>.
 Any byte value can be represented with decimal notation <code>\032</code> or octal notation '\o040' or hexadecimal notation <code>\x20</code>.
 The escape functions in OCaml has a preference for the decimal notation over the hexadecimal notation.
 Finally I also want to mention the Unicode code point escape sequence <code>\u{3bb}</code> which represents the UTF-8 encoding of U+3BB.
 While the escape functions do not use it, it will become handy later on.
 Illegal escape sequences (escape sequences that are not recognized) will emit a warning but otherwise result in the escape sequence as-is.
 It is common to compile OCaml programs with warnings-as-errors, however.</p>
 <h3 id="python"><a class="anchor" aria-hidden="true" href="#python"></a>Python</h3>
 <p>Python has a number of different string literals and string-like literals.
 They all use single quote or double quote to delimit the string (or string-like) literals.
 There is a preference towards single quotes in <code>str.__repr__()</code>.
 You can also triple the quotes if you like to write a string that uses a lot of both quote characters.
 This format is not used by <code>str.__repr__()</code> so I will not cover it further, but you can read about it in the <a href="https://docs.python.org/3/reference/lexical_analysis.html#strings">Python reference manual</a>.
 The string literal can optionally have a prefix character that modifies what type the string literal is and how its content is interpreted.</p>
 <p>The <code>r</code>-prefixed strings are called <em>raw strings</em>.
 That means backslash escape sequences are not interpreted.
 In my experiments they seem to be quasi-interpreted, however!
 The string <code>r&quot;\&quot;</code> is considered unterminated!
 But <code>r&quot;\&quot;&quot;</code> is fine as is interpreted as <code>'\\&quot;'</code>[^raw-escape-example].
 Why this is the case I have not found a good explanation for.</p>
 <p>The <code>b</code>-prefixed strings are <code>bytes</code> literals.
 This is close to OCaml strings.</p>
 <p>Finally there are the unprefixed strings which are <code>str</code> literals.
 These are the ones we are most interested in.
 They use the usual escape <code>\[ntr&quot;]</code> we know from OCaml as well as <code>\'</code>.
 <code>\032</code> is <strong>octal</strong> notation and <code>\x20</code> is hexadecimal notation.
 There is as far as I know <strong>no</strong> decimal notation.
 The output of <code>str.__repr__()</code> uses the hexadecimal notation over the octal notation.
 As Python strings are Unicode code point sequences we need more than two hexadecimal digits to be able to represent all valid &quot;characters&quot;.
 Thus there are the longer <code>\u0032</code> and the longest <code>\U00000032</code>.</p>
 <h2 id="intermezzo"><a class="anchor" aria-hidden="true" href="#intermezzo"></a>Intermezzo</h2>
 <p>While studying Python string literals I discovered several odd corners of the syntax and semantics besides the raw string quasi-escape sequence mentioned earlier.
 One fact is that Python doesn't have a separate character or Unicode code point type.
 Instead, a character is a one element string.
 This leads to some interesting indexing shenanigans: <code>&quot;a&quot;[0][0][0] == &quot;a&quot;</code>.
 Furthermore, strings separated by spaces only are treated as one single concatenated string: <code>&quot;a&quot; &quot;b&quot; &quot;c&quot; == &quot;abc&quot;</code>.
 These two combined makes it possible to write this unusual snippet: <code>&quot;a&quot; &quot;b&quot; &quot;c&quot;[0] == &quot;a&quot;</code>!
 For byte sequences, or <code>b</code>-prefixed strings, things are different.
 Indexing a bytes object returns the integer value of that byte (or character):</p>
 <pre><code class="language-python">&gt;&gt;&gt; b&quot;a&quot;[0]
 97
 &gt;&gt;&gt; b&quot;a&quot;[0][0]
 &lt;stdin&gt;:1: SyntaxWarning: 'int' object is not subscriptable; perhaps you missed a comma?
 Traceback (most recent call last):
  File &quot;&lt;stdin&gt;&quot;, line 1, in &lt;module&gt;
 TypeError: 'int' object is not subscriptable
 </code></pre>
 <p>For strings <code>\x32</code> can be said to be shorthand for <code>&quot;\u0032&quot;</code> (or <code>&quot;\u00000032&quot;</code>).
 But for bytes <code>&quot;\x32&quot; != &quot;\u0032&quot;</code>!
 Why is this?!
 Well, bytes is a byte sequence and <code>b&quot;\u0032&quot;</code> is not interpreted as an escape sequence and is instead <strong>silently</strong> treated as <code>b&quot;\\u0032&quot;</code>!
 Writing <code>&quot;\xff&quot;.encode()</code> which encodes the string <code>&quot;\xff&quot;</code> to UTF-8 is <strong>not</strong> the same as <code>b&quot;\xff&quot;</code>.
 The bytes <code>&quot;\xff&quot;</code> consist of a single byte with decimal value 255,
 and the Unicode wizards reading will know that the Unicode code point 255 (or U+FF) is encoded in two bytes in UTF-8.</p>
 <h2 id="where-is-the-python-code"><a class="anchor" aria-hidden="true" href="#where-is-the-python-code"></a>Where is the Python code?</h2>
 <p>Finding the implementation of <code>str.__repr__()</code> turned out to not be so easy.
 In the end I asked on the Internet and got a link to <a href="https://github.com/python/cpython/blob/963904335e579bfe39101adf3fd6a0cf705975ff/Objects/unicodeobject.c#L12245-L12405">cpython's <code>Objects/unicodeobject.c</code></a>.
 And holy cow!
 That's some 160 lines of C code with two loops, a switch statement and I don't know how many chained and nested if statements!
 Meanwhile the OCaml implementation is a much less daunting 52 lines of which about a fifth is a long comment.
 It also has two loops which each contain one much more tame match expression (roughly a C switch statement).
 In both cases they first loop over the string to compute the size of the output string.
 The Python implementation also counts the number of double quotes and single quotes as well as the highest code point value.
 The latter I'm not sure why they do, but my guess it's so they can choose an efficient internal representation.
 Then the Python code decides what quote character to use with the following algorithm:<br>
 Does the string contain single quotes but no double quotes? Then use double quotes. Otherwise use single quotes.
 Then the output size estimate is adjusted with the number of backslashes to escape the quote character chosen and the two quotes surrounding the string.</p>
 <p>Already here it's clear that a regular expression substitution is not enough by itself to fix OCaml escaping to be Python escaping.
 My first step then was to implement the algorithm only for US ASCII.
 This is simpler as we don't have to worry much about Unicode, and I could implement it relatively quickly.
 The first 32 characters and the last US ASCII character (DEL or <code>\x7f</code>) are considered non-printable and must be escaped.
 I then wrote some simple tests by hand.
 Then I discovered the OCaml <a href="https://github.com/zshipko/ocaml-py">py</a> library which provides bindings to Python from OCaml.
 Great! This I can use to test my implementation against Python!</p>
 <h2 id="how-about-unicode"><a class="anchor" aria-hidden="true" href="#how-about-unicode"></a>How about Unicode?</h2>
 <p>For the non-ascii characters (or code points rather) they are either considered <em>printable</em> or <em>non-printable</em>.
 For now let's look at what that means for the output.
 A printable character is copied as-is.
 That is, there is no escaping done.
 Non-printable characters must be escaped, and python wil use <code>\xHH</code>, <code>\uHHHH</code> or <code>\UHHHHHHHH</code> depending on how many hexadecimal digits are necessary to represent the code point.
 That is, the latin-1 subset of ASCII (<code>0x80</code>-<code>0xff</code>) can be represented using <code>\xHH</code> and neither <code>\u00HH</code> nor <code>\U000000HH</code> will be used etc.</p>
 <h3 id="what-is-a-printable-unicode-character"><a class="anchor" aria-hidden="true" href="#what-is-a-printable-unicode-character"></a>What is a printable Unicode character?</h3>
 <p>In the cpython <a href="https://github.com/python/cpython/blob/963904335e579bfe39101adf3fd6a0cf705975ff/Objects/unicodeobject.c#L12245-L12405">function</a> mentioned earlier they use the function <code>Py_UNICODE_ISPRINTABLE</code>.
 I had a local clone of the cpython git repository where I ran <code>git grep Py_UNICODE_ISPRINTABLE</code> to find information about it.
 In <a href="https://github.com/python/cpython/blob/963904335e579bfe39101adf3fd6a0cf705975ff/Doc/c-api/unicode.rst?plain=1#L257-L265">unicode.rst</a> I found a documentation string for the function that describes it to return false if the character is nonprintable with the definition of nonprintable as the code point being in the categories &quot;Other&quot; or &quot;Separator&quot; in the Unicode character database <strong>with the exception of ASCII space</strong> (U+20 or <code> </code>).</p>
 <p>What are those &quot;Other&quot; and &quot;Separator&quot; categories?
 Further searching for the function definition we find in <a href="https://github.com/python/cpython/blob/963904335e579bfe39101adf3fd6a0cf705975ff/Include/cpython/unicodeobject.h#L683"><code>Include/cpython/unicodeobject.h</code></a> the definition.
 Well, we find <code>#define Py_UNICODE_ISPRINTABLE(ch) _PyUnicode_IsPrintable(ch)</code>.
 On to <code>git grep _PyUnicode_IsPrintable</code> then.
 That function is defined in <a href="https://github.com/python/cpython/blob/963904335e579bfe39101adf3fd6a0cf705975ff/Objects/unicodectype.c#L158-L163"><code>Objects/unicodectype.c</code></a>.</p>
 <pre><code class="language-C">/* Returns 1 for Unicode characters to be hex-escaped when repr()ed,
   0 otherwise.
   All characters except those characters defined in the Unicode character
   database as following categories are considered printable.
      * Cc (Other, Control)
      * Cf (Other, Format)
      * Cs (Other, Surrogate)
      * Co (Other, Private Use)
      * Cn (Other, Not Assigned)
      * Zl Separator, Line ('\u2028', LINE SEPARATOR)
      * Zp Separator, Paragraph ('\u2029', PARAGRAPH SEPARATOR)
      * Zs (Separator, Space) other than ASCII space('\x20').
 */
 int _PyUnicode_IsPrintable(Py_UCS4 ch)
 {
    const _PyUnicode_TypeRecord *ctype = gettyperecord(ch);
    return (ctype-&gt;flags &amp; PRINTABLE_MASK) != 0;
 }
 </code></pre>
 <p>Ok, now we're getting close to something.
 Searching for <code>PRINTABLE_MASK</code> we find in <a href="https://github.com/python/cpython/blob/963904335e579bfe39101adf3fd6a0cf705975ff/Tools/unicode/makeunicodedata.py#L450-L451"><code>Tools/unicode/makeunicodedata.py</code></a> the following line of code:</p>
 <pre><code class="language-Python">if char == ord(&quot; &quot;) or category[0] not in (&quot;C&quot;, &quot;Z&quot;):
    flags |= PRINTABLE_MASK
 </code></pre>
 <p>So the algorithm is really if the character is a space character or if its Unicode general category doesn't start with a <code>C</code> or <code>Z</code>.
 This can be implemented in OCaml using the uucp library as follows:</p>
 <pre><code class="language-OCaml">let py_unicode_isprintable uchar =
  (* {[if char == ord(&quot; &quot;) or category[0] not in (&quot;C&quot;, &quot;Z&quot;):
           flags |= PRINTABLE_MASK]} *)
  Uchar.equal uchar (Uchar.of_char ' ')
  ||
  let gc = Uucp.Gc.general_category uchar in
  (* Not those categories starting with 'C' or 'Z' *)
  match gc with
  | `Cc | `Cf | `Cn | `Co | `Cs | `Zl | `Zp | `Zs -&gt; false
  | `Ll | `Lm | `Lo | `Lt | `Lu | `Mc | `Me | `Mn | `Nd | `Nl | `No | `Pc | `Pd
  | `Pe | `Pf | `Pi | `Po | `Ps | `Sc | `Sk | `Sm | `So -&gt;
      true
 </code></pre>
 <p>After implementing unicode I expanded the tests to generate arbitrary OCaml strings and compare the results of calling my function and Python's <code>str.__repr__()</code> on the string.
 Well, that didn't go quite well.
 OCaml strings are just any byte sequence, and ocaml-py expects it to be a UTF-8 encoded string and fails on invalid UTF-8.
 Then in qcheck you can &quot;assume&quot; a predicate which means if a predicate doesn't hold on the generated value then the test is skipped for that input.
 So I implement a simple verification of UTF-8.
 This is far from optimal because qcheck will generate a lot of invalid utf-8 strings.</p>
 <p>The next test failure is some unassigned code point.
 So I add to <code>py_unicode_isprintable</code> a check that the code point is assigned using <code>Uucp.Age.age uchar &lt;&gt; `Unassigned</code>.</p>
 <p>Still, qcheck found a case I hadn't considered: U+61D.
 My python version (Python 3.9.2 (default, Feb 28 2021, 17:03:44)) renders this as <code>'\u061'</code> while my OCaml function prints it as-is.
 In other words my implementation considers it printable while python does not.
 I try to enter this Unicode character in my terminal, but nothing shows up.
 Then I look it up and its name is <code>ARABIC END OF TEXT MARKER</code>.
 The general category according to uucp is <code>`Po</code>.
 So this <strong>should</strong> be a printable character‽</p>
 <p>After being stumped by this for a while I get the suspicion it may be dependent on the Python version.
 I am still on Debian 11 and my Python version is far from being the latest and greatest.
 I ask someone with a newer Python version to write <code>'\u061d'</code> in a python session.
 And 'lo! It prints something that looks like <code>''</code>!
 Online I figure out how to get the unicode version compiled into Python:</p>
 <pre><code class="language-Python">&gt;&gt;&gt; import unicodedata
 &gt;&gt;&gt; unicodedata.unidata_version
 '13.0.0'
 </code></pre>
 <p>Aha! And with uucp we find that the unicode version that introduced U+61D to be 14.0:</p>
 <pre><code class="language-OCaml"># Uucp.Age.age (Uchar.of_int 0x61D);;
 - : Uucp.Age.t = `Version (14, 0)
 </code></pre>
 <p>My reaction is this is seriously some ungodly mess we are in.
 Not only is the code that instigated this journey highly dependent on Python-specifics - it's also dependent on the specific version of unicode and thus the version of Python!</p>
 <p>I modify our <code>py_unicode_isprintable</code> function to take an optional <code>?unicode_version</code> argument and replace the &quot;is this unassigned?&quot; check with the following snippet:</p>
 <pre><code class="language-OCaml">let age = Uucp.Age.age uchar in
 (match (age, unicode_version) with
 | `Unassigned, _ -&gt; false
 | `Version _, None -&gt; true
 | `Version (major, minor), Some (major', minor') -&gt;
    major &lt; major' || (major = major' &amp;&amp; minor &lt;= minor'))
 </code></pre>
 <p>Great! I modify the test suite to first detect the unicode version python uses and then pass that version to the OCaml function.
 Now I can't find anymore failing test cases!</p>
 <h2 id="epilogue"><a class="anchor" aria-hidden="true" href="#epilogue"></a>Epilogue</h2>
 <p>What can we learn from this?
 It is easy to say in hindsight that a different representation should have been chosen.
 However, arriving at this insight takes time.
 The exact behavior of <code>str.__repr__()</code> is poorly documented.
 Reaching my understanding of <code>str.__repr__()</code> took hours of research and reading the C implementation.
 It often doesn't seem to be worth it to spend so much time on research for a small function.
 Technical debt is a real thing and often hard to predict.
 Below is the output of <code>help(str.__repr__)</code>:</p>
 <pre><code class="language-Python">__repr__(self, /)
    Return repr(self)
 </code></pre>
 <p>Language and (standard) library designers could consider whether the slightly nicer looking strings are worth the added complexity users eventually are going to rely on - inadvertently or not.
 I do think strings and bytes in Python are a bit too complex.
 It is not easy to get a language lawyer[^language-lawyer] level understanding.
 In my opinion it is a mistake to not at least print a warning if there are illegal escape sequences - especially considering there are escape sequences that are valid in one string literal but not another.</p>
 <p>Unfortunately it is often the case that to get a precise specification it is necessary to look at the implementation.
 For testing your implementation hand-written tests are good.
 Testing against the original implementation is great, and if combined with property-based testing or fuzzing you may find failing test cases you couldn't dream up!
 I certainly didn't see it coming that the output depends on the Unicode version.
 As is said, testing can only show the presence of bugs, but with a, in a sense, limited domain like this function you can get pretty close to showing absence of bugs.</p>
 <p>I enjoyed working on this.
 Sure, it was frustrating and at times I discovered some ungodly properties, but it's a great feeling to study and understand something at a deeper level.
 It may be the last time I need to understand Python's <code>str.__repr__()</code> this well, but if I do I now have the OCaml code and this blog post to reread.</p>
 <p>If you are curious to read the resulting code you may find it on github at <a href="https://github.com/reynir/python-str-repr">github.com/reynir/python-str-repr</a>.
 I have documented the code to make it more approachable and maintainable by others.
 Hopefully it is not something that you need, but in case it is useful to you it is licensed under a permissive license.</p>
 <p>If you have a project in OCaml or want to port something to OCaml and would like help from me and my colleagues at <a href="https://robur.coop/">Robur</a> please <a href="https://robur.coop/Contact">get in touch</a> with us and we will figure something out.</p>
 <p>[^python-bytes]: There is as well the <code>bytes</code> type which is a byte sequence like OCaml's <code>string</code>.
 The Python code in question is using <code>str</code> however.</p>
 <p>[^raw-escape-example]: Note I use single quotes for the output. This is what Python would do. It would be equivalent to <code>&quot;\\\&quot;&quot;</code>.</p>
 <p>[^language-lawyer]: <a href="http://catb.org/jargon/html/L/language-lawyer.html">A person, usually an experienced or senior software engineer, who is intimately familiar with many or most of the numerous restrictions and features (both useful and esoteric) applicable to one or more computer programming languages. A language lawyer is distinguished by the ability to show you the five sentences scattered through a 200-plus-page manual that together imply the answer to your question “if only you had thought to look there”.</a></p>
    </article>
        </main>
    <footer>
        <a href="https://github.com/xhtmlboi/yocaml">Powered by <strong>YOCaml</strong></a>
        <br />
    </footer>
    <script>hljs.highlightAll();</script>
  </body>
 </html>
--- a/articles/2024-08-21-OpenVPN-and-MirageVPN.html
+++ b/articles/2024-08-21-OpenVPN-and-MirageVPN.html
@ -0,0 +1,229 @@
 <!doctype html>
 <html lang="en">
  <head>
    <meta charset="utf-8">
    <meta http-equiv="x-ua-compatible" content="ie=edge">
    <meta name="viewport" content="width=device-width, initial-scale=1">
    <title>
        Robur's blogMirageVPN and OpenVPN
    </title>
    <meta name="description" content="Discoveries made implementing MirageVPN, a OpenVPN-compatible VPN library">
    <link type="text/css" rel="stylesheet" href="https://blog.robur.coop/css/hl.css">
    <link type="text/css" rel="stylesheet" href="https://blog.robur.coop/css/style.css">
    <script src="https://blog.robur.coop/js/hl.js"></script>
    <link rel="alternate" type="application/rss+xml" href="https://blog.robur.coop/feed.xml" title="blog.robur.coop">
  </head>
  <body>
    <header>
      <h1>blog.robur.coop</h1>
      <blockquote>
        The <strong>Robur</strong> cooperative blog.
      </blockquote>
    </header>
    <main><a href="https://blog.robur.coop/index.html">Back to index</a>
 <article>
    <h1>MirageVPN and OpenVPN</h1>
    <ul class="tags-list"><li>MirageVPN</li><li>OpenVPN</li><li>security</li></ul><p>At <a href="https://robur.coop/">Robur</a> we have been busy at work implementing our OpenVPN™-compatible MirageVPN software.
 Recently we have implemented the <a href="https://blog.robur.coop/articles/miragevpn-server.html">server side</a>.
 In order to implement this side of the protocol I studied parts of the OpenVPN™ source code and performed experiments to understand what the implementation does at the protocol level.
 Studying the OpenVPN™ implementation has lead me to discover two security issues: CVE-2024-28882 and CVE-2024-5594.
 In this article I will talk about the relevant parts of the protocol, and describe the security issues in detail.</p>
 <p>A VPN establishes a secure tunnel in which (usually) IP packets are sent.
 The OpenVPN protocol establishes a TLS tunnel[^openvpn-tls] with which key material and configuration options are negotiated.
 Once established the TLS tunnel is used to exchange so-called control channel messages.
 They are NUL-terminated (well, more on that later) text messages sent in a single TLS record frame (mostly, more on that later).</p>
 <p>I will describe two (groups) of control channel messages (and a bonus control channel message):</p>
 <ul>
 <li><code>EXIT</code>, <code>RESTART</code>, and <code>HALT</code></li>
 <li><code>PUSH_REQUEST</code> / <code>PUSH_REPLY</code></li>
 <li>(<code>AUTH_FAILED</code>)</li>
 </ul>
 <p>The <code>EXIT</code>, <code>RESTART</code>, and <code>HALT</code> messages share similarity.
 They are all three used to signal to the client that it should disconnect[^disconnect] from the server.
 <code>HALT</code> tells the client to disconnect and suggests the client should terminate.
 <code>RESTART</code> also tells the client to disconnect and suggests the client can reconnect either to the same server or the next server if multiple are configured depending on flags in the message.
 <code>EXIT</code> tells the <em>peer</em> that it is exiting and the <em>peer</em> should disconnect.
 The last one can be sent by either the server or the client and is useful when the underlying transport is UDP.
 It informs the peer that the sender is exiting and will (soon) not be receiving and ACK'ing messages; for UDP the peer would otherwise (re)send messages until a timeout.</p>
 <p>Because the underlying transport can either be TCP or UDP the sender may have no guarantees that the message arrives.
 OpenVPN's control channel implements a reliable layer with ACKs and retransmissions to work around that.
 To accomodate this OpenVPN™ will wait five seconds before disconnecting to allow for retransmission of the exit message.</p>
 <h3 id="the-bug"><a class="anchor" aria-hidden="true" href="#the-bug"></a>The bug</h3>
 <p>While I was working on implementing more control channel message types I modified a client application that connects to a server and sends pings over the tunnel - instead of ICMPv4 echo requests I modified it to send the <code>EXIT</code> control channel message once a second.
 In the server logs I saw that the server successfully received the <code>EXIT</code> message!
 But nothing else happened.
 The server just kept receiving <code>EXIT</code> messages but for some reason it never disconnected the client.</p>
 <p>Curious about this behavior I dived into the OpenVPN™ source code and found that on each <code>EXIT</code> message it (re)schedules an exit (disconnect) timer! That is, every time the server receives an <code>EXIT</code> message it'll go &quot;OK! I'll shut down this connection in five seconds&quot; forgetting it promised to do so earlier, too.</p>
 <h3 id="implications"><a class="anchor" aria-hidden="true" href="#implications"></a>Implications</h3>
 <p>At first this seemed like a relatively benign bug.
 What's the worst that could happen if a client says &quot;let's stop in five second! No, five seconds from now! No, five seconds from now!&quot; etc?
 Well, it turns out the same timer is used when the server sends an exit message.
 Ok, so what?
 The client can hold open a resource it <em>was</em> authorized to use <em>longer</em>.
 So we have a somewhat boring potential denial of service attack.</p>
 <p>Then I learned more about the management interface.
 The management interface is a text protocol to communicate with the OpenVPN server (or client) and query for information or send commands.
 One command is the <code>client-kill</code> command.
 The documentation says to use this command to &quot;[i]mmediately kill a client instance[...]&quot;.
 In practice it sends an exit message to the client (either a custom one or the default <code>RESTART</code>).
 I learnt that it shares code paths with the exit control messages to schedule an exit (disconnect)[^kill-immediately].
 That is, <code>client-kill</code> schedules the same five second timer.</p>
 <p>Thus a malicious client can, instead of exiting on receiving an exit or <code>RESTART</code> message, send back repeatedly <code>EXIT</code> to the server to reset the five second timer.
 This way the client can indefinitely delay the exit/disconnect assuming sufficiently stable and responsive network.
 This is suddenly not so good.
 The application using the management interface might be enforcing a security policy which we can now circumvent!
 The client might be a former employee in a company, and the security team might want to revoke access to the internal network for the former employee, and in that process uses <code>client-kill</code> to kick off all of his connecting clients.
 The former employee, if prepared, can circumvent this by sending back <code>EXIT</code> messages repeatedly and thus keep unauthorized access.
 Or a commercial VPN service may try to enforce a data transfer limit with the same mechanism which is then rather easily circumvented by sending <code>EXIT</code> messages.</p>
 <p>Does anyone use the management interface in this way?
 I don't know.
 If you do or are aware of software that enforces policies this way please do reach out to <a href="https://reyn.ir/contact.html">me</a>.
 It would be interesting to hear and discuss.
 The OpenVPN security@ mailing list took it seriously enough to assign it CVE-2024-28882.</p>
 <h2 id="openvpn-configuration-language"><a class="anchor" aria-hidden="true" href="#openvpn-configuration-language"></a>OpenVPN configuration language</h2>
 <p>Next up we have <code>PUSH_REQUEST</code> / <code>PUSH_REPLY</code>.
 As the names suggest it's a request/response protocol.
 It is used to communicate configuration options from the server to the client.
 These options include routes, ip address configuration, negotiated cryptographic algorithms.
 The client signals it would like to receive configuration options from the server by sending the <code>PUSH_REQUEST</code> control channel message[^proto-push-request].
 The server then sends a <code>PUSH_REPLY</code> message.</p>
 <p>The format of a <code>PUSH_REPLY</code> message is <code>PUSH_REPLY,</code> followed by a comma separated list of OpenVPN configuration directives terminated by a NUL byte as in other control channel messages.
 Note that this means pushed configuration directives cannot contain commas.</p>
 <p>When implementing the <code>push</code> server configuration directive, which tells the server to send the parameter of <code>push</code> as a configuration option to the client in the <code>PUSH_REPLY</code>, I studied how exactly OpenVPN™ parses configuration options.
 I learned some quirks of the configuration language which I find surprising and somewhat hard to implement.
 I will not cover all corners of the configuration language.</p>
 <p>In some sense you could say the configuration language of OpenVPN™ is line based.
 At least, the first step to parsing configuration directives as OpenVPN 2.X does is to read one line at a time and parse it as one configuration directive[^inline-files].
 A line is whatever <code>fgets()</code> says it is - this includes the newline if not at the end of the file[^configuration-newlines].
 This is how it is for configuration files.
 However, if it is a <code>PUSH_REPLY</code> a <em>&quot;line&quot;</em> is the text string up to a comma or the end of file (or, importantly, a NUL byte).
 This &quot;line&quot; tokenization is done by repeatedly calling OpenVPN™'s <code>buf_parse(buf, ',', line, sizeof(line))</code> function.</p>
 <pre><code class="language-C">/* file: src/openvpn/buffer.c */
 bool
 buf_parse(struct buffer *buf, const int delim, char *line, const int size)
 {
    bool eol = false;
    int n = 0;
    int c;
    ASSERT(size &gt; 0);
    do
    {
        c = buf_read_u8(buf);
        if (c &lt; 0)
        {
            eol = true;
        }
        if (c &lt;= 0 || c == delim)
        {
            c = 0;
        }
        if (n &gt;= size)
        {
            break;
        }
        line[n++] = c;
    }
    while (c);
    line[size-1] = '\0';
    return !(eol &amp;&amp; !strlen(line));
 }
 </code></pre>
 <p><code>buf_parse()</code> takes a <code>struct buffer*</code> which is a pointer to a byte array with a offset and length field, a delimiter character (in our case <code>','</code>), a destination buffer <code>line</code> and its length <code>size</code>.
 It calls <code>buf_read_u8()</code> which returns the first character in the buffer and advances the offset and decrements the length, or returns <code>-1</code> if the buffer is empty.
 In essence, <code>buf_parse()</code> &quot;reads&quot; from the buffer and copies over to <code>line</code> until it encounters <code>delim</code>, a NUL byte or the end of the buffer.
 In that case a NUL byte is written to <code>line</code>.</p>
 <p>What is interesting is that a NUL byte is effectively considered a delimiter, too, and that it is consumed by <code>buf_parse()</code>.
 Next, let's look at how incoming control channel messages are handled (modified for brevity):</p>
 <pre><code class="language-C">/* file: src/openvpn/forward.c (before fix) */
 /*
 * Handle incoming configuration
 * messages on the control channel.
 */
 static void
 check_incoming_control_channel(struct context *c, struct buffer buf)
 {
    /* force null termination of message */
    buf_null_terminate(&amp;buf);
    /* enforce character class restrictions */
    string_mod(BSTR(&amp;buf), CC_PRINT, CC_CRLF, 0);
    if (buf_string_match_head_str(&amp;buf, &quot;AUTH_FAILED&quot;))
    {
        receive_auth_failed(c, &amp;buf);
    }
    else if (buf_string_match_head_str(&amp;buf, &quot;PUSH_&quot;))
    {
        incoming_push_message(c, &amp;buf);
    }
    /* SNIP */
 }
 </code></pre>
 <p>First, the buffer is ensured to be NUL terminated by replacing the last byte with a NUL byte.
 This is already somewhat questionable as it could make an otherwise invalid message valid.
 Next, character class restrictions are &quot;enforced&quot;.
 What this roughly does is removing non-printable characters and carriage returns and line feeds from the C string.
 The macro <code>BSTR()</code> returns the underlying buffer behind the <code>struct buffer</code> with the offset added.
 Notably, <code>string_mod()</code> works on (NUL terminated) C strings and not <code>struct buffer</code>s.
 As an example, the string (with the usual C escape sequences):</p>
 <pre><code>&quot;PUSH_REPLY,line \nfeeds\n,are\n,removed\n\000&quot;
 </code></pre>
 <p>becomes</p>
 <pre><code>&quot;PUSH_REPLY,line feeds,are,removed\000ed\n\000&quot;
 </code></pre>
 <p>As you can see, if interpreted as a C string we have removed the line feeds.
 But what is this at the end?
 It is the same last 4 bytes from the original string.
 More generally, it is the last N bytes from the original string if the original string has N line feeds (or other disallowed characters).</p>
 <p>The whole buffer is still passed to the push reply parsing.
 Remember that the &quot;line&quot; parser will not only consume commas as the line delimiter, but also NUL bytes!
 This means the configuration directives are parsed as lines:</p>
 <pre><code class="language-C">&quot;line feeds&quot;
 &quot;are&quot;
 &quot;removed&quot;
 &quot;ed\n&quot;
 </code></pre>
 <p>With this technique we can now inject (almost; the exception is NUL) arbitrary bytes as configuration directive lines.
 This is bad because the configuration directive is printed to the console if it doesn't parse.
 As a proof of concept I sent a <code>PUSH_REPLY</code> with an embedded BEL character, and the OpenVPN™ client prints to console (abbreviated):</p>
 <pre><code>Unrecognized option or missing or extra parameter(s): ^G
 </code></pre>
 <p>The <code>^G</code> is how the BEL character is printed in my terminal.
 I was also able to hear an audible bell.</p>
 <p>A more thorough explanation on how terminal escape sequences can be exploited can be found on <a href="https://www.gresearch.com/news/g-research-the-terminal-escapes/">G-Reasearch's blog</a>.</p>
 <h3 id="the-fix"><a class="anchor" aria-hidden="true" href="#the-fix"></a>The fix</h3>
 <p>The fix also is also a first step towards decoupling the control channel messaging from the TLS record frames.
 First, the data is split on NUL bytes in order to get the control channel message(s), and then messages are rejected if they contain illegal characters.
 This solves the vulnerability described previously.</p>
 <p>Unfortunately, it turns out that especially for the <code>AUTH_FAILED</code> control channel message it is easy to create invalid messages:
 If 2FA is implemented using the script mechanism sending custom messages they easily end with a newline asking the client to enter the verification code.
 I believe in 2.6.12 the client tolerates trailing newline characters.</p>
 <h2 id="conclusion"><a class="anchor" aria-hidden="true" href="#conclusion"></a>Conclusion</h2>
 <p>The first bug, the timer rescheduling bug, is at least 20 years old!
 It hasn't always been exploitable, but the bug itself goes back as far as the git history does.
 I haven't attempted further software archeology to find the exact time of introduction.
 Either way, it's old and gone unnoticed for quite a while.</p>
 <p>I think this shows that diversity in implementations is a great way to exercise corner cases, push forward (protocol) documentation efforts and get thorough code review by motivated peers.
 This work was funded by <a href="https://nlnet.nl/project/MirageVPN/">the EU NGI Assure Fund through NLnet</a>.
 In my opinion, this shows that funding one open source project can have a positive impact on other open source projects, too.</p>
 <p>[^openvpn-tls]: This is not always the case. It is possible to use static shared secret keys, but it is mostly considered deprecated.
 [^disconnect]: I say &quot;disconnect&quot; even when the underlying transport is the connection-less UDP.
 [^kill-immediately]: As the alert reader might have realized this is inaccurate. It does not kill the client &quot;immediately&quot; as it will wait five seconds after the exit message is sent before exiting. At best this will kill a cooperating client once it's received the kill message.
 [^proto-push-request]: There is another mechanism to request a <code>PUSH_REPLY</code> earlier with less roundtrips, but let's ignore that for now. The exact message is <code>PUSH_REQUEST&lt;NUL-BYTE&gt;</code> as messages need to be NUL-terminated.
 [^inline-files]: An exception being inline files which can span multiple lines. They vaguely resemble XML tags with an open <code>&lt;tag&gt;</code> and close <code>&lt;/tag&gt;</code> each on their own line with the data in between. I doubt these are sent in <code>PUSH_REPLY</code>s, but I can't rule out without diving into the source code that it isn't possible to send inline files.
 [^configuration-newlines]: This results in the quirk that it is possible to sort-of escape a newline in a configuration directive. But since the line splitting is done <em>first</em> it's not possible to continue the directive on the next line! I believe this is mostly useless, but it is a way to inject line feeds in configuration options without modifying the OpenVPN source code.</p>
    </article>
        </main>
    <footer>
        <a href="https://github.com/xhtmlboi/yocaml">Powered by <strong>YOCaml</strong></a>
        <br />
    </footer>
    <script>hljs.highlightAll();</script>
  </body>
 </html>
--- a/articles/gptar.html
+++ b/articles/gptar.html
@ -0,0 +1,145 @@
 <!doctype html>
 <html lang="en">
  <head>
    <meta charset="utf-8">
    <meta http-equiv="x-ua-compatible" content="ie=edge">
    <meta name="viewport" content="width=device-width, initial-scale=1">
    <title>
        Robur's blogGPTar
    </title>
    <meta name="description" content="Hybrid GUID partition table and tar archive">
    <link type="text/css" rel="stylesheet" href="https://blog.robur.coop/css/hl.css">
    <link type="text/css" rel="stylesheet" href="https://blog.robur.coop/css/style.css">
    <script src="https://blog.robur.coop/js/hl.js"></script>
    <link rel="alternate" type="application/rss+xml" href="https://blog.robur.coop/feed.xml" title="blog.robur.coop">
  </head>
  <body>
    <header>
      <h1>blog.robur.coop</h1>
      <blockquote>
        The <strong>Robur</strong> cooperative blog.
      </blockquote>
    </header>
    <main><a href="https://blog.robur.coop/index.html">Back to index</a>
 <article>
    <h1>GPTar</h1>
    <ul class="tags-list"><li>OCaml</li><li>gpt</li><li>tar</li><li>mbr</li><li>persistent storage</li></ul><p>At <a href="https://robur.coop/">Robur</a> we developed a piece of software for mirroring or exposing an <a href="https://opam.ocaml.org/">opam</a> repository.
 We have it deployed at <a href="https://opam.robur.coop/">opam.robur.coop</a>, and you can use it as an alternative to opam.ocaml.org.
 It is usually more up-to-date with the git <a href="https://github.com/ocaml/opam-repository">opam-repository</a> than opam.ocaml.org although in the past it suffered from <a href="https://blog.osau.re/articles/lwt_pause.html">occasional availability issues</a>.
 I can recommend reading Hannes' post about <a href="https://hannes.robur.coop/Posts/OpamMirror">opam-mirror</a>.
 This article is about adding a partition table to the disk <a href="https://hannes.robur.coop/Posts/OpamMirror#code-development-and-improvements">as used by opam-mirror</a>.
 For background I can recommend reading the previously linked subsection of the opam-mirror article.</p>
 <h2 id="the-opam-mirror-persistent-storage-scheme"><a class="anchor" aria-hidden="true" href="#the-opam-mirror-persistent-storage-scheme"></a>The opam-mirror persistent storage scheme</h2>
 <p>Opam-mirror uses a single block device for its persistent storage.
 On the block device it stores cached source code archives from the opam repository.
 These are stored in a <a href="https://en.wikipedia.org/wiki/Tar_(computing)">tar archive </a> consisting of files whose file name is the sha256 checksum of the file contents.
 Furthermore, at the end of the block device some space is allocated for dumping the cloned git state of the upstream (git) opam repository as well as caches storing maps from md5 and sha512 checksums respectively to the sha256 checksums.
 The partitioning scheme is entirely decided by command line arguments.
 In other words, there is no partition table on the disk image.</p>
 <p>This scheme has the nice property that the contents of the tar archive can be inspected by regular tar utilities in the host system.
 Due to the append-only nature of tar and in the presence of concurrent downloads a file written to the archive may be partial or corrupt.
 Opam-mirror handles this by prepending a <code>pending/</code> directory to partial downloads and <code>to-delete/</code> directory for corrupt downloads.
 If there are no files after the failed download in the tar archive the file can be removed without any issues.
 Otherwise a delete would involve moving all subsequent files further back in the archive - which is too error prone to do robustly.
 So using the tar utilities in the host we can inspect how much garbage has accumulated in the tar file system.</p>
 <p>The big downside to this scheme is that since the disk partitioning is not stored on the disk the contents can easily become corrupt if the wrong offsets are passed on the command line.
 Therefore I have for a long time been wanting to use an on-disk partition table.
 The problem is both <a href="https://en.wikipedia.org/wiki/MBR_partition_table">MBR</a> and <a href="https://en.m.wikipedia.org/wiki/GUID_Partition_Table">GPT</a> (GUID Partition Table) store the table at the beginning of the disk.
 If we write a partition table at the beginning it is suddenly not a valid tar archive anymore.
 Of course, in Mirage we can just write and read the table at the end if we please, but then we lose the ability to inspect the partition table in the host system.</p>
 <h2 id="gpt-header-as-tar-file-name"><a class="anchor" aria-hidden="true" href="#gpt-header-as-tar-file-name"></a>GPT header as tar file name</h2>
 <p>My first approach, which turned out to be a dead end, was when I realized that a GPT header consists of 92 bytes at the beginning followed by reserved space for the remainder of the LBA.
 The reserved space should be all zeroes, but it seems no one bothers to enforce this.
 What's more is that a tar header starts with the file name in the first 100 bytes.
 This got me thinking we could embed a GPT header inside a tar header by using the GPT header as the tar header file name!</p>
 <p>I started working on implementing this, but I quickly realized that 1) the tar header has a checksum, and 2) the gpt header has a checksum as well.
 Having two checksums that cover each other is tricky.
 Updating one checksum affects the other checksum.
 So I started reading a paper written by Martin Stigge et al. about <a href="https://sar.informatik.hu-berlin.de/research/publications/SAR-PR-2006-05/SAR-PR-2006-05_.pdf">reversing CRC</a> as the GPT header use CRC32 checksum.
 I ended up writing <a href="https://github.com/reynir/gptar/commit/e8269c4959e98aa0cf339cf250ff4e6671fd344c">something</a> that I knew was incorrect.</p>
 <p>Next, I realized the GPT header's checksum only covers the first 92 bytes - that is, the reserved space is not checksummed!
 I find this and the fact that the reserved space should be all zeroes but no one checks odd about GPT.
 This simplified things a lot as we don't have to reverse any checksums!
 Then I implemented a test binary that produces a half megabyte disk image with a hybrid GPT and tar header followed by a tar archive with a file <code>test.txt</code> whose content is <code>Hello, World!</code>.
 I had used the byte <code>G</code> as the link indicator.
 In POSIX.1-1988 the link indicators <code>A</code>-<code>Z</code> are reserved for vendor specific extensions, and it seemed <code>G</code> was unused.
 A mistake I made was to not update the tar header checksum - the <a href="https://github.com/mirage/ocaml-tar">ocaml-tar</a> library doesn't support this link indicator value so I had manually updated the byte value in the serialized header but forgot to update the checksum.
 This was easily remediated as the checksum is a simple sum of the bytes in the header.
 The changes made are viewable on <a href="https://github.com/reynir/gptar/compare/e8269c4959e98aa0cf339cf250ff4e6671fd344c...03a57a1cef17eeb66b0d883dbe441c9c4b9093bd">GitHub</a>.
 I also had to work around a <a href="https://github.com/mirage/ocaml-tar/issues/144">bug in ocaml-tar</a>.
 GNU tar was successfully able to list the archive.
 A quirk is that the archive will start with a dummy file <code>GPTAR</code> which consists of any remaining space in the first LBA if the sector size is greater than 512 bytes followed by the partition table.</p>
 <h2 id="protective-mbr"><a class="anchor" aria-hidden="true" href="#protective-mbr"></a>Protective MBR</h2>
 <p>Unfortunately, neither fdisk nor parted recognized the GPT partition table.
 I was able to successfully read the partition table using <a href="https://github.com/mirage/ocaml-gpt">ocaml-gpt</a> however.
 This puzzled me.
 Then I got a hunch: I had read about <a href="https://en.m.wikipedia.org/wiki/GUID_Partition_Table#Protective_MBR_(LBA_0)">protective MBRs</a> on the Wikipedia page on GPT.
 I had always thought it was optional and not needed in a new system such as Mirage that doesn't have to care too much about legacy code and operating systems.</p>
 <p>So I started comparing the layout of MBR and tar.
 The V7 tar format only uses the first 257 bytes of the 512 byte block.
 The V7 format is differentiated by the UStar, POSIX/pax and old GNU tar formats by not having the string <code>ustar</code> at byte offset 257[^tar-ustar].
 The master boot record format starts with the bootstrap code area.
 In the classic format it is the first 446 bytes.
 In the modern standard MBR format the first 446 bytes are mostly bootstrap code too with the exception of a handful bytes at offset 218 or so which are used for a timestamp or so.
 This section overlaps with the tar V7 linked file name field.
 In both formats these bytes can be zero without any issues, thankfully.</p>
 <p>This is great!
 This means we can put a tar header in the bootstrap code area of the MBR and have it be a valid tar header <em>and</em> MBR record at the same time.
 The protective MBR has one partition of type <code>0xEE</code> whose LBA starts at sector 1 and the number of LBAs should cover the whole disk, or be <code>0xFFFFFFFF</code> (maximum representable number in unsigned 32 bit).
 In practice this means we can get away with only touching byte offsets 446-453 and 510-511 for the protective MBR.
 The MBR does not have a checksum which also makes things easier.
 Using this I could create a disk image that parted and fdisk recognized as a GPT partitioned disk!
 With the caveat that they both reported that the backup GPT header was corrupt.
 I had just copied the primary GPT header to the end of the disk.
 It turns out that the alternate, or backup, GPT header should have the current LBA and backup LBA fields swapped (and the header crc32 recomputed).
 I updated the ocaml-gpt code so that it can <a href="https://github.com/mirage/ocaml-gpt/pull/14">marshal alternate GPT headers</a>.</p>
 <p>Finally we can produce GPT partitioned disks that can be inspected with tar utilities!</p>
 <pre><code>$ /usr/sbin/parted disk.img print
 WARNING: You are not superuser.  Watch out for permissions.
 Model:  (file)
 Disk /home/reynir/workspace/gptar/disk.img: 524kB
 Sector size (logical/physical): 512B/512B
 Partition Table: gpt
 Disk Flags: pmbr_boot
 Number  Start   End     Size   File system  Name              Flags
 1      17.4kB  19.5kB  2048B               Real tar archive  hidden
 </code></pre>
 <pre><code>$ tar -tvf disk.img
 ?r-------- 0/0           16896 1970-01-01 01:00 GPTAR unknown file type ‘G’
 -r-------- 0/0              14 1970-01-01 01:00 test.txt
 </code></pre>
 <p>The <a href="https://github.com/reynir/gptar">code</a> is freely available on GitHub.</p>
 <h2 id="future-work"><a class="anchor" aria-hidden="true" href="#future-work"></a>Future work</h2>
 <p>One thing that bothers me a bit is the dummy file <code>GPTAR</code>.
 By using the <code>G</code> link indicator GNU tar will print a warning about the unknown file type <code>G</code>,
 but it will still extract the dummy file when extracting the archive.
 I have been thinking about what tar header I could put in the MBR so tar utilities will skip the partition table but not try to extract the dummy file.
 Ideas I've had is to:</p>
 <ul>
 <li>Pretend it is a directory with a non-zero file size which is nonsense.
 I'm unsure what tar utilities would do in that case.
 I fear not all implementations will skip to the next header correctly as a non-zero directory is nonsense.
 I may give it a try and check how GNU tar, FreeBSD tar and ocaml-tar react.</li>
 <li>Say it is a PAX extended header and use a nonsense tag or attribute whose value covers the GPT header and partition table.
 The problem is the PAX extended header content format is <code>&lt;length&gt; &lt;tag&gt;=&lt;value&gt;\n</code> where <code>&lt;length&gt;</code> is the decimal string encoding of the length of <code>&lt;tag&gt;=&lt;value&gt;\n</code>.
 In other words it must start with the correct length.
 For sector size 512 this is a problem because the PAX extended header content would start with the GPT header which starts with the string <code>EFI PART</code>.
 If the sector size is greater than 512 we can use the remaining space in LBA 0 to write a length, dummy tag and some padding.
 I may try this for a sector size of 4096, but I'm not happy that it doesn't work with sector size 512 which solo5 will default to.</li>
 </ul>
 <p>If you have other ideas what I can do please reach out!</p>
 <p>[^tar-ustar]: This is somewhat simplified. There are some more nuances between the different formats, but for this purpose they don't matter much.</p>
    </article>
        </main>
    <footer>
        <a href="https://github.com/xhtmlboi/yocaml">Powered by <strong>YOCaml</strong></a>
        <br />
    </footer>
    <script>hljs.highlightAll();</script>
  </body>
 </html>
--- a/articles/lwt_pause.html
+++ b/articles/lwt_pause.html
@ -0,0 +1,263 @@
 <!doctype html>
 <html lang="en">
  <head>
    <meta charset="utf-8">
    <meta http-equiv="x-ua-compatible" content="ie=edge">
    <meta name="viewport" content="width=device-width, initial-scale=1">
    <title>
        Robur's blogCooperation and Lwt.pause
    </title>
    <meta name="description" content="A disgression about Lwt and Miou">
    <link type="text/css" rel="stylesheet" href="https://blog.robur.coop/css/hl.css">
    <link type="text/css" rel="stylesheet" href="https://blog.robur.coop/css/style.css">
    <script src="https://blog.robur.coop/js/hl.js"></script>
    <link rel="alternate" type="application/rss+xml" href="https://blog.robur.coop/feed.xml" title="blog.robur.coop">
  </head>
  <body>
    <header>
      <h1>blog.robur.coop</h1>
      <blockquote>
        The <strong>Robur</strong> cooperative blog.
      </blockquote>
    </header>
    <main><a href="https://blog.robur.coop/index.html">Back to index</a>
 <article>
    <h1>Cooperation and Lwt.pause</h1>
    <ul class="tags-list"><li>OCaml</li><li>Scheduler</li><li>Community</li><li>Unikernel</li><li>Git</li></ul><p>Here's a concrete example of the notion of availability and the scheduler used
 (in this case Lwt). As you may know, at Robur we have developed a unikernel:
 <a href="https://git.robur.coop/robur/opam-mirror">opam-mirror</a>. It launches an HTTP service that can be used as an
 OPAM overlay available from a Git repository (with <code>opam repository add &lt;name&gt; &lt;url&gt;</code>).</p>
 <p>The purpose of such an unikernel was to respond to a failure of the official
 repository which fortunately did not last long and to offer decentralisation
 of such a service. You can use https://opam.robur.coop!</p>
 <p>It was also useful at the Mirage retreat, where we don't usually have a
 great internet connection. Caching packages for our OCaml users on the local
 network has benefited us in terms of our Internet bill by allowing the OCaml
 users to fetch opam packages over the local network instead of over the shared,
 metered 4G Internet conncetion.</p>
 <p>Finally, it's a unikernel that I also use on my server for my software
 <a href="https://blog.osau.re/articles/reproducible.html">reproducibility service</a> in order to have an overlay for my
 software like <a href="https://bob.osau.re/">Bob</a>.</p>
 <p>In short, I advise you to use it, you can see its installation
 <a href="https://blog.osau.re/articles/reproducible.html">here</a> (I think that in the context of a company, internally, it
 can be interesting to have such a unikernel available).</p>
 <p>However, this unikernel had a long-standing problem. We were already talking
 about it at the Mirleft retreat, when we tried to get the repository from Git,
 we had a (fairly long) unavailability of our HTTP server. Basically, we had to
 wait ~10 min before the service offered by the unikernel was available.</p>
 <h2 id="availability"><a class="anchor" aria-hidden="true" href="#availability"></a>Availability</h2>
 <p>If you follow my <a href="https://blog.osau.re/tags/scheduler.html">articles</a>, as far as Miou is concerned, from
 the outset I talk of the notion of availability if we were to make yet another
 new scheduler for OCaml 5. We emphasised this notion because we had quite a few
 problems on this subject and Lwt.</p>
 <p>In this case, the notion of availability requires the scheduler to be able to
 observe system events as often as possible. The problem is that Lwt doesn't
 really offer this approach.</p>
 <p>Indeed, Lwt offers a way of observing system events (<code>Lwt.pause</code>) but does not
 do so systematically. The only time you really give the scheduler the
 opportunity to see whether you can read or write is when you want to...
 read or write...</p>
 <p>More generally, it is said that Lwt's <strong>bind</strong> does not <em>yield</em>. In other words,
 you can chain any number of functions together (via the <code>&gt;&gt;=</code> operator), but
 from Lwt's point of view, there is no opportunity to see if an event has
 occurred. Lwt always tries to go as far down your chain as possible:</p>
 <ul>
 <li>and finish your promise</li>
 <li>or come across an operation that requires a system event (read or write)</li>
 <li>or come across an <code>Lwt.pause</code> (as a <em>yield</em> point)</li>
 </ul>
 <p>Lwt is rather sparse in adding cooperation points besides <code>Lwt.pause</code> and
 read/write operations, in contrast with Async where the bind operator is a
 cooperation point.</p>
 <h3 id="if-there-is-no-io-do-not-wrap-in-lwt"><a class="anchor" aria-hidden="true" href="#if-there-is-no-io-do-not-wrap-in-lwt"></a>If there is no I/O, do not wrap in Lwt</h3>
 <p>It was (bad<sup><a href="#fn1">1</a></sup>) advice I was given. If a function doesn't do
 I/O, there's no point in putting it in Lwt. At first glance, however, the idea
 may be a good one. If you have a function that doesn't do I/O, whether it's in
 the Lwt monad or not won't make any difference to the way Lwt tries to execute
 it. Once again, Lwt should go as far as possible. So Lwt tries to solve both
 functions in the same way:</p>
 <pre><code class="language-ocaml">val merge : int array -&gt; int array -&gt; int array
 let rec sort0 arr =
  if Array.length arr &lt;= 1 then arr
  else
    let m = Array.length arr / 2 in
    let arr0 = sort0 (Array.sub arr 0 m) in
    let arr1 = sort0 (Array.sub arr m (Array.length arr - m)) in
    merge arr0 arr1
 let rec sort1 arr =
  let open Lwt.Infix in
  if Array.length arr &lt;= 1 then Lwt.return arr
  else
    let m = Array.length arr / 2 in
    Lwt.both
      (sort1 (Array.sub arr m (Array.length arr - m)))
      (sort1 (Array.sub arr 0 m))
    &gt;|= fun (arr0, arr1) -&gt;
    merge arr0 arr1
 </code></pre>
 <p>If we trace the execution of the two functions (for example, by displaying our
 <code>arr</code> each time), we see the same behaviour whether Lwt is used or not. However,
 what is interesting in the Lwt code is the use of <code>both</code>, which suggests that
 the processes are running <em>at the same time</em>.</p>
 <p>&quot;At the same time&quot; does not necessarily suggest the use of several cores or &quot;in
 parallel&quot;, but the possibility that the right-hand side may also have the
 opportunity to be executed even if the left-hand side has not finished. In other
 words, that the two processes can run <strong>concurrently</strong>.</p>
 <p>But factually, this is not the case, because even if we had the possibility of
 a point of cooperation (with the <code>&gt;|=</code> operator), Lwt tries to go as far as
 possible and decides to finish the left part before launching the right part:</p>
 <pre><code class="language-shell">$ ./a.out
 sort0: [|3; 4; 2; 1; 7; 5; 8; 9; 0; 6|]
 sort0: [|3; 4; 2; 1; 7|]
 sort0: [|3; 4|]
 sort0: [|2; 1; 7|]
 sort0: [|1; 7|]
 sort0: [|5; 8; 9; 0; 6|]
 sort0: [|5; 8|]
 sort0: [|9; 0; 6|]
 sort0: [|0; 6|]
 sort1: [|3; 4; 2; 1; 7; 5; 8; 9; 0; 6|]
 sort1: [|3; 4; 2; 1; 7|]
 sort1: [|3; 4|]
 sort1: [|2; 1; 7|]
 sort1: [|1; 7|]
 sort1: [|5; 8; 9; 0; 6|]
 sort1: [|5; 8|]
 sort1: [|9; 0; 6|]
 sort1: [|0; 6|]
 </code></pre>
 <hr>
 <p><strong><tag id="fn1">1</tag></strong>: However, if you are not interested in availability
 and would like the scheduler to try to resolve your promises as quickly as
 possible, this advice is clearly valid.</p>
 <h4 id="performances"><a class="anchor" aria-hidden="true" href="#performances"></a>Performances</h4>
 <p>It should be noted, however, that Lwt has an impact. Even if the behaviour is
 the same, the Lwt layer is not free. A quick benchmark shows that there is an
 overhead:</p>
 <pre><code class="language-ocaml">let _ =
  let t0 = Unix.gettimeofday () in
  for i = 0 to 1000 do let _ = sort0 arr in () done;
  let t1 = Unix.gettimeofday () in
  Fmt.pr &quot;sort0 %fs\n%!&quot; (t1 -. t0)
 let _ =
  let t0 = Unix.gettimeofday () in
  Lwt_main.run @@ begin
    let open Lwt.Infix in
    let rec go idx = if idx = 1000 then Lwt.return_unit
      else sort1 arr &gt;&gt;= fun _ -&gt; go (succ idx) in
    go 0 end;
  let t1 = Unix.gettimeofday () in
  Fmt.pr &quot;sort1 %fs\n%!&quot; (t1 -. t0)
 </code></pre>
 <pre><code class="language-sh">$ ./a.out
 sort0 0.000264s
 sort1 0.000676s
 </code></pre>
 <p>This is the fairly obvious argument for not using Lwt when there's no I/O. Then,
 if the Lwt monad is really needed, a simple <code>Lwt.return</code> at the very last
 instance is sufficient (or, better, the use of <code>Lwt.map</code> / <code>&gt;|=</code>).</p>
 <h4 id="cooperation-and-concrete-example"><a class="anchor" aria-hidden="true" href="#cooperation-and-concrete-example"></a>Cooperation and concrete example</h4>
 <p>So <code>Lwt.both</code> is the one to use when we want to run two processes
 &quot;at the same time&quot;. For the example, <a href="https://github.com/mirage/ocaml-git">ocaml-git</a> attempts <em>both</em> to
 retrieve a repository and also to analyse it. This can be seen in this snippet
 of <a href="https://github.com/mirage/ocaml-git/blob/a36c90404b149ab85f429439af8785bb1dde1bee/src/not-so-smart/smart_git.ml#L476-L481">code</a>.</p>
 <p>In our example with ocaml-git, the problem &quot;shouldn't&quot; appear because, in this
 case, both the left and right side do I/O (the left side binds into a socket
 while the right side saves Git objects in your file system). So, in our tests
 with <code>Git_unix</code>, we were able to see that the analysis (right-hand side) was
 well executed and 'interleaved' with the reception of objects from the network.</p>
 <h3 id="composability"><a class="anchor" aria-hidden="true" href="#composability"></a>Composability</h3>
 <p>However, if we go back to our initial problem, we were talking about our
 opam-mirror unikernel. As you might expect, there is no standalone MirageOS file
 system (and many of our unikernels don't need one). So, in the case of
 opam-mirror, we use the ocaml-git memory implementation: <code>Git_mem</code>.</p>
 <p><code>Git_mem</code> is different in that Git objects are simply stored in a <code>Hashtbl</code>.
 There is no cooperation point when it comes to obtaining Git objects from this
 <code>Hashtbl</code>. So let's return to our original advice:</p>
 <blockquote>
 <p>don't wrap code in Lwt if it doesn't do I/O.</p>
 </blockquote>
 <p>And, of course, <code>Git_mem</code> doesn't do I/O. It does, however, require the process
 to be able to work with Lwt. In this case, <code>Git_mem</code> wraps the results in Lwt
 <strong>as late as possible</strong> (as explained above, so as not to slow down our
 processes unnecessarily). The choice inevitably means that the right-hand side
 can no longer offer cooperation points. And this is where our problem begins:
 composition.</p>
 <p>In fact, we had something like:</p>
 <pre><code class="language-ocaml">let clone socket git =
  Lwt.both (receive_pack socket) (analyse_pack git) &gt;&gt;= fun ((), ()) -&gt;
  Lwt.return_unit
 </code></pre>
 <p>However, our <code>analyse_pack</code> function is an injection of a functor representing
 the Git backend. In other words, <code>Git_unix</code> or <code>Git_mem</code>:</p>
 <pre><code class="language-ocaml">module Make (Git : Git.S) = struct
  let clone socket git =
    Lwt.both (receive_pack socket) (Git.analyse_pack git) &gt;&gt;= fun ((), ()) -&gt;
    Lwt.return_unit
 end
 </code></pre>
 <p>Composability poses a problem here because even if <code>Git_unix</code> and <code>Git_mem</code>
 offer the same function (so both modules can be used), the fact remains that one
 will always offer a certain availability to other services (such as an HTTP
 service) while the other will offer a Lwt function which will try to go as far
 as possible quite to make other services unavailable.</p>
 <p>Composing with one or the other therefore does not produce the same behavior.</p>
 <h4 id="where-to-put-lwtpause"><a class="anchor" aria-hidden="true" href="#where-to-put-lwtpause"></a>Where to put <code>Lwt.pause</code>?</h4>
 <p>In this case, our <code>analyse_pack</code> does read/write on the Git store. As far as
 <code>Git_mem</code> is concerned, we said that these read/write accesses were just
 accesses to a <code>Hashtbl</code>.</p>
 <p>Thanks to <a href="https://hannes.robur.coop/">Hannes</a>' help, it took us an afternoon to work out where we
 needed to add cooperation points in <code>Git_mem</code> so that <code>analyse_pack</code> could give
 another service such as HTTP the opportunity to work. Basically, this series of
 <a href="https://github.com/mirage/ocaml-git/pull/631/files">commits</a> shows where we needed to add <code>Lwt.pause</code>.</p>
 <p>However, this points to a number of problems:</p>
 <ol>
 <li>it is not necessarily true that on the basis of composability alone (by
 <em>functor</em> or by value), Lwt reacts in the same way</li>
 <li>Subtly, you have to dig into the code to find the right opportunities where
 to put, by hand, <code>Lwt.pause</code>.</li>
 <li>In the end, Lwt has no mechanisms for ensuring the availability of a service
 (this is something that must be taken into account by the implementer).</li>
 </ol>
 <h3 id="in-depth-knowledge-of-lwt"><a class="anchor" aria-hidden="true" href="#in-depth-knowledge-of-lwt"></a>In-depth knowledge of Lwt</h3>
 <p>I haven't mentioned another problem we encountered with <a href="https://cambium.inria.fr/~agueneau/">Armael</a> when
 implementing <a href="https://discuss.ocaml.org/t/ann-release-of-multipart-form-0-2-0/7704#memory-bound-implementation">multipart_form</a> where the use of stream meant that
 Lwt didn't interleave the two processes and the use of a <em>bounded stream</em> was
 required. Again, even when it comes to I/O, Lwt always tries to go as far as
 possible in one of two branches of a <code>Lwt.both</code>.</p>
 <p>This allows us to conclude that beyond the monad, Lwt has subtleties in its
 behaviour which may be different from another scheduler such as Async (hence the
 incompatibility between the two, which is not just of the <code>'a t</code> type).</p>
 <h3 id="digression-on-miou"><a class="anchor" aria-hidden="true" href="#digression-on-miou"></a>Digression on Miou</h3>
 <p>That's why we put so much emphasis on the notion of availability when it comes
 to Miou: to avoid repeating the mistakes of the past. The choices that can be
 made with regard to this notion in particular have a major impact, and can be
 unsatisfactory to the user in certain cases (for example, so-called pure
 calculations could take longer with Miou than with another scheduler).</p>
 <p>In this sense, we have tried to constrain ourselves in the development of Miou
 through the use of <code>Effect.Shallow</code> which requires us to always re-attach our
 handler (our scheduler) as soon as an effect is produced, unlike <code>Effect.Deep</code>
 which can re-use the same handler for several effects. In other words, and as
 we've described here, <strong>an effect yields</strong>!</p>
 <h2 id="conclusion"><a class="anchor" aria-hidden="true" href="#conclusion"></a>Conclusion</h2>
 <p>As far as opam-mirror is concerned, we now have an unikernel that is available
 even if it attempts to clone a Git repository and save Git objects in memory. At
 least, an HTTP service can co-exist with ocaml-git!</p>
 <p>I hope we'll be able to use it at <a href="https://retreat.mirage.io/">the next retreat</a>, which I invite
 you to attend to talk more about Lwt, scheduler, Git and unikernels!</p>
    </article>
        </main>
    <footer>
        <a href="https://github.com/xhtmlboi/yocaml">Powered by <strong>YOCaml</strong></a>
        <br />
    </footer>
    <script>hljs.highlightAll();</script>
  </body>
 </html>
--- a/articles/miragevpn-ncp.html
+++ b/articles/miragevpn-ncp.html
@ -0,0 +1,58 @@
 <!doctype html>
 <html lang="en">
  <head>
    <meta charset="utf-8">
    <meta http-equiv="x-ua-compatible" content="ie=edge">
    <meta name="viewport" content="width=device-width, initial-scale=1">
    <title>
        Robur's blogMirageVPN updated (AEAD, NCP)
    </title>
    <meta name="description" content="How we resurrected MirageVPN from its bitrot state">
    <link type="text/css" rel="stylesheet" href="https://blog.robur.coop/css/hl.css">
    <link type="text/css" rel="stylesheet" href="https://blog.robur.coop/css/style.css">
    <script src="https://blog.robur.coop/js/hl.js"></script>
    <link rel="alternate" type="application/rss+xml" href="https://blog.robur.coop/feed.xml" title="blog.robur.coop">
  </head>
  <body>
    <header>
      <h1>blog.robur.coop</h1>
      <blockquote>
        The <strong>Robur</strong> cooperative blog.
      </blockquote>
    </header>
    <main><a href="https://blog.robur.coop/index.html">Back to index</a>
 <article>
    <h1>MirageVPN updated (AEAD, NCP)</h1>
    <ul class="tags-list"><li>OCaml</li><li>MirageOS</li><li>VPN</li><li>security</li></ul><h2 id="updating-miragevpn"><a class="anchor" aria-hidden="true" href="#updating-miragevpn"></a>Updating MirageVPN</h2>
 <p>As announced <a href="https://blog.robur.coop/articles/miragevpn.html">earlier this month</a>, we've been working hard over the last months on MirageVPN (initially developed in 2019, targeting OpenVPN™ 2.4.7, now 2.6.6). We managed to receive funding from <a href="https://www.assure.ngi.eu/">NGI Assure</a> call (via <a href="https://nlnet.nl">NLnet</a>). We've made over 250 commits with more than 10k lines added, and 18k lines removed. We closed nearly all old issues, and opened 100 fresh ones, of which we already closed more than half of them. :D</p>
 <h3 id="actual-bugs-fixed-that-were-leading-to-non-working-miragevpn-applications"><a class="anchor" aria-hidden="true" href="#actual-bugs-fixed-that-were-leading-to-non-working-miragevpn-applications"></a>Actual bugs fixed (that were leading to non-working MirageVPN applications)</h3>
 <p>In more detail, we had a specific configuration running over all the years, namely UDP mode with static keys (no TLS handshake, etc.). There were several issues (bitrot) that we encountered and solved along the path, amongst others:</p>
 <ul>
 <li>related to the <a href="https://github.com/robur-coop/miragevpn/pull/111">static-key mode and TCP/IP</a>,</li>
 <li>the <a href="https://github.com/robur-coop/miragevpn/pull/98">order of ACK between the client and the server</a>,</li>
 <li><a href="https://github.com/robur-coop/miragevpn/pull/110">outgoing TLS packets</a>.</li>
 </ul>
 <p>To avoid any future breakage while revising the code (cleaning it up, extending it), we are now building several unikernels as part of our CI system. We also have setup OpenVPN™ servers with various configurations that we periodically test with our new code (we'll also work on further automation thereof).</p>
 <h3 id="new-features-aead-ciphers-supporting-more-configuration-primitives"><a class="anchor" aria-hidden="true" href="#new-features-aead-ciphers-supporting-more-configuration-primitives"></a>New features: AEAD ciphers, supporting more configuration primitives</h3>
 <p>We added various configuration primitives, amongst them configuratble tls ciphersuites, minimal and maximal tls version to use, <a href="https://blog.robur.coop/articles/miragevpn.html">tls-crypt-v2</a>, verify-x509-name, cipher,  remote-random, ...</p>
 <p>From a cryptographic point of view, we are now supporting more <a href="https://github.com/robur-coop/miragevpn/pull/108">authentication hashes</a> via the configuration directive <code>auth</code>, namely the SHA2 family - previously, only SHA1 was supported, <a href="https://github.com/robur-coop/miragevpn/pull/125">AEAD ciphers</a> (AES-128-GCM, AES-256-GCM, CHACHA20-POLY1305) - previously only AES-256-CBC was supported.</p>
 <h3 id="ncp---negotiation-of-cryptographic-parameters"><a class="anchor" aria-hidden="true" href="#ncp---negotiation-of-cryptographic-parameters"></a>NCP - Negotiation of cryptographic parameters</h3>
 <p>OpenVPN™ has a way to negotiate cryptographic parameters, instead of hardcoding them in the configuration. The client can propose its supported ciphers, and other features (MTU, directly request a push message for IP configuration, use TLS exporter secret instead of the hand-crafted (TLS 1.0 based PRF), ...) once the TLS handshake has been completed.</p>
 <p>We are now supporting this negotiation protocol, and have been working on the different extensions that are useful to us. Namely, transmitting the <a href="https://github.com/robur-coop/miragevpn/pull/121">supported ciphers</a>, <a href="https://github.com/robur-coop/miragevpn/pull/129">request push</a> (which deletes an entire round-trip), <a href="https://github.com/robur-coop/miragevpn/pull/163">TLS-exporter</a>. This will also be part of the <a href="https://git.robur.coop/robur/openvpn-spec">protocol specification</a> that we're working on while finishing our implementation.</p>
 <h3 id="cleanups-and-refactorings"><a class="anchor" aria-hidden="true" href="#cleanups-and-refactorings"></a>Cleanups and refactorings</h3>
 <p>We also took some time to cleanup our code base, removing <code>Lwt.fail</code> (which doesn't produce proper backtraces), using lzo from the decompress package (since that code has been upstreamed a while ago), remove unneeded dependencies (rresult, astring), avoiding <code>assert false</code> in pattern matches by improving types, improve the log output (include a timestamp, show log source, use colors).</p>
 <h3 id="future"><a class="anchor" aria-hidden="true" href="#future"></a>Future</h3>
 <p>There is still some work that we want to do, namely a QubesOS client implementation, an operators manual, extending our specification, resurrecting and adapting the server implementation, supporting more NCP features (if appropriate), etc. So stay tuned, we'll also provide reproducible binaries once we're ready.</p>
 <p>Don't hesitate to reach out to us on <a href="https://github.com/robur-coop/miragevpn/issues">GitHub</a>, <a href="https://robur.coop/Contact">by mail</a> or me personally <a href="https://mastodon.social/@hannesm">on Mastodon</a> if you're stuck.</p>
    </article>
        </main>
    <footer>
        <a href="https://github.com/xhtmlboi/yocaml">Powered by <strong>YOCaml</strong></a>
        <br />
    </footer>
    <script>hljs.highlightAll();</script>
  </body>
 </html>
--- a/articles/miragevpn-performance.html
+++ b/articles/miragevpn-performance.html
@ -0,0 +1,57 @@
 <!doctype html>
 <html lang="en">
  <head>
    <meta charset="utf-8">
    <meta http-equiv="x-ua-compatible" content="ie=edge">
    <meta name="viewport" content="width=device-width, initial-scale=1">
    <title>
        Robur's blogSpeeding up MirageVPN and use it in the wild
    </title>
    <meta name="description" content="Performance engineering of MirageVPN, speeding it up by a factor of 25.">
    <link type="text/css" rel="stylesheet" href="https://blog.robur.coop/css/hl.css">
    <link type="text/css" rel="stylesheet" href="https://blog.robur.coop/css/style.css">
    <script src="https://blog.robur.coop/js/hl.js"></script>
    <link rel="alternate" type="application/rss+xml" href="https://blog.robur.coop/feed.xml" title="blog.robur.coop">
  </head>
  <body>
    <header>
      <h1>blog.robur.coop</h1>
      <blockquote>
        The <strong>Robur</strong> cooperative blog.
      </blockquote>
    </header>
    <main><a href="https://blog.robur.coop/index.html">Back to index</a>
 <article>
    <h1>Speeding up MirageVPN and use it in the wild</h1>
    <ul class="tags-list"><li>OCaml</li><li>MirageOS</li><li>cryptography</li><li>security</li><li>VPN</li><li>performance</li></ul><p>As we were busy continuing to work on <a href="https://github.com/robur-coop/miragevpn">MirageVPN</a>, we got in touch with <a href="https://eduvpn.org">eduVPN</a>, who are interested about deploying MirageVPN. We got example configuration from their side, and <a href="https://github.com/robur-coop/miragevpn/pull/201">fixed</a> <a href="https://github.com/robur-coop/miragevpn/pull/168">some</a> <a href="https://github.com/robur-coop/miragevpn/pull/202">issues</a>, and also implemented <a href="https://github.com/robur-coop/miragevpn/pull/169">tls-crypt</a> - which was straightforward since we earlier spend time to implement <a href="https://blog.robur.coop/articles/miragevpn.html">tls-crypt-v2</a>.</p>
 <p>In January, they gave MirageVPN another try, and <a href="https://github.com/robur-coop/miragevpn/issues/206">measured the performance</a> -- which was very poor -- MirageVPN (run as a Unix binary) provided a bandwith of 9.3Mb/s, while OpenVPN provided a bandwidth of 360Mb/s (using a VPN tunnel over TCP).</p>
 <p>We aim at spending less resources for computing, thus the result was not satisfying for us. We re-read a lot of code, refactored a lot, and are now at ~250Mb/s.</p>
 <h2 id="tooling-for-performance-engineering-of-ocaml"><a class="anchor" aria-hidden="true" href="#tooling-for-performance-engineering-of-ocaml"></a>Tooling for performance engineering of OCaml</h2>
 <p>As a first approach we connected with the MirageVPN unix client &amp; OpenVPN client to a eduVPN server and ran speed tests using <a href="https://fast.com">fast.com</a>. This was sufficient to show the initial huge gap in download speeds between MirageVPN and OpenVPN. There is <em>a lot</em> of noise in this approach as there are many computers and routers involved in this setup, and it is hard to reproduce.</p>
 <p>To get more reproducible results we set up a local VM with openvpn and iperf3 installed. On the host machine we can then connect to the VM's OpenVPN server and run iperf3 against the <em>VPN</em> ip address. This worked more reliably. However, it was still noisy and not suitable to measure single digit percentage changes in performance.
 To better guide the performance engineering, we also developed <a href="https://github.com/robur-coop/miragevpn/pull/230">a microbenchmark</a> using OCaml tooling. This will setup a client and server without any input and output, and transfer data in memory.</p>
 <p>We also re-read our code and used the Linux utility <a href="https://perf.wiki.kernel.org/index.php/Main_Page"><code>perf</code></a> together with <a href="https://github.com/brendangregg/FlameGraph">Flamegraph</a> to graph its output. This works nicely with OCaml programs (we're using the 4.14.1 compiler and runtime system). We did the performance engineering on Unix binaries, i.e. not on MirageOS unikernels - but the MirageVPN protocol is used in both scenarios - thus the performance improvements described here are also in the MirageVPN unikernels.</p>
 <h2 id="takeaway-of-performance-engineering"><a class="anchor" aria-hidden="true" href="#takeaway-of-performance-engineering"></a>Takeaway of performance engineering</h2>
 <p>The learnings of our performance engineering are in three areas:</p>
 <ul>
 <li>Formatting strings is computational expensive -- thus if in an error case a hexdump is produced of a packet, its construction must be delayed for when the error case is executed (we have <a href="https://github.com/robur-coop/miragevpn/pull/220">this PR</a> and <a href="https://github.com/robur-coop/miragevpn/pull/209">that PR</a>). Alain Frisch wrote a nice <a href="https://www.lexifi.com/blog/ocaml/note-about-performance-printf-and-format/#">blog post</a> at LexiFi about performance of <code>Printf</code> and <code>Format</code>[^lexifi-date].</li>
 <li>Rethink allocations: fundamentally, only a single big buffer (to be send out) for each incoming packet should be allocated, not a series of buffers that are concatenated (see <a href="https://github.com/robur-coop/miragevpn/pull/217">this PR</a> and <a href="https://github.com/robur-coop/miragevpn/pull/219">that PR</a>). Additionally, not zeroing out the just allocated buffer (if it is filled with data anyways) removes some further instructions (see <a href="https://github.com/robur-coop/miragevpn/pull/218">this PR</a>). And we figured that appending to an empty buffer nevertheless allocated and copied in OCaml, so we worked on <a href="https://github.com/robur-coop/miragevpn/pull/214">this PR</a>.</li>
 <li>Still an open topic is: we are in the memory-safe language OCaml, and we sometimes extract data out of a buffer (or set data in a buffer). Now, each operation lead to bounds checks (that we do not touch memory that is not allocated or not ours). However, if we just checked for the buffer being long enough (either by checking the length, or by allocating a specific amount of data), these bounds checks are superfluous. So far, we don't have an automated solution for this issue, but we are <a href="https://discuss.ocaml.org/t/bounds-checks-for-string-and-bytes-when-retrieving-or-setting-subparts-thereof/">discussing it in the OCaml community</a>, and are eager to find a solution to avoid unneeded computations.</li>
 </ul>
 <h2 id="conclusion"><a class="anchor" aria-hidden="true" href="#conclusion"></a>Conclusion</h2>
 <p>To conclude: we already achieved a factor of 25 in performance by adapting the code in various ways. We have ideas to improve the performance even more in the future - we also work on using OCaml string and bytes, instead of off-the-OCaml-heap-allocated bigarrays (see <a href="https://blog.robur.coop/articles/speeding-ec-string.html">our previous article</a>, which provided some speedups).</p>
 <p>Don't hesitate to reach out to us on <a href="https://github.com/robur-coop/miragevpn/issues">GitHub</a>, or <a href="https://robur.coop/Contact">by mail</a> if you're stuck.</p>
 <p>We want to thank <a href="https://nlnet.nl">NLnet</a> for their funding (via <a href="https://www.assure.ngi.eu/">NGI assure</a>), and <a href="https://eduvpn.org">eduVPN</a> for their interest.</p>
 <p>[^lexifi-date]: It has come to our attention that the blog post is rather old (2012) and that the implementation has been completely rewritten since then.</p>
    </article>
        </main>
    <footer>
        <a href="https://github.com/xhtmlboi/yocaml">Powered by <strong>YOCaml</strong></a>
        <br />
    </footer>
    <script>hljs.highlightAll();</script>
  </body>
 </html>
--- a/articles/miragevpn-server.html
+++ b/articles/miragevpn-server.html
@ -0,0 +1,46 @@
 <!doctype html>
 <html lang="en">
  <head>
    <meta charset="utf-8">
    <meta http-equiv="x-ua-compatible" content="ie=edge">
    <meta name="viewport" content="width=device-width, initial-scale=1">
    <title>
        Robur's blogMirageVPN server
    </title>
    <meta name="description" content="Announcement of our MirageVPN server.">
    <link type="text/css" rel="stylesheet" href="https://blog.robur.coop/css/hl.css">
    <link type="text/css" rel="stylesheet" href="https://blog.robur.coop/css/style.css">
    <script src="https://blog.robur.coop/js/hl.js"></script>
    <link rel="alternate" type="application/rss+xml" href="https://blog.robur.coop/feed.xml" title="blog.robur.coop">
  </head>
  <body>
    <header>
      <h1>blog.robur.coop</h1>
      <blockquote>
        The <strong>Robur</strong> cooperative blog.
      </blockquote>
    </header>
    <main><a href="https://blog.robur.coop/index.html">Back to index</a>
 <article>
    <h1>MirageVPN server</h1>
    <ul class="tags-list"><li>OCaml</li><li>MirageOS</li><li>cryptography</li><li>security</li><li>VPN</li></ul><p>It is a great pleasure to finally announce that we have finished a server implementation for MirageVPN (OpenVPN™-compatible). This allows to setup a very robust VPN network on both the client and the server side.</p>
 <p>As announced last year, <a href="https://blog.robur.coop/articles/miragevpn.html">MirageVPN</a> is a reimplemtation of OpenVPN™ in OCaml, with <a href="https://mirage.io">MirageOS</a> unikernels.</p>
 <h2 id="why-a-miragevpn-server"><a class="anchor" aria-hidden="true" href="#why-a-miragevpn-server"></a>Why a MirageVPN server?</h2>
 <p>Providing Internet services with programming languages that have not much safety requires a lot of discipline by the developers to avoid issues which may lead to exploitable services that are attacked (and thus will circumvent any security goals). Especially services that are critical for security and privacy, it is crucial to avoid common memory safety pitfalls.</p>
 <p>Some years back, when we worked on the client implementation, we also drafted a server implementation. The reasoning was that a lot of the code was already there, and just a few things needed to be developed to allow clients to connect there.</p>
 <p>Now, we spend several months to push our server implementation into a state where it is usable and we are happy for everyone who wants to test it. We also adapted the modern ciphers we recently implemented for the client, and also tls-crypt and tls-crypt-v2 for the server implementation.</p>
 <p>The overall progress was tracked in <a href="https://github.com/robur-coop/miragevpn/issues/15">this issue</a>. We developed, next to the MirageOS unikernel, also a test server that doesn't use any tun interface.</p>
 <p>Please move along to our handbook with the <a href="https://robur-coop.github.io/miragevpn-handbook/miragevpn_server.html">chapter on MirageVPN server</a>.</p>
 <p>If you encounter any issues, please open an issue at <a href="https://github.com/robur-coop/miragevpn">the repository</a>.</p>
    </article>
        </main>
    <footer>
        <a href="https://github.com/xhtmlboi/yocaml">Powered by <strong>YOCaml</strong></a>
        <br />
    </footer>
    <script>hljs.highlightAll();</script>
  </body>
 </html>
--- a/articles/miragevpn.html
+++ b/articles/miragevpn.html
@ -0,0 +1,120 @@
 <!doctype html>
 <html lang="en">
  <head>
    <meta charset="utf-8">
    <meta http-equiv="x-ua-compatible" content="ie=edge">
    <meta name="viewport" content="width=device-width, initial-scale=1">
    <title>
        Robur's blogMirageVPN &amp; tls-crypt-v2
    </title>
    <meta name="description" content="How we implementated tls-crypt-v2 for miragevpn">
    <link type="text/css" rel="stylesheet" href="https://blog.robur.coop/css/hl.css">
    <link type="text/css" rel="stylesheet" href="https://blog.robur.coop/css/style.css">
    <script src="https://blog.robur.coop/js/hl.js"></script>
    <link rel="alternate" type="application/rss+xml" href="https://blog.robur.coop/feed.xml" title="blog.robur.coop">
  </head>
  <body>
    <header>
      <h1>blog.robur.coop</h1>
      <blockquote>
        The <strong>Robur</strong> cooperative blog.
      </blockquote>
    </header>
    <main><a href="https://blog.robur.coop/index.html">Back to index</a>
 <article>
    <h1>MirageVPN &amp; tls-crypt-v2</h1>
    <ul class="tags-list"><li>OCaml</li><li>MirageOS</li><li>VPN</li><li>security</li></ul><p>In 2019 <a href="https://robur.coop/">Robur</a> started working on a <a href="https://github.com/robur-coop/miragevpn/">OpenVPN™-compatible implementation in OCaml</a>.
 The project was funded for 6 months in 2019 by <a href="https://prototypefund.de">prototypefund</a>.
 In late 2022 we applied again for funding this time to the <a href="https://www.assure.ngi.eu/">NGI Assure</a> open call, and our application was eventually accepted.
 In this blog post I will explain why reimplementing the OpenVPN™ protocol in OCaml is a worthwhile effort, and describe the Miragevpn implementation and in particular the <code>tls-crypt-v2</code> mechanism.</p>
 <h2 id="what-even-is-openvpn™"><a class="anchor" aria-hidden="true" href="#what-even-is-openvpn™"></a>What even is OpenVPN™?</h2>
 <p><a href="https://openvpn.net/">OpenVPN™</a> is a protocol and software implementation to provide <a href="https://en.wikipedia.org/wiki/Virtual_private_network">virtual private networks</a>: computer networks that do not exist in hardware and are encrypted and tunnelled through existing networks.
 Common use cases for this is to provide access to internal networks for remote workers, and for routing internet traffic through another machine for various reasons e.g. when using untrusted wifi, privacy from a snooping ISP, circumventing geoblock etc.</p>
 <p>It is a protocol that has been worked on and evolved over the decades.
 OpenVPN™ has a number of modes of operations as well as a number of options in the order of hundreds.
 The modes can be categorized into two main categories: static mode and TLS mode.
 The former mode uses static symmetric keys, and will be removed in the upcoming OpenVPN™ 2.7 (community edition).
 I will not focus on static mode in this post.
 The latter uses separate data &amp; control channels where the control channel uses TLS - more on that later.</p>
 <h3 id="why-reimplement-it-and-why-in-ocaml"><a class="anchor" aria-hidden="true" href="#why-reimplement-it-and-why-in-ocaml"></a>Why reimplement it? And why in OCaml?</h3>
 <p>Before diving into TLS mode and eventually tls-crypt-v2 it's worth to briefly discuss why we spend time reimplementing the OpenVPN™ protocol.
 You may ask yourself: why not just use the existing tried and tested implementation?</p>
 <p>OpenVPN™ community edition is implemented in the C programming language.
 It heavily uses the OpenSSL library[^mbedtls] which is as well written in C and has in the past had some notable security vulnerabilities.
 Many vulnerabilities and bugs in C can be easily avoided in other languages due to bounds checking and stricter and more expressive type systems.
 The state machine of the protocol can be more easily be expressed in OCaml, and some properties of the protocol can be encoded in the type system.</p>
 <p>Another reason is <a href="https://mirage.io/">Mirage OS</a>, a library operating system implemented in OCaml.
 We work on the Mirage project and write applications (unikernels) using Mirage.
 In many cases it would be desirable to be able to connect to an existing VPN network[^vpn-network],
 or be able to offer a VPN network to clients using OpenVPN™.</p>
 <p>Consider a VPN provider:
 The VPN provider runs many machines that run an operating system in order to run the user-space OpenVPN™ service.
 There are no <em>real</em> users on the system, and a lot of unrelated processes and legacy layers are around that are not needed.
 With a Mirage OS unikernel, which is basically a statically linked binary and operating system such a setup becomes simpler with fewer layers.
 With <a href="https://robur.coop/Projects/Reproducible_builds">reproducible builds</a> deployment and updates will be straightforward.</p>
 <p>Another very interesting example is a unikernel for <a href="https://www.qubes-os.org/">Qubes OS</a> that we have planned.
 Qubes OS is an operating system with a high focus on security.
 It offers an almost seamless experience of running applications in different virtual machines on the same machine.
 The networking provided to a application (virtual machine) can be restricted to only go through the VPN.
 It is possible to use OpenVPN™ for such a setup, but that requires running OpenVPN™ in a full Linux virtual machine.
 With Mirage OS the resource footprint is typically much smaller than an equivalent application running in a Linux virtual machine; often the memory footprint is smaller by an order.</p>
 <p>Finally, while it's not an explicit goal of ours, reimplementing a protocol without an explicit specification can help uncover bugs and things that need better documentation in the original implementation.</p>
 <h3 id="tls-mode"><a class="anchor" aria-hidden="true" href="#tls-mode"></a>TLS mode</h3>
 <p>There are different variants of TLS mode, but what they share is separate &quot;control&quot; channel and &quot;data&quot; channel.
 The control channel is used to do a TLS handshake, and with the established TLS session data channel encryption keys, username/password authentication, etc. is negotiated.
 Once this dance has been performed and data channel encryption keys have been negotiated the peers can exchange IP packets over the data channel.</p>
 <p>Over the years a number of mechanisms has been implemented to protect the TLS stack from being exposed to third parties, protect against denial of service attacks and to hide information exchanged during a TLS handshake such as certificates (which was an isue before TLS 1.3).
 These are known as <code>tls-auth</code>, <code>tls-crypt</code> and <code>tls-crypt-v2</code>.
 The <code>tls-auth</code> mechanism adds a pre-shared key for hmac authentication on the control channel.
 This makes it possible for an OpenVPN™ server to reject early clients that don't know the shared key before any TLS handshakes are performed.
 In <code>tls-crypt</code> the control channel is encrypted as well as hmac authenticated using a pre-shared key.
 Common to both is that the pre-shared key is shared between the server and all clients.
 For large deployments this significantly reduces the usefulness - the key is more likely to be leaked the greate the number of clients who share this key.</p>
 <h3 id="tls-crypt-v2"><a class="anchor" aria-hidden="true" href="#tls-crypt-v2"></a>tls-crypt-v2</h3>
 <p>To improve on <code>tls-crypt</code>, <code>tls-crypt-v2</code> uses one pre-shared key per client.
 This could be a lot of keys for the server to keep track of, so instead of storing all the client keys on the server the server has a special tls-crypt-v2 server key that is used to <em><a href="https://en.wikipedia.org/wiki/Key_wrap">wrap</a></em> the client keys.
 That is, each client has their own client key as well as the client key wrapped using the server key.
 The protocol is then extended so the client in the first message appends the wrapped key <em>unencrypted</em>.
 The server can then decrypt and verify the client key and decrypt the rest of the packet.
 Then the client and server use the client key just as in <code>tls-crypt</code>.</p>
 <p>This is great!
 Each client can have their own key, and the server doesn't need to keep a potentially large database of client keys.
 What if the client's key is leaked?
 A detail I didn't mention is that the wrapped key contains metadata.
 By default this is the current timestamp, but it is possible on creation to put any (relative short) binary data in there as the metadata.
 The server can then be configured to check the metadata by calling a script.</p>
 <p>An issue exists that an initial packet takes up resources on the server because the server needs to</p>
 <ol>
 <li>decrypt the wrapped key, and</li>
 <li>keep the unwrapped key and other data in memory while waiting for the handshake to complete.</li>
 </ol>
 <p>This can be abused in an attack very similar to a TCP <a href="https://en.wikipedia.org/wiki/SYN_flood">SYN flood</a>.
 Without <code>tls-crypt-v2</code> OpenVPN uses a specially crafted session ID (a 64 bit identifier) to avoid this issue similar to <a href="https://en.wikipedia.org/wiki/SYN_cookies">SYN cookies</a>.
 To address this in OpenVPN 2.6 the protocol for <code>tls-crypt-v2</code> was extended yet further with a 'HMAC cookie' mechanism.
 The client sends the same packet as before, but uses a sequence number <code>0x0f000001</code> instead of <code>1</code> to signal support of this mechanism.
 The server responds in a similar manner with a sequence number of <code>0x0f000001</code> and the packet is appended with a tag-length-value encoded list of flags.
 At the moment only one tag and one value is defined which signifies the server supports HMAC cookies - this seems unnecessarily complex, but is done to allow future extensibility.
 Finally, if the server supports HMAC cookies, the client sends a packet where the wrapped key is appended in cleartext.
 The server is now able to decrypt the third packet without having to keep the key from the first packet around and can verify the session id.</p>
 <h2 id="cool-lets-deploy-it"><a class="anchor" aria-hidden="true" href="#cool-lets-deploy-it"></a>Cool! Let's deploy it!</h2>
 <p>Great!
 We build on a daily basis unikernels in our <a href="https://builds.robur.coop/">reproducible builds setup</a>.
 At the time of writing we have published a <a href="https://builds.robur.coop/job/miragevpn-router">Miragevpn router unikernel</a> acting as a client.
 For general instructions on running Mirage unikernels see our <a href="https://robur.coop/Projects/Reproducible_builds">reproducible builds</a> blog post.
 The unikernel will need a block device containing the OpenVPN™ configuration and a network device.
 More detailed instructions Will Follow Soon™!
 Don't hesitate to reach out to us on <a href="https://github.com/robur-coop/miragevpn/issues">GitHub</a>, <a href="https://robur.coop/Contact">by mail</a> or me personally <a href="https://bsd.network/@reynir">on Mastodon</a> if you're stuck.</p>
 <p>[^mbedtls]: It is possible to compile OpenVPN™ community edition with Mbed TLS instead of OpenSSL which is written in C as well.</p>
 <p>[^vpn-network]: I use the term &quot;VPN network&quot; to mean the virtual private network itself. It is a bit odd because the 'N' in 'VPN' is 'Network', but without disambiguation 'VPN' could refer to the network itself, the software or the service.</p>
    </article>
        </main>
    <footer>
        <a href="https://github.com/xhtmlboi/yocaml">Powered by <strong>YOCaml</strong></a>
        <br />
    </footer>
    <script>hljs.highlightAll();</script>
  </body>
 </html>
--- a/articles/qubes-miragevpn.html
+++ b/articles/qubes-miragevpn.html
@ -0,0 +1,83 @@
 <!doctype html>
 <html lang="en">
  <head>
    <meta charset="utf-8">
    <meta http-equiv="x-ua-compatible" content="ie=edge">
    <meta name="viewport" content="width=device-width, initial-scale=1">
    <title>
        Robur's blogqubes-miragevpn, a MirageVPN client for QubesOS
    </title>
    <meta name="description" content="A new OpenVPN client for QubesOS">
    <link type="text/css" rel="stylesheet" href="https://blog.robur.coop/css/hl.css">
    <link type="text/css" rel="stylesheet" href="https://blog.robur.coop/css/style.css">
    <script src="https://blog.robur.coop/js/hl.js"></script>
    <link rel="alternate" type="application/rss+xml" href="https://blog.robur.coop/feed.xml" title="blog.robur.coop">
  </head>
  <body>
    <header>
      <h1>blog.robur.coop</h1>
      <blockquote>
        The <strong>Robur</strong> cooperative blog.
      </blockquote>
    </header>
    <main><a href="https://blog.robur.coop/index.html">Back to index</a>
 <article>
    <h1>qubes-miragevpn, a MirageVPN client for QubesOS</h1>
    <ul class="tags-list"><li>OCaml</li><li>vpn</li><li>unikernel</li><li>QubesOS</li></ul><p>We are pleased to announce the arrival of a new unikernel:
 <a href="https://github.com/robur-coop/qubes-miragevpn">qubes-miragevpn</a>. The latter is the result of work begun
 several months ago on <a href="https://github.com/robur-coop/miragevpn">miragevpn</a>.</p>
 <p>Indeed, with the ambition of completing our unikernel suite and the success of
 <a href="https://github.com/mirage/qubes-mirage-firewall">qubes-mirage-firewall</a> - as well as the general aims of
 QubesOS - we thought it would be a good idea to offer this community a unikernel
 capable of acting as an OpenVPN client, from which other virtual machines (app
 qubes) can connect so that all their connections pass through the OpenVPN
 tunnel.</p>
 <h2 id="qubesos--mirageos"><a class="anchor" aria-hidden="true" href="#qubesos--mirageos"></a>QubesOS &amp; MirageOS</h2>
 <p>Unikernels and QubesOS have always been a tempting idea for users in the sense
 that a network application (such as a firewall or VPN client) could be smaller
 than a Linux kernel: no keyboard, mouse, wifi management, etc. Just network
 management via virtual interfaces should suffice.</p>
 <p>In this case, the unikernel corresponds to this ideal where, starting from a
 base (<a href="https://github.com/Solo5/solo5">Solo5</a>) that only allows the strictly necessary (reading and
 writing on a virtual interface or block device) and building on top of it all
 the application logic strictly necessary to the objective we wish to achieve
 reduces, in effect, drastically:</p>
 <ol>
 <li>the unikernel's attack surface</li>
 <li>its weight</li>
 <li>its memory usage</li>
 </ol>
 <p>We won't go into all the work that's been done to maintain and improve
 <a href="https://github.com/mirage/qubes-mirage-firewall">qubes-mirage-firewall</a> over the last 10
 years<sup><a href="#fn1">1</a></sup>, but it's clear that this particular unikernel has
 found its audience, who aren't necessarily OCaml and MirageOS aficionados.</p>
 <p>In other words, <a href="https://github.com/mirage/qubes-mirage-firewall">qubes-mirage-firewall</a> may well be a
 fine example of what can actually be done with MirageOS, and of real utility.</p>
 <hr>
 <p><tag id="fn1"><strong>1</strong></tag>: <a href="https://github.com/marmarek">marmarek</a>, <a href="https://github.com/yomimono">Mindy</a> or
 <a href="https://github.com/mato">mato</a> were (and still are) heavily involved in the work between QubesOS
 and MirageOS. We'd also like to thank them, because if we're able to continue
 this adventure, it's also thanks to them.</p>
 <h2 id="qubesos--miragevpn"><a class="anchor" aria-hidden="true" href="#qubesos--miragevpn"></a>QubesOS &amp; MirageVPN</h2>
 <p>So, after a lengthy development phase for MirageVPN, we set about developing a
 unikernel for QubesOS to offer an OpenVPN client as an operating system. We'd
 like to give special thanks to <a href="https://github.com/palainp">Pierre Alain</a>, who helped us to better
 understand QubesOS and its possibilities.</p>
 <p>The unikernel is available here: https://github.com/robur-coop/qubes-miragevpn
 A tutorial has just been created to help QubesOS users install and configure
 such an unikernel: https://robur-coop.github.io/miragevpn-handbook/</p>
 <p>In the same way as <a href="https://github.com/mirage/qubes-mirage-firewall">qubes-mirage-firewall</a>, we hope to
 offer a solution that works and expand the circle of MirageOS and unikernel
 users!</p>
    </article>
        </main>
    <footer>
        <a href="https://github.com/xhtmlboi/yocaml">Powered by <strong>YOCaml</strong></a>
        <br />
    </footer>
    <script>hljs.highlightAll();</script>
  </body>
 </html>
--- a/articles/speeding-ec-string.html
+++ b/articles/speeding-ec-string.html
@ -0,0 +1,82 @@
 <!doctype html>
 <html lang="en">
  <head>
    <meta charset="utf-8">
    <meta http-equiv="x-ua-compatible" content="ie=edge">
    <meta name="viewport" content="width=device-width, initial-scale=1">
    <title>
        Robur's blogSpeeding elliptic curve cryptography
    </title>
    <meta name="description" content="How we improved the performance of elliptic curves by only modifying the underlying byte array">
    <link type="text/css" rel="stylesheet" href="https://blog.robur.coop/css/hl.css">
    <link type="text/css" rel="stylesheet" href="https://blog.robur.coop/css/style.css">
    <script src="https://blog.robur.coop/js/hl.js"></script>
    <link rel="alternate" type="application/rss+xml" href="https://blog.robur.coop/feed.xml" title="blog.robur.coop">
  </head>
  <body>
    <header>
      <h1>blog.robur.coop</h1>
      <blockquote>
        The <strong>Robur</strong> cooperative blog.
      </blockquote>
    </header>
    <main><a href="https://blog.robur.coop/index.html">Back to index</a>
 <article>
    <h1>Speeding elliptic curve cryptography</h1>
    <ul class="tags-list"><li>OCaml</li><li>MirageOS</li><li>cryptography</li><li>security</li></ul><p>TL;DR: replacing cstruct with string, we gain a factor of 2.5 in performance.</p>
 <h2 id="mirage-crypto-ec"><a class="anchor" aria-hidden="true" href="#mirage-crypto-ec"></a>Mirage-crypto-ec</h2>
 <p>In April 2021 We published our implementation of <a href="https://hannes.robur.coop/Posts/EC">elliptic curve cryptography</a> (as <code>mirage-crypto-ec</code> opam package) - this is DSA and DH for NIST curves P224, P256, P384, and P521, and also Ed25519 (EdDSA) and X25519 (ECDH). We use <a href="https://github.com/mit-plv/fiat-crypto/">fiat-crypto</a> for the cryptographic primitives, which emits C code that by construction is correct (note: earlier we stated &quot;free of timing side-channels&quot;, but this is a huge challenge, and as <a href="https://discuss.systems/@edwintorok/111925959867297453">reported by Edwin Török</a> likely impossible on current x86 hardware). More C code (such as <code>point_add</code>, <code>point_double</code>, and further 25519 computations including tables) have been taken from the BoringSSL code base. A lot of OCaml code originates from our TLS 1.3 work in 2018, where Etienne Millon, Nathan Rebours, and Clément Pascutto interfaced <a href="https://github.com/mirage/fiat/">elliptic curves for OCaml</a> (with the goal of being usable with MirageOS).</p>
 <p>The goal of mirage-crypto-ec was: develop elliptic curve support for OCaml &amp; MirageOS quickly - which didn't leave much time to focus on performance. As time goes by, our mileage varies, and we're keen to use fewer resources - and thus fewer CPU time and a smaller memory footprint is preferable.</p>
 <h2 id="memory-allocation-and-calls-to-c"><a class="anchor" aria-hidden="true" href="#memory-allocation-and-calls-to-c"></a>Memory allocation and calls to C</h2>
 <p>OCaml uses managed memory with a generational copying collection. To safely call a C function at any point in time when the arguments are OCaml values (memory allocated on the OCaml heap), it is crucial that while the C function is executed, the arguments should stay at the same memory location, and not being moved by the GC. Otherwise the C code may be upset retrieving wrong data or accessing unmapped memory.</p>
 <p>There are several strategies to achieve this, ranging from &quot;let's use another memory area where the GC doesn't mess around with&quot;, &quot;do not run any GC while executing the C code&quot; (read further in the OCaml <a href="https://v2.ocaml.org/releases/4.14/htmlman/intfc.html#ss:c-direct-call">cheaper C calls</a> manual), &quot;deeply copy the arguments to a non-moving memory area before executing C code&quot;, and likely others.</p>
 <p>For our elliptic curve operations, the C code is pretty simple - there are no memory allocations happening in C, neither are exceptions raised. Also, the execution time of the code is constant and pretty small.</p>
 <h2 id="ocaml-cstruct"><a class="anchor" aria-hidden="true" href="#ocaml-cstruct"></a>ocaml-cstruct</h2>
 <p>In the <a href="https://mirage.io">MirageOS</a> ecosystem, a core library is <a href="https://github.com/mirage/ocaml-cstruct">cstruct</a> - which purpose is manifold: provide ppx rewriters to define C structure layouts in OCaml (getter/setter functions are generated), as well as enums; also a fundamental idea is to use OCaml bigarray which is non-moving memory not allocated on the OCaml heap but directly by calling <code>malloc</code>. The memory can even be page-aligned, as required by some C software, such as Xen. Convenient functionality, such as &quot;retrieve a big-endian unsigned 32 bit integer from offset X in this buffer&quot; are provided as well.</p>
 <p>But there's a downside to it - as time moves along, Xen is no longer the only target for MirageOS, and other virtualization mechanisms (such as KVM / virtio) do not require page-aligned memory ranges that are retained at a given memory address. It also turns out that cstruct spends a lot of time in bounds checks. Another huge downside is that OCaml tooling (such as statmemprof) was for a long time (maybe still is not?) unaware of out-of-OCaml-GC allocated memory (cstruct uses bigarray as underlying buffer). Freeing up the memory requires finalizers to be executed - after all pretty tedious (expensive) and against the OCaml runtime philosophy.</p>
 <p>As time moves forward, also the OCaml standard library got support for (a) strings are immutable byte vectors now (since 4.06 - released in 2017 -- there's as well an interface for mutable/immutable cstruct, but that is not used as far as I can tell), (b) retrieve a certain amount of octets in a string or byte as (unsigned) integer number (since 4.08 - released in 2019, while some additional functionality is only available in 4.13).</p>
 <p>Still, bigarrays are necessary in certain situations - if you need to have a non-moving (shared) area of memory, as in the Xen interface, but also e.g. when you compute in parallel in different processes, or when you need mmap()ed files.</p>
 <h2 id="putting-it-together"><a class="anchor" aria-hidden="true" href="#putting-it-together"></a>Putting it together</h2>
 <p>Already in October 2021, Romain <a href="https://github.com/mirage/mirage-crypto/pull/146">proposed</a> to not use cstruct, but bytes for mirage-crypto-ec. The PR was sitting around since there were benchmarks missing, and developer time was small. But recently, Virgile Robles <a href="https://github.com/mirage/mirage-crypto/pull/191">proposed</a> another line of work to use pre-computed tables for NIST curves to speed up the elliptic curve cryptography. Conducting performance evaluation resulted that the &quot;use bytes instead of cstruct&quot; combined with pre-computed tables made a huge difference (factor of 6) compared to the latest release.</p>
 <p>To ease reviewing changes, we decided to focus on landing the &quot;use bytes instead of cstruct&quot; first, and gladly Pierre Alain had already rebased the existing patch onto the latest release of mirage-crypto-ec. We also went further and use string where applicable instead of bytes. For safety reasons we also introduced an API layer which (a) allocates a byte vector for the result (b) calls the primitive, and (c) transforms the byte vector into an immutable string. This API is more in line with functional programming (immutable values), and since allocations and deallocations of values are cheap, there's no measurable performance decrease.</p>
 <p>All the changes are internal, there's no external API that needs to be adjusted - still there's at the API boundary one conversion of cstruct to string (and back for the return value) done.</p>
 <p>We used <code>perf</code> to construct some flame graphs (of the ECDSA P256 sign), shown below.</p>
 <p><img src="../images/trace-cstruct-440.svg" alt="Flamegraph of ECDSA sign with cstruct" ></p>
 <p>The flame graph of P256 ECDSA sign using the mirage-crypto release 0.11.2. The majority of time is spent in &quot;do_sign&quot;, which calls <code>inv</code> (inversion), <code>scalar_mult</code> (majority of time), and <code>x_of_finite_point_mod_n</code>. The scalar multiplication spends time in <code>add</code>, <code>double</code> and <code>select</code>. Several towers starting at <code>Cstruct.create_919</code> are visible.</p>
 <p>With PR#146, the flame graph looks different:</p>
 <p><img src="../images/trace-string-770.svg" alt="Flamegraph of ECDSA sign with string" ></p>
 <p>Now, the allocation towers do not exist anymore. The time of a sign operation is spend in <code>inv</code>, <code>scalar_mult</code>, and <code>x_of_finite_point_mod_n</code>. There's still room for improvements in these operations.</p>
 <h2 id="performance-numbers"><a class="anchor" aria-hidden="true" href="#performance-numbers"></a>Performance numbers</h2>
 <p>All numbers were gathered on a Lenovo X250 laptop with a Intel i7-5600U CPU @ 2.60GHz. We used OCaml 4.14.1 as compiler. The baseline is OpenSSL 3.0.12. All numbers are in operations per second.</p>
 <p>NIST P-256</p>
 <p>| op | 0.11.2 | PR#146 | speedup | OpenSSL | speedup |
 | - | - | - | - | - | - |
 | sign | 748 | 1806 | 2.41x | 34392 | 19.04x |
 | verify | 285 | 655 | 2.30x | 12999 | 19.85x |
 | ecdh | 858 | 1785 | 2.08x | 16514 | 9.25x |</p>
 <p>Curve 25519</p>
 <p>| op | 0.11.2 | PR#146 | speedup | OpenSSL | speedup |
 | - | - | - | - | - | - |
 | sign | 10713 | 11560 | 1.08x | 21943 | 1.90x |
 | verify | 7600 | 8314 | 1.09x | 7081 | 0.85x |
 | ecdh | 12144 | 13457 | 1.11x | 26201 | 1.95x |</p>
 <p>Note: to re-create the performance numbers, you can run <code>openssl speed ecdsap256 ecdhp256 ed25519 ecdhx25519</code> - for the OCaml site, use <code>dune bu bench/speed.exe --rel</code> and <code>_build/default/bench/speed.exe ecdsa-sign ecdsa-verify ecdh-share</code>.</p>
 <p>The performance improvements are up to 2.5 times compared to the latest mirage-crypto-ec release (look at the 4th column). In comparison to OpenSSL, we still lack a factor of 20 for the NIST curves, and up to a factor of 2 for 25519 computations (look at the last column).</p>
 <p>If you have ideas for improvements, let us know via an issue, eMail, or a pull request :) We started to <a href="https://github.com/mirage/mirage-crypto/issues/193">gather some</a> for 25519 by comparing our code with changes in BoringSSL over the last years.</p>
 <p>As a spoiler, for P-256 sign there's another improvement of around 4.5 with <a href="https://github.com/mirage/mirage-crypto/pull/191">Virgile's PR</a> using pre-computed tables also for NIST curves.</p>
 <h2 id="the-road-ahead-for-2024"><a class="anchor" aria-hidden="true" href="#the-road-ahead-for-2024"></a>The road ahead for 2024</h2>
 <p>Remove all cstruct, everywhere, apart from in mirage-block-xen and mirage-net-xen ;). It was a fine decision in the early MirageOS days, but from a performance point of view, and for making our packages more broadly usable without many dependencies, it is time to remove cstruct. Earlier this year we already <a href="https://github.com/mirage/ocaml-tar/pull/137">removed cstruct from ocaml-tar</a> for similar reasons.</p>
 <p>Our MirageOS work is only partially funded, we cross-fund our work by commercial contracts and public (EU) funding. We are part of a non-profit company, you can make a (tax-deducable - at least in the EU) <a href="https://aenderwerk.de/donate/">donation</a> (select &quot;DONATION robur&quot; in the dropdown menu).</p>
 <p>We're keen to get MirageOS deployed in production - if you would like to do that, don't hesitate to reach out to us via eMail team at robur.coop</p>
    </article>
        </main>
    <footer>
        <a href="https://github.com/xhtmlboi/yocaml">Powered by <strong>YOCaml</strong></a>
        <br />
    </footer>
    <script>hljs.highlightAll();</script>
  </body>
 </html>
--- a/articles/tar-release.html
+++ b/articles/tar-release.html
@ -0,0 +1,405 @@
 <!doctype html>
 <html lang="en">
  <head>
    <meta charset="utf-8">
    <meta http-equiv="x-ua-compatible" content="ie=edge">
    <meta name="viewport" content="width=device-width, initial-scale=1">
    <title>
        Robur's blogThe new Tar release, a retrospective
    </title>
    <meta name="description" content="A little retrospective to the new Tar release and changes">
    <link type="text/css" rel="stylesheet" href="https://blog.robur.coop/css/hl.css">
    <link type="text/css" rel="stylesheet" href="https://blog.robur.coop/css/style.css">
    <script src="https://blog.robur.coop/js/hl.js"></script>
    <link rel="alternate" type="application/rss+xml" href="https://blog.robur.coop/feed.xml" title="blog.robur.coop">
  </head>
  <body>
    <header>
      <h1>blog.robur.coop</h1>
      <blockquote>
        The <strong>Robur</strong> cooperative blog.
      </blockquote>
    </header>
    <main><a href="https://blog.robur.coop/index.html">Back to index</a>
 <article>
    <h1>The new Tar release, a retrospective</h1>
    <ul class="tags-list"><li>OCaml</li><li>Cstruct</li><li>functors</li></ul><p>We are delighted to announce the new release of <code>ocaml-tar</code>. A small library for
 reading and writing tar archives in OCaml. Since this is a major release, we'll
 take the time in this article to explain the work that's been done by the
 cooperative on this project.</p>
 <p>Tar is an <strong>old</strong> project. Originally written by David Scott as part of Mirage,
 this project is particularly interesting for building bridges between the tools
 we can offer and what already exists. Tar is, in fact, widely used. So we're
 both dealing with a format that's older than I am (but I'm used to it by email)
 and a project that's been around since... 2012 (over 10 years!).</p>
 <p>But we intend to maintain and improve it, since we're using it for the
 <a href="https://hannes.robur.coop/Posts/OpamMirror">opam-mirror</a> project among other things - this unikernel is to
 provide an opam-repository &quot;tarball&quot; for opam when you do <code>opam update</code>.</p>
 <h2 id="cstructt--bytes"><a class="anchor" aria-hidden="true" href="#cstructt--bytes"></a><code>Cstruct.t</code> &amp; bytes</h2>
 <p>As some of you may have noticed, over the last few months we've begun a fairly
 substantial change to the Mirage ecosystem, replacing the use of <code>Cstruct.t</code> in
 key places with bytes/string.</p>
 <p>This choice is based on 2 considerations:</p>
 <ul>
 <li>we came to realize that <code>Cstruct.t</code> could be very costly in terms of
 performance</li>
 <li><code>Cstruct.t</code> remains a &quot;Mirage&quot; structure; outside the Mirage ecosystem, the
 use of <code>Cstruct.t</code> is not so &quot;obvious&quot;.</li>
 </ul>
 <p>The pull-request is available here: https://github.com/mirage/ocaml-tar/pull/137.
 The discussion can be interesting in discovering common bugs (uninitialized
 buffer, invalid access). There's also a small benchmark to support our initial
 intuition<sup><a href="#fn1">1</a></sup>.</p>
 <p>But this PR can also be an opportunity to understand the existence of
 <code>Cstruct.t</code> in the Mirage ecosystem and the reasons for this historic choice.</p>
 <h3 id="cstructt-as-a-non-moveable-data"><a class="anchor" aria-hidden="true" href="#cstructt-as-a-non-moveable-data"></a><code>Cstruct.t</code> as a non-moveable data</h3>
 <p>I've already <a href="https://discuss.ocaml.org/t/buffered-io-bytes-vs-bigstring/8978/3">made</a> a list of pros/cons when it comes to
 bigarrays. Indeed, <code>Cstruct.t</code> is based on a bigarray:</p>
 <pre><code class="language-ocaml">type buffer = (char, Bigarray.int8_unsigned_elt, Bigarray.c_layout) Bigarray.Array1.t
 type t =
  { buffer : buffer
  ; off : int
  ; len : int }
 </code></pre>
 <p>The experienced reader may rightly wonder why Cstruct.t is a bigarray with <code>off</code>
 and <code>len</code>. First, we need to clarify what a bigarray is for OCaml.</p>
 <p>A bigarray is a somewhat special value in OCaml. This value is allocated in the
 C heap. In other words, its contents are not in OCaml's garbage collector, but
 exist outside it. The first (and very important) implication of this feature is
 that the contents of a bigarray do not move (even if the GC tries to defragment
 the memory). This feature has several advantages:</p>
 <ul>
 <li>in parallel programming, it can be very interesting to use a bigarray knowing
 that, from the point of view of the 2 processes, the position of the bigarray
 will never change - this is essentially what <a href="https://github.com/rdicosmo/parmap">parmap</a> does (before
 OCaml 5).</li>
 <li>for calculations such as checksum or hash, it can be interesting to use a
 bigarray. The calculation would not be interrupted by the GC since the
 bigarray does not move. The calculation can therefore be continued at the same
 point, which can help the CPU to better predict the next stage of the
 calculation. This is what <a href="https://github.com/mirage/digestif">digestif</a> offers and what
 <a href="https://github.com/mirage/decompress">decompress</a> requires.</li>
 <li>for one reason or another, particularly when interacting with something other
 than OCaml, you need to offer a memory zone that cannot move. This is
 particularly true for unikernels as Xen guests (where the <em>net device</em>
 corresponds to a fixed memory zone with which we need to interact) or
 <a href="https://ocaml.org/manual/5.2/api/Unix.html#1_Mappingfilesintomemory">mmap</a>.</li>
 <li>there are other subtleties more related to the way OCaml compiles. For
 example, using bigarray layouts to manipulate &quot;bigger words&quot; can really have
 an impact on performance, as <a href="https://github.com/robur-coop/utcp/pull/29">this PR</a> on <a href="https://github.com/robur-coop/utcp">utcp</a> shows.</li>
 <li>finally, it may be useful to store sensitive information in a bigarray so as
 to have the opportunity to clean up this information as quickly as possible
 (ensuring that the GC has not made a copy) in certain situations.</li>
 </ul>
 <p>All these examples show that bigarrays can be of real interest as long as
 <strong>their uses are properly contextualized</strong> - which ultimately remains very
 specific. Our experience of using them in Mirage has shown us their advantages,
 but also, and above all, their disadvantages:</p>
 <ul>
 <li>keep in mind that bigarray allocation uses either a system call like <code>mmap</code> or
 <code>malloc()</code>. The latter, compared with what OCaml can offer, is slow. As soon
 as you need to allocate bytes/strings smaller than
 <a href="https://github.com/ocaml/ocaml/blob/744006bfbfa045cc1ca442ff7b52c2650d2abe32/runtime/alloc.c#L175"><code>(256 * words)</code></a>, these values are allocated in the minor heap,
 which is incredibly fast to allocate (3 processor instructions which can be
 predicted very well). So, preferring to allocate a 10-byte bigarray rather
 than a 10-byte <code>bytes</code> penalizes you enormously.</li>
 <li>since the bigarray exists in the C heap, the GC has a special mechanism for
 knowing when to <code>free()</code> the zone as soon as the value is no longer in use.
 Reference-counting is used to then allocate &quot;small&quot; values in the OCaml heap
 and use them to manipulate <em>indirectly</em> the bigarray.</li>
 </ul>
 <h4 id="ownership-proxy-and-gc"><a class="anchor" aria-hidden="true" href="#ownership-proxy-and-gc"></a>Ownership, proxy and GC</h4>
 <p>This last point deserves a little clarification, particularly with regard to the
 <code>Bigarray.sub</code> function. This function will not create a new, smaller bigarray
 and copy what was in the old one to the new one (as <code>Bytes.sub</code>/<code>String.sub</code>
 does). In fact, OCaml will allocate a &quot;proxy&quot; of your bigarray that represents a
 subfield. This is where <em>reference-counting</em> comes in. This proxy value needs
 the initial bigarray to be manipulated. So, as long as proxies exist, the GC
 cannot <code>free()</code> the initial bigarray.</p>
 <p>This poses several problems:</p>
 <ul>
 <li>the first is the allocation of these proxies. They can help us to manipulate
 the initial bigarray in several places without copying it, but as time goes
 by, these proxies could be very expensive</li>
 <li>the second is GC intervention. You still need to scan the bigarray, in a
 particular way, to know whether or not to keep it. This particular scan, once
 again in time immemorial, was not all that common.</li>
 <li>the third concerns bigarray ownership. Since we're talking about proxies, we
 can imagine 2 competing tasks having access to the same bigarray.</li>
 </ul>
 <p>As far as the first point is concerned, <code>Bigarray.sub</code> could still be &quot;slow&quot; for
 small data since it was, <em>de facto</em> (since a bigarray always has a finalizer -
 don't forget reference counting!), allocated in the major heap. And, in truth,
 this is perhaps the main reason for the existence of Cstruct! To have a &quot;proxy&quot;
 to a bigarray allocated in the minor heap (and, be fast). But since
 <a href="https://github.com/ocaml/ocaml/pull/92">Pierre Chambart's PR#92</a>, the problem is no more.</p>
 <p>The second point, on the other hand, is still topical, even if we can see that
 <a href="https://github.com/ocaml/ocaml/pull/1738">considerable efforts</a> have been made. What we see every
 day on our unikernels is <a href="https://github.com/ocaml/ocaml/issues/7750">the pressure</a> that can be put on
 the GC when it comes to bigarrays. Indeed, bigarrays use memory and making the C
 heap cohabit with the OCaml heap inevitably comes at a cost. As far as
 unikernels are concerned, which have a more limited memory than an OCaml
 application, we reach this limit rather quickly and we therefore ask the GC to
 work more specifically on our 10 or 20 byte bigarrays...</p>
 <p>Finally, the third point can be the toughest. On several occasions, we've
 noticed competing accesses on our bigarrays that we didn't want (for example,
 <code>http-lwt-client</code> had <a href="https://github.com/robur-coop/http-lwt-client/pull/16">this problem</a>). In our experience,
 it's very difficult to observe and know that there is indeed an unauthorized
 concurrent access changing the contents of our buffer. In this respect, the
 question remains open as regards <code>Cstruct.t</code> and the possibility of encoding
 ownership of a <code>Cstruct.t</code> in the type to prevent unauthorized access.
 <a href="https://github.com/mirage/ocaml-cstruct/pull/237">This PR</a> is interesting to see all the discussions that have taken
 place on this subject<sup><a href="#fn2">2</a></sup>.</p>
 <p>It should be noted that, with regard to the third point, the problem also
 applies to bytes and the use of <code>Bytes.unsafe_to_string</code>!</p>
 <h3 id="conclusion-about-cstruct"><a class="anchor" aria-hidden="true" href="#conclusion-about-cstruct"></a>Conclusion about Cstruct</h3>
 <p>We hope we've been thorough enough in our experience with Cstruct. If we go back
 to the initial definition of our <code>Cstruct.t</code> shown above and take all the
 history into account, it becomes increasingly difficult to argue for a
 <strong>systematic</strong> use of Cstruct in our unikernels. In fact, the question of
 <code>Cstruct.t</code> versus bytes/string remains completely open.</p>
 <p>It's worth noting that the original reasons for <code>Cstruct.t</code> are no longer really
 relevant if we consider how OCaml has evolved. It should also be noted that this
 systematic approach to using <code>Cstruct.t</code> rather than bytes/string has cost us.</p>
 <p>This is not to say that <code>Cstruct.t</code> is obsolete. The library is very good and
 offers an API where manipulating bytes to extract information such as a TCP/IP
 packet remains more pleasant than directly using bytes (even if, here too,
 <a href="https://github.com/ocaml/ocaml/pull/1864">efforts</a> have been made).</p>
 <p>As far as <code>ocaml-tar</code> is concerned, what really counts is the possibility for
 other projects to use this library without requiring <code>Cstruct.t</code> - thus
 facilitating its adoption. In other words, given the advantages/disadvantages of
 <code>Cstruct.t</code>, we felt it would be a good idea to remove this dependency.</p>
 <hr />
 <p><tag id="fn1"><strong>1</strong></tag>: It should be noted that the benchmark also concerns
 compression. In this case, we use <code>decompress</code>, which uses bigarrays. So there's
 some copying involved (from bytes to bigarrays)! But despite this copying, it
 seems that the change is worthwhile.</p>
 <p><tag id="fn2"><strong>2</strong></tag>: It reminds me that we've been experimenting with
 capabilities and using the type system to enforce certain characteristics. To
 date, <code>Cstruct_cap</code> has not been used anywhere, which raises a real question
 about the advantages/disadvantages in everyday use.</p>
 <h2 id="functors"><a class="anchor" aria-hidden="true" href="#functors"></a>Functors</h2>
 <p>This is perhaps the other point of the Mirage ecosystem that is also the subject
 of debate. Functors! Before we talk about functors, we need to understand their
 relevance in the context of Mirage.</p>
 <p>Mirage transforms an application into an operating system. What's the difference
 between a &quot;normal&quot; application and a unikernel: the &quot;subsystem&quot; with which you
 interact. In this case, a normal application will interact with the host system,
 whereas a unikernel will have to interact with the Solo5 <em>mini-system</em>.</p>
 <p>What Mirage is trying to offer is the ability for an application to transform
 itself into either without changing a thing! Mirage's aim is to <strong>inject</strong> the
 subsystem into your application. In this case:</p>
 <ul>
 <li>inject <code>unix.cmxa</code> when you want a Mirage application to become a simple
 executable</li>
 <li>inject <a href="https://github.com/mirage/ocaml-solo5">ocaml-solo5</a> when you want to produce a unikernel</li>
 </ul>
 <p>So we're not going to talk about the pros and cons of this approach here, but
 consider this feature as one that requires us to use functors.</p>
 <p>Indeed, what's the best way in OCaml to inject one implementation into another:
 functors? There are definite advantages here too, but we're going to concentrate
 on one in particular: the expressiveness of types at module level (which can be
 used as arguments to our functors).</p>
 <p>For example, did you know that OCaml has a dependent type system?</p>
 <pre><code class="language-ocaml">type 'a nat = Zero : zero nat | Succ : 'a nat -&gt; 'a succ nat
 and zero = |
 and 'a succ = S
 module type T = sig type t val v : t nat end
 module type Rec = functor (T:T) -&gt; T
 module type Nat = functor (S:Rec) -&gt; functor (Z:T) -&gt; T
 module Zero = functor (S:Rec) -&gt; functor (Z:T) -&gt; Z
 module Succ = functor (N:Nat) -&gt; functor (S:Rec) -&gt; functor (Z:T) -&gt; S(N(S)(Z))
 module Add = functor (X:Nat) -&gt; functor (Y:Nat) -&gt; functor (S:Rec) -&gt; functor (Z:T) -&gt; X(S)(Y(S)(Z))
 module One = Succ(Zero)
 module Two_a = Add(One)(One)
 module Two_b = Succ(One)
 module Z : T with type t = zero = struct
  type t = zero
  let v = Zero
 end
 module S (T:T) : T with type t = T.t succ = struct
  type t = T.t succ
  let v = Succ T.v
 end
 module A = Two_a(S)(Z)
 module B = Two_b(S)(Z)
 type ('a, 'b) refl = Refl : ('a, 'a) refl
 let _ : (A.t, B.t) refl = Refl (* 1+1 == succ 1 *)
 </code></pre>
 <p>The code is ... magical, but it shows that two differently constructed modules
 (<code>Two_a</code> &amp; <code>Two_b</code>) ultimately produce the same type, and OCaml is able to prove
 this equality. Above all, the example shows just how powerful functors can be.
 But it also shows just how difficult functors can be to understand and use.</p>
 <p>In fact, this is one of Mirage's biggest drawbacks: the overuse of functors
 makes the code difficult to read and understand. It can be difficult to deduce
 in your head the type that results from an application of functors, and the
 constraints associated with it... (yes, I don't use <code>merlin</code>).</p>
 <p>But back to our initial problem: injection! In truth, the functor is a
 fly-killing sledgehammer in most cases. There are many other ways of injecting
 what the system would be (and how to do a <code>read</code> or <code>write</code>) into an
 implementation. The best example, as <a href="https://discuss.ocaml.org/t/best-practices-and-design-patterns-for-supporting-concurrent-io-in-libraries/15001/4?u=dinosaure">@nojb pointed out</a>, is of
 course <a href="https://github.com/mirleft/ocaml-tls">ocaml-tls</a> - this answer also shows a contrast between the
 functor approach (with <a href="https://github.com/mirage/ocaml-cohttp">CoHTTP</a> for example) and the &quot;pure value-passing
 interface&quot; of <code>ocaml-tls</code>.</p>
 <p>What's more, we've been trying to find other approaches for injecting the system
 we want for several years now. We can already list several:</p>
 <ul>
 <li><code>ocaml-tls</code>' &quot;value-passing&quot; approach, of course, but also <code>decompress</code></li>
 <li>of course, there's the passing of <a href="https://github.com/mirage/colombe/blob/07cd4cf134168ecd841924ee7ddda1a9af8fbd5a/src/sigs.ml#L13-L16">a record</a> (a sort of
 mini-module with fewer possibilities with types, but which does the job - a
 poor man's functor, in short) which would have the functions to perform the
 system's operations</li>
 <li><a href="https://github.com/dinosaure/mimic">mimic</a> can be used to inject a module as an implementation of a
 flow/stream according to a resolution mechanism (DNS, <code>/etc/services</code>, etc.) -
 a little closer to the idea of <em>runtime-resolved implicit implementations</em></li>
 <li>there are, of course, the variants (but if we go back to 2010, this solution
 wasn't so obvious) popularized by <a href="https://github.com/dbuenzli/ptime">ptime</a>/<a href="https://github.com/dbuenzli/mtime">mtime</a>, <code>digestif</code> &amp;
 <a href="https://github.com/ocaml/dune/pull/1207">dune</a></li>
 <li>and finally, <a href="https://github.com/mirage/decompress/blob/c8301ba674e037b682338958d6d0bb5c42fd720e/lib/lzo.ml#L164-L175">GADTs</a>, which describe what the process should
 do, then let the user implement the <code>run</code> function according to the system.</li>
 </ul>
 <p>In short, based on this list and the various experiments we've carried out on a
 number of projects, we've decided to remove the functors from <code>ocaml-tar</code>! The
 crucial question now is: which method to choose?</p>
 <h3 id="the-best-answers"><a class="anchor" aria-hidden="true" href="#the-best-answers"></a>The best answers</h3>
 <p>There's no real answer to that, and in truth it depends on what level of
 abstraction you're at. In fact, you'd like to have a fairly simple method of
 abstraction from the system at the start and at the lowest level, to end up
 proposing a functor that does all the <em>ceremony</em> (the glue between your
 implementation and the system) at the end - that's what <a href="https://github.com/mirage/ocaml-git">ocaml-git</a>
 does, for example.</p>
 <p>The abstraction you choose also depends on how the process is going to work. As
 far as streams/protocols are concerned, the <code>ocaml-tls</code>/<code>decompress</code> approach
 still seems the best. But when it comes to introspecting a file/block-device, it
 may be preferable to use a GADT that will force the user to implement an
 arbitrary memory access rather than consume a sequence of bytes. In short, at
 this stage, experience speaks for itself and, just as we were wrong about
 functors, we won't be advising you to use this or that solution.</p>
 <p>But based on our experience of <code>ocaml-tls</code> &amp; <code>decompress</code> with LZO (which
 requires arbitrary access to the content) and the way Tar works, we decided to
 use a &quot;value-passing&quot; approach (to describe when we need to read/write) and a
 GADT to describe calculations such as:</p>
 <ul>
 <li>iterating over the files/folders contained in a Tar document</li>
 <li>producing a Tar file according to a &quot;dispenser&quot; of inputs</li>
 </ul>
 <pre><code class="language-ocaml">val decode : decode_state -&gt; string -&gt;
  decode_state *
   * [ `Read of int
     | `Skip of int
     | `Header of Header.t ] option
   * Header.Extended.t option
 (** [decode state] returns a new state and what the user should do next:
    - [`Skip] skip bytes
    - [`Read] read bytes
    - [`Header hdr] do something according the last header extracted
      (like stream-out the contents of a file). *)
 type ('a, 'err) t =
  | Really_read : int -&gt; (string, 'err) t
  | Read : int -&gt; (string, 'err) t
  | Seek : int -&gt; (unit, 'err) t
  | Bind : ('a, 'err) t * ('a -&gt; ('b, 'err) t) -&gt; ('b, 'err) t
  | Return : ('a, 'err) result -&gt; ('a, 'err) t
  | Write : string -&gt; (unit, 'err) t
 </code></pre>
 <p>However, and this is where we come back to OCaml's limitations and where
 functors could help us: higher kinded polymorphism!</p>
 <h3 id="higher-kinded-polymorphism"><a class="anchor" aria-hidden="true" href="#higher-kinded-polymorphism"></a>Higher kinded Polymorphism</h3>
 <p>If we return to our functor example above, there's one element that may be of
 interest: <code>T with type t = T.t succ</code></p>
 <p>In other words, add a constraint to a signature type. A constraint often seen
 with Mirage (but deprecated now according to <a href="https://github.com/mirage/mirage/issues/1004#issue-507517315">this issue</a>) is the
 type <code>io</code> and its constraint: <code>type 'a io</code>, <code>with type 'a io = 'a Lwt.t</code>.</p>
 <p>So we had this type in Tar. The problem is that our GADT can't understand that
 sometimes it will have to manipulate <em>Lwt</em> values, sometimes <em>Async</em> or
 sometimes <em>Eio</em> (or <em>Miou</em>!). In other words: how do we compose our <code>Bind</code> with
 the <code>Bind</code> of these three targets? The difficulty lies above all in history?
 Supporting this library requires us to assume a certain compatibility with
 applications over which we have no control. What's more, we need to maintain
 support for all three libraries without imposing one.</p>
 <hr />
 <p>A small disgression at this stage seems important to us, as we've been working
 in this way for over 10 years. Of course, despite all the solutions mentioned
 above, not depending on a system (and/or a scheduler) also allows us to ensure
 the existence of libraries like Tar over more than a decade! The OCaml ecosystem
 is changing, and choosing this or that library to facilitate the development of
 an application has implications we might regret 10 years down the line (for
 example... <code>Cstruct.t</code>!). So, it can be challenging to ensure compatibility with
 all systems, but the result is libraries steeped in the experience and know-how
 of many developers!</p>
 <hr />
 <p>So, and this is why we talk about Higher Kinded Polymorphism, how do we abstract
 the <code>t</code> from <code>'a t</code> (to replace it with <code>Lwt.t</code> or even with a type such as
 <code>type 'a t = 'a</code>)? This is where we're going to use the trick explained in
 <a href="https://www.cl.cam.ac.uk/~jdy22/papers/lightweight-higher-kinded-polymorphism.pdf">this paper</a>. The trick is to consider a &quot;new type&quot; that will represent our
 monad (lwt or async) and inject/project a value from this monad to something
 understandable by our GADT: <code>High : ('a, 't) io -&gt; ('a, 't) t</code>.</p>
 <pre><code class="language-ocaml">type ('a, 't) io
 type ('a, 'err, 't) t =
  | Really_read : int -&gt; (string, 'err, 't) t
  | Read : int -&gt; (string, 'err, 't) t
  | Seek : int -&gt; (unit, 'err, 't) t
  | Bind : ('a, 'err, 't) t * ('a -&gt; ('b, 'err, 't) t) -&gt; ('b, 'err, 't) t
  | Return : ('a, 'err) result -&gt; ('a, 'err, 't) t
  | Write : string -&gt; (unit, 'err, 't) t
  | High : ('a, 't) io -&gt; ('a, 'err, 't) t
 </code></pre>
 <p>Next, we need to create this new type according to the chosen scheduler. Let's
 take <em>Lwt</em> as an example:</p>
 <pre><code class="language-ocaml">module Make (X : sig type 'a t end) = struct
  type t (* our new type *)
  type 'a s = 'a X.t
  external inj : 'a s -&gt; ('a, t) io = &quot;%identity&quot;
  external prj : ('a, t) io -&gt; 'a s = &quot;%identity&quot;
 end
 module L = Make(Lwt)
 let rec run
  : type a err. (a, err, L.t) t -&gt; (a, err) result Lwt.t
  = function
  | High v -&gt; Ok (L.prj v)
  | Return v -&gt; Lwt.return v
  | Bind (x, f) -&gt;
    run x &gt;&gt;= fun value -&gt; run (f value)
  | _ -&gt; ...
 </code></pre>
 <p>So, as you can see, it's a real trick to avoid doing at home without a
 companion. Indeed, the use of <code>%identity</code> corresponds to an <code>Obj.magic</code>! So even
 if the <code>io</code> type is exposed (to let the user derive Tar for their own system),
 this trick is not exposed for other packages, and we instead suggest helpers
 such as:</p>
 <pre><code class="language-ocaml">val lwt : 'a Lwt.t -&gt; ('a, 'err, lwt) t
 val miou : 'a -&gt; ('a, 'err, miou) t
 </code></pre>
 <p>But this way, Tar can always be derived from another system, and the process for
 extracting entries from a Tar file is the same for <strong>all</strong> systems!</p>
 <h2 id="conclusion"><a class="anchor" aria-hidden="true" href="#conclusion"></a>Conclusion</h2>
 <p>This Tar release isn't as impressive as this article, but it does sum up all the
 work we've been able to do over the last few months and years. We hope that our
 work is appreciated and that this article, which sets out all the thoughts we've
 had (and still have), helps you to better understand our work!</p>
    </article>
        </main>
    <footer>
        <a href="https://github.com/xhtmlboi/yocaml">Powered by <strong>YOCaml</strong></a>
        <br />
    </footer>
    <script>hljs.highlightAll();</script>
  </body>
 </html>
--- a/atom.xml
+++ b/atom.xml
@ -0,0 +1,137 @@
 <?xml version="1.0" encoding="utf-8"?>
 <feed xmlns="http://www.w3.org/2005/Atom">
  <id>https://blog.robur.coop/atom.xml</id>
  <title type="text">The Robur&apos;s blog</title>
  <generator uri="https://github.com/xhtmlboi/yocaml" version="2">YOCaml</generator>
  <updated>2024-08-21T00:00:00Z</updated>
  <author>
    <name>The Robur Team</name>
  </author>
  <link href="https://blog.robur.coop/"/>
  <link href="https://blog.robur.coop/atom.xml" rel="self"/>
  <entry>
    <id>https://blog.robur.coop//articles/2024-08-21-OpenVPN-and-MirageVPN.html</id>
    <title type="text">MirageVPN and OpenVPN</title>
    <updated>2024-08-21T00:00:00Z</updated>
    <summary type="text">Discoveries made implementing MirageVPN, a OpenVPN-compatible VPN library</summary>
    <link href="https://blog.robur.coop//articles/2024-08-21-OpenVPN-and-MirageVPN.html" rel="alternate" title="MirageVPN and OpenVPN"/>
    <category term="MirageVPN"/>
    <category term="OpenVPN"/>
    <category term="security"/>
  </entry>
  <entry>
    <id>https://blog.robur.coop//articles/tar-release.html</id>
    <title type="text">The new Tar release, a retrospective</title>
    <updated>2024-08-15T00:00:00Z</updated>
    <summary type="text">A little retrospective to the new Tar release and changes</summary>
    <link href="https://blog.robur.coop//articles/tar-release.html" rel="alternate" title="The new Tar release, a retrospective"/>
    <category term="OCaml"/>
    <category term="Cstruct"/>
    <category term="functors"/>
  </entry>
  <entry>
    <id>https://blog.robur.coop//articles/qubes-miragevpn.html</id>
    <title type="text">qubes-miragevpn, a MirageVPN client for QubesOS</title>
    <updated>2024-06-24T00:00:00Z</updated>
    <summary type="text">A new OpenVPN client for QubesOS</summary>
    <link href="https://blog.robur.coop//articles/qubes-miragevpn.html" rel="alternate" title="qubes-miragevpn, a MirageVPN client for QubesOS"/>
    <category term="OCaml"/>
    <category term="vpn"/>
    <category term="unikernel"/>
    <category term="QubesOS"/>
  </entry>
  <entry>
    <id>https://blog.robur.coop//articles/miragevpn-server.html</id>
    <title type="text">MirageVPN server</title>
    <updated>2024-06-17T00:00:00Z</updated>
    <summary type="text">Announcement of our MirageVPN server.</summary>
    <link href="https://blog.robur.coop//articles/miragevpn-server.html" rel="alternate" title="MirageVPN server"/>
    <category term="OCaml"/>
    <category term="MirageOS"/>
    <category term="cryptography"/>
    <category term="security"/>
    <category term="VPN"/>
  </entry>
  <entry>
    <id>https://blog.robur.coop//articles/miragevpn-performance.html</id>
    <title type="text">Speeding up MirageVPN and use it in the wild</title>
    <updated>2024-04-16T00:00:00Z</updated>
    <summary type="text">Performance engineering of MirageVPN, speeding it up by a factor of 25.</summary>
    <link href="https://blog.robur.coop//articles/miragevpn-performance.html" rel="alternate" title="Speeding up MirageVPN and use it in the wild"/>
    <category term="OCaml"/>
    <category term="MirageOS"/>
    <category term="cryptography"/>
    <category term="security"/>
    <category term="VPN"/>
    <category term="performance"/>
  </entry>
  <entry>
    <id>https://blog.robur.coop//articles/gptar.html</id>
    <title type="text">GPTar</title>
    <updated>2024-02-21T00:00:00Z</updated>
    <summary type="text">Hybrid GUID partition table and tar archive</summary>
    <link href="https://blog.robur.coop//articles/gptar.html" rel="alternate" title="GPTar"/>
    <category term="OCaml"/>
    <category term="gpt"/>
    <category term="tar"/>
    <category term="mbr"/>
    <category term="persistent storage"/>
  </entry>
  <entry>
    <id>https://blog.robur.coop//articles/speeding-ec-string.html</id>
    <title type="text">Speeding elliptic curve cryptography</title>
    <updated>2024-02-13T00:00:00Z</updated>
    <summary type="text">
      How we improved the performance of elliptic curves by only modifying the underlying byte array
    </summary>
    <link href="https://blog.robur.coop//articles/speeding-ec-string.html" rel="alternate" title="Speeding elliptic curve cryptography"/>
    <category term="OCaml"/>
    <category term="MirageOS"/>
    <category term="cryptography"/>
    <category term="security"/>
  </entry>
  <entry>
    <id>https://blog.robur.coop//articles/lwt_pause.html</id>
    <title type="text">Cooperation and Lwt.pause</title>
    <updated>2024-02-11T00:00:00Z</updated>
    <summary type="text">A disgression about Lwt and Miou</summary>
    <link href="https://blog.robur.coop//articles/lwt_pause.html" rel="alternate" title="Cooperation and Lwt.pause"/>
    <category term="OCaml"/>
    <category term="Scheduler"/>
    <category term="Community"/>
    <category term="Unikernel"/>
    <category term="Git"/>
  </entry>
  <entry>
    <id>https://blog.robur.coop//articles/2024-02-03-python-str-repr.html</id>
    <title type="text">Python&apos;s `str.__repr__()`</title>
    <updated>2024-02-03T00:00:00Z</updated>
    <summary type="text">Reimplementing Python string escaping in OCaml</summary>
    <link href="https://blog.robur.coop//articles/2024-02-03-python-str-repr.html" rel="alternate" title="Python&apos;s `str.__repr__()`"/>
    <category term="OCaml"/>
    <category term="Python"/>
    <category term="unicode"/>
  </entry>
  <entry>
    <id>https://blog.robur.coop//articles/miragevpn-ncp.html</id>
    <title type="text">MirageVPN updated (AEAD, NCP)</title>
    <updated>2023-11-20T00:00:00Z</updated>
    <summary type="text">How we resurrected MirageVPN from its bitrot state</summary>
    <link href="https://blog.robur.coop//articles/miragevpn-ncp.html" rel="alternate" title="MirageVPN updated (AEAD, NCP)"/>
    <category term="OCaml"/>
    <category term="MirageOS"/>
    <category term="VPN"/>
    <category term="security"/>
  </entry>
  <entry>
    <id>https://blog.robur.coop//articles/miragevpn.html</id>
    <title type="text">MirageVPN &amp; tls-crypt-v2</title>
    <updated>2023-11-14T00:00:00Z</updated>
    <summary type="text">How we implementated tls-crypt-v2 for miragevpn</summary>
    <link href="https://blog.robur.coop//articles/miragevpn.html" rel="alternate" title="MirageVPN &amp; tls-crypt-v2"/>
    <category term="OCaml"/>
    <category term="MirageOS"/>
    <category term="VPN"/>
    <category term="security"/>
  </entry>
 </feed>
--- a/feed.xml
+++ b/feed.xml
@ -0,0 +1,95 @@
 <?xml version="1.0" encoding="utf-8"?>
 <rss version="2.0" xmlns:atom="http://www.w3.org/2005/Atom">
  <channel>
    <title>The Robur&apos;s blog</title>
    <link>https://blog.robur.coop/</link>
    <description><![CDATA[The Robur cooperative blog]]></description>
    <atom:link href="https://blog.robur.coop/feed.xml" rel="self" type="application/rss+xml"/>
    <lastBuildDate>Wed, 21 Aug 2024 00:00:00 GMT</lastBuildDate>
    <docs>https://www.rssboard.org/rss-specification</docs>
    <generator>YOCaml</generator>
    <item>
      <title>MirageVPN and OpenVPN</title>
      <link>https://blog.robur.coop//articles/2024-08-21-OpenVPN-and-MirageVPN.html</link>
      <description>
        <![CDATA[Discoveries made implementing MirageVPN, a OpenVPN-compatible VPN library]]>
      </description>
      <guid isPermaLink="true">https://blog.robur.coop//articles/2024-08-21-OpenVPN-and-MirageVPN.html</guid>
      <pubDate>Wed, 21 Aug 2024 00:00:00 GMT</pubDate>
    </item>
    <item>
      <title>The new Tar release, a retrospective</title>
      <link>https://blog.robur.coop//articles/tar-release.html</link>
      <description><![CDATA[A little retrospective to the new Tar release and changes]]></description>
      <guid isPermaLink="true">https://blog.robur.coop//articles/tar-release.html</guid>
      <pubDate>Thu, 15 Aug 2024 00:00:00 GMT</pubDate>
    </item>
    <item>
      <title>qubes-miragevpn, a MirageVPN client for QubesOS</title>
      <link>https://blog.robur.coop//articles/qubes-miragevpn.html</link>
      <description><![CDATA[A new OpenVPN client for QubesOS]]></description>
      <guid isPermaLink="true">https://blog.robur.coop//articles/qubes-miragevpn.html</guid>
      <pubDate>Mon, 24 Jun 2024 00:00:00 GMT</pubDate>
    </item>
    <item>
      <title>MirageVPN server</title>
      <link>https://blog.robur.coop//articles/miragevpn-server.html</link>
      <description><![CDATA[Announcement of our MirageVPN server.]]></description>
      <guid isPermaLink="true">https://blog.robur.coop//articles/miragevpn-server.html</guid>
      <pubDate>Mon, 17 Jun 2024 00:00:00 GMT</pubDate>
    </item>
    <item>
      <title>Speeding up MirageVPN and use it in the wild</title>
      <link>https://blog.robur.coop//articles/miragevpn-performance.html</link>
      <description>
        <![CDATA[Performance engineering of MirageVPN, speeding it up by a factor of 25.]]>
      </description>
      <guid isPermaLink="true">https://blog.robur.coop//articles/miragevpn-performance.html</guid>
      <pubDate>Tue, 16 Apr 2024 00:00:00 GMT</pubDate>
    </item>
    <item>
      <title>GPTar</title>
      <link>https://blog.robur.coop//articles/gptar.html</link>
      <description><![CDATA[Hybrid GUID partition table and tar archive]]></description>
      <guid isPermaLink="true">https://blog.robur.coop//articles/gptar.html</guid>
      <pubDate>Wed, 21 Feb 2024 00:00:00 GMT</pubDate>
    </item>
    <item>
      <title>Speeding elliptic curve cryptography</title>
      <link>https://blog.robur.coop//articles/speeding-ec-string.html</link>
      <description>
        <![CDATA[How we improved the performance of elliptic curves by only modifying the underlying byte array]]>
      </description>
      <guid isPermaLink="true">https://blog.robur.coop//articles/speeding-ec-string.html</guid>
      <pubDate>Tue, 13 Feb 2024 00:00:00 GMT</pubDate>
    </item>
    <item>
      <title>Cooperation and Lwt.pause</title>
      <link>https://blog.robur.coop//articles/lwt_pause.html</link>
      <description><![CDATA[A disgression about Lwt and Miou]]></description>
      <guid isPermaLink="true">https://blog.robur.coop//articles/lwt_pause.html</guid>
      <pubDate>Sun, 11 Feb 2024 00:00:00 GMT</pubDate>
    </item>
    <item>
      <title>Python&apos;s `str.__repr__()`</title>
      <link>https://blog.robur.coop//articles/2024-02-03-python-str-repr.html</link>
      <description><![CDATA[Reimplementing Python string escaping in OCaml]]></description>
      <guid isPermaLink="true">https://blog.robur.coop//articles/2024-02-03-python-str-repr.html</guid>
      <pubDate>Sat, 03 Feb 2024 00:00:00 GMT</pubDate>
    </item>
    <item>
      <title>MirageVPN updated (AEAD, NCP)</title>
      <link>https://blog.robur.coop//articles/miragevpn-ncp.html</link>
      <description><![CDATA[How we resurrected MirageVPN from its bitrot state]]></description>
      <guid isPermaLink="true">https://blog.robur.coop//articles/miragevpn-ncp.html</guid>
      <pubDate>Mon, 20 Nov 2023 00:00:00 GMT</pubDate>
    </item>
    <item>
      <title>MirageVPN &amp; tls-crypt-v2</title>
      <link>https://blog.robur.coop//articles/miragevpn.html</link>
      <description><![CDATA[How we implementated tls-crypt-v2 for miragevpn]]></description>
      <guid isPermaLink="true">https://blog.robur.coop//articles/miragevpn.html</guid>
      <pubDate>Tue, 14 Nov 2023 00:00:00 GMT</pubDate>
    </item>
  </channel>
 </rss>
--- a/index.html
+++ b/index.html
@ -0,0 +1,179 @@
 <!doctype html>
 <html lang="en">
  <head>
    <meta charset="utf-8">
    <meta http-equiv="x-ua-compatible" content="ie=edge">
    <meta name="viewport" content="width=device-width, initial-scale=1">
    <title>
        Robur's blogIndex
    </title>
    <meta name="description" content="The famous root of the website">
    <link type="text/css" rel="stylesheet" href="https://blog.robur.coop/css/hl.css">
    <link type="text/css" rel="stylesheet" href="https://blog.robur.coop/css/style.css">
    <script src="https://blog.robur.coop/js/hl.js"></script>
    <link rel="alternate" type="application/rss+xml" href="https://blog.robur.coop/feed.xml" title="blog.robur.coop">
  </head>
  <body>
    <header>
      <h1>blog.robur.coop</h1>
      <blockquote>
        The <strong>Robur</strong> cooperative blog.
      </blockquote>
    </header>
    <main><a class="small-button rss" href="https://blog.robur.coop/feed.xml">RSS</a><p>The Robur blog.</p>
 <h3>Essays and ramblings</h3>
 <ol reversed class="list-articles"><li>
        <div class="side">
            <a href="https://reyn.ir/">
                <img src="https://www.gravatar.com/avatar/54a15736b37879bc9708c1618a7cc130">
            </a></div>
        <div class="content">
            <span class="date">2024-08-21</span>
            <a href="https://blog.robur.coop/articles/2024-08-21-OpenVPN-and-MirageVPN.html">MirageVPN and OpenVPN</a><br />
            <p>Discoveries made implementing MirageVPN, a OpenVPN-compatible VPN library</p>
            <div class="bottom">
                <ul class="tags-list"><li>MirageVPN</li><li>OpenVPN</li><li>security</li></ul>
            </div>
        </div>
    </li><li>
        <div class="side">
            <a href="https://blog.osau.re">
                <img src="https://www.gravatar.com/avatar/e243d18f97471424ca390e85820797ac">
            </a></div>
        <div class="content">
            <span class="date">2024-08-15</span>
            <a href="https://blog.robur.coop/articles/tar-release.html">The new Tar release, a retrospective</a><br />
            <p>A little retrospective to the new Tar release and changes</p>
            <div class="bottom">
                <ul class="tags-list"><li>OCaml</li><li>Cstruct</li><li>functors</li></ul>
            </div>
        </div>
    </li><li>
        <div class="side">
            <a href="https://blog.osau.re/">
                <img src="https://www.gravatar.com/avatar/e243d18f97471424ca390e85820797ac">
            </a></div>
        <div class="content">
            <span class="date">2024-06-24</span>
            <a href="https://blog.robur.coop/articles/qubes-miragevpn.html">qubes-miragevpn, a MirageVPN client for QubesOS</a><br />
            <p>A new OpenVPN client for QubesOS</p>
            <div class="bottom">
                <ul class="tags-list"><li>OCaml</li><li>vpn</li><li>unikernel</li><li>QubesOS</li></ul>
            </div>
        </div>
    </li><li>
        <div class="side">
            <a href="https://hannes.robur.coop">
                <img src="https://www.gravatar.com/avatar/25558b4457cf73159f5427fdf2b4a717">
            </a></div>
        <div class="content">
            <span class="date">2024-06-17</span>
            <a href="https://blog.robur.coop/articles/miragevpn-server.html">MirageVPN server</a><br />
            <p>Announcement of our MirageVPN server.</p>
            <div class="bottom">
                <ul class="tags-list"><li>OCaml</li><li>MirageOS</li><li>cryptography</li><li>security</li><li>VPN</li></ul>
            </div>
        </div>
    </li><li>
        <div class="side">
            <a href="https://hannes.robur.coop">
                <img src="https://www.gravatar.com/avatar/25558b4457cf73159f5427fdf2b4a717">
            </a></div>
        <div class="content">
            <span class="date">2024-04-16</span>
            <a href="https://blog.robur.coop/articles/miragevpn-performance.html">Speeding up MirageVPN and use it in the wild</a><br />
            <p>Performance engineering of MirageVPN, speeding it up by a factor of 25.</p>
            <div class="bottom">
                <ul class="tags-list"><li>OCaml</li><li>MirageOS</li><li>cryptography</li><li>security</li><li>VPN</li><li>performance</li></ul>
            </div>
        </div>
    </li><li>
        <div class="side">
            <a href="https://reyn.ir/">
                <img src="https://www.gravatar.com/avatar/54a15736b37879bc9708c1618a7cc130">
            </a></div>
        <div class="content">
            <span class="date">2024-02-21</span>
            <a href="https://blog.robur.coop/articles/gptar.html">GPTar</a><br />
            <p>Hybrid GUID partition table and tar archive</p>
            <div class="bottom">
                <ul class="tags-list"><li>OCaml</li><li>gpt</li><li>tar</li><li>mbr</li><li>persistent storage</li></ul>
            </div>
        </div>
    </li><li>
        <div class="side">
            <a href="https://hannes.robur.coop">
                <img src="https://www.gravatar.com/avatar/25558b4457cf73159f5427fdf2b4a717">
            </a></div>
        <div class="content">
            <span class="date">2024-02-13</span>
            <a href="https://blog.robur.coop/articles/speeding-ec-string.html">Speeding elliptic curve cryptography</a><br />
            <p>How we improved the performance of elliptic curves by only modifying the underlying byte array</p>
            <div class="bottom">
                <ul class="tags-list"><li>OCaml</li><li>MirageOS</li><li>cryptography</li><li>security</li></ul>
            </div>
        </div>
    </li><li>
        <div class="side">
            <a href="https://blog.osau.re/">
                <img src="https://www.gravatar.com/avatar/e243d18f97471424ca390e85820797ac">
            </a></div>
        <div class="content">
            <span class="date">2024-02-11</span>
            <a href="https://blog.robur.coop/articles/lwt_pause.html">Cooperation and Lwt.pause</a><br />
            <p>A disgression about Lwt and Miou</p>
            <div class="bottom">
                <ul class="tags-list"><li>OCaml</li><li>Scheduler</li><li>Community</li><li>Unikernel</li><li>Git</li></ul>
            </div>
        </div>
    </li><li>
        <div class="side">
            <a href="https://reyn.ir/">
                <img src="https://www.gravatar.com/avatar/54a15736b37879bc9708c1618a7cc130">
            </a></div>
        <div class="content">
            <span class="date">2024-02-03</span>
            <a href="https://blog.robur.coop/articles/2024-02-03-python-str-repr.html">Python&apos;s `str.__repr__()`</a><br />
            <p>Reimplementing Python string escaping in OCaml</p>
            <div class="bottom">
                <ul class="tags-list"><li>OCaml</li><li>Python</li><li>unicode</li></ul>
            </div>
        </div>
    </li><li>
        <div class="side">
            <a href="https://hannes.robur.coop">
                <img src="https://www.gravatar.com/avatar/25558b4457cf73159f5427fdf2b4a717">
            </a></div>
        <div class="content">
            <span class="date">2023-11-20</span>
            <a href="https://blog.robur.coop/articles/miragevpn-ncp.html">MirageVPN updated (AEAD, NCP)</a><br />
            <p>How we resurrected MirageVPN from its bitrot state</p>
            <div class="bottom">
                <ul class="tags-list"><li>OCaml</li><li>MirageOS</li><li>VPN</li><li>security</li></ul>
            </div>
        </div>
    </li><li>
        <div class="side">
            <a href="https://reyn.ir/">
                <img src="https://www.gravatar.com/avatar/54a15736b37879bc9708c1618a7cc130">
            </a></div>
        <div class="content">
            <span class="date">2023-11-14</span>
            <a href="https://blog.robur.coop/articles/miragevpn.html">MirageVPN &amp; tls-crypt-v2</a><br />
            <p>How we implementated tls-crypt-v2 for miragevpn</p>
            <div class="bottom">
                <ul class="tags-list"><li>OCaml</li><li>MirageOS</li><li>VPN</li><li>security</li></ul>
            </div>
        </div>
    </li></ol>
        </main>
    <footer>
        <a href="https://github.com/xhtmlboi/yocaml">Powered by <strong>YOCaml</strong></a>
        <br />
    </footer>
    <script>hljs.highlightAll();</script>
  </body>
 </html>
--- a/rss1.xml
+++ b/rss1.xml
@ -0,0 +1,84 @@
 <?xml version="1.0" encoding="utf-8"?>
 <rdf:RDF xmlns="http://purl.org/rss/1.0/" xmlns:rdf="http://www.w3.org/1999/02/22-rdf-syntax-ns#">
  <channel rdf:about="https://blog.robur.coop/rss1.xml">
    <title>The Robur&apos;s blog</title>
    <link>https://blog.robur.coop/</link>
    <description><![CDATA[The Robur cooperative blog]]></description>
    <items>
      <rdf:Seq>
        <rdf:li resource="https://blog.robur.coop//articles/2024-08-21-OpenVPN-and-MirageVPN.html"/>
        <rdf:li resource="https://blog.robur.coop//articles/tar-release.html"/>
        <rdf:li resource="https://blog.robur.coop//articles/qubes-miragevpn.html"/>
        <rdf:li resource="https://blog.robur.coop//articles/miragevpn-server.html"/>
        <rdf:li resource="https://blog.robur.coop//articles/miragevpn-performance.html"/>
        <rdf:li resource="https://blog.robur.coop//articles/gptar.html"/>
        <rdf:li resource="https://blog.robur.coop//articles/speeding-ec-string.html"/>
        <rdf:li resource="https://blog.robur.coop//articles/lwt_pause.html"/>
        <rdf:li resource="https://blog.robur.coop//articles/2024-02-03-python-str-repr.html"/>
        <rdf:li resource="https://blog.robur.coop//articles/miragevpn-ncp.html"/>
        <rdf:li resource="https://blog.robur.coop//articles/miragevpn.html"/>
      </rdf:Seq>
    </items>
  </channel>
  <item rdf:about="https://blog.robur.coop//articles/2024-08-21-OpenVPN-and-MirageVPN.html">
    <title>MirageVPN and OpenVPN</title>
    <link>https://blog.robur.coop//articles/2024-08-21-OpenVPN-and-MirageVPN.html</link>
    <description>
      <![CDATA[Discoveries made implementing MirageVPN, a OpenVPN-compatible VPN library]]>
    </description>
  </item>
  <item rdf:about="https://blog.robur.coop//articles/tar-release.html">
    <title>The new Tar release, a retrospective</title>
    <link>https://blog.robur.coop//articles/tar-release.html</link>
    <description><![CDATA[A little retrospective to the new Tar release and changes]]></description>
  </item>
  <item rdf:about="https://blog.robur.coop//articles/qubes-miragevpn.html">
    <title>qubes-miragevpn, a MirageVPN client for QubesOS</title>
    <link>https://blog.robur.coop//articles/qubes-miragevpn.html</link>
    <description><![CDATA[A new OpenVPN client for QubesOS]]></description>
  </item>
  <item rdf:about="https://blog.robur.coop//articles/miragevpn-server.html">
    <title>MirageVPN server</title>
    <link>https://blog.robur.coop//articles/miragevpn-server.html</link>
    <description><![CDATA[Announcement of our MirageVPN server.]]></description>
  </item>
  <item rdf:about="https://blog.robur.coop//articles/miragevpn-performance.html">
    <title>Speeding up MirageVPN and use it in the wild</title>
    <link>https://blog.robur.coop//articles/miragevpn-performance.html</link>
    <description>
      <![CDATA[Performance engineering of MirageVPN, speeding it up by a factor of 25.]]>
    </description>
  </item>
  <item rdf:about="https://blog.robur.coop//articles/gptar.html">
    <title>GPTar</title>
    <link>https://blog.robur.coop//articles/gptar.html</link>
    <description><![CDATA[Hybrid GUID partition table and tar archive]]></description>
  </item>
  <item rdf:about="https://blog.robur.coop//articles/speeding-ec-string.html">
    <title>Speeding elliptic curve cryptography</title>
    <link>https://blog.robur.coop//articles/speeding-ec-string.html</link>
    <description>
      <![CDATA[How we improved the performance of elliptic curves by only modifying the underlying byte array]]>
    </description>
  </item>
  <item rdf:about="https://blog.robur.coop//articles/lwt_pause.html">
    <title>Cooperation and Lwt.pause</title>
    <link>https://blog.robur.coop//articles/lwt_pause.html</link>
    <description><![CDATA[A disgression about Lwt and Miou]]></description>
  </item>
  <item rdf:about="https://blog.robur.coop//articles/2024-02-03-python-str-repr.html">
    <title>Python&apos;s `str.__repr__()`</title>
    <link>https://blog.robur.coop//articles/2024-02-03-python-str-repr.html</link>
    <description><![CDATA[Reimplementing Python string escaping in OCaml]]></description>
  </item>
  <item rdf:about="https://blog.robur.coop//articles/miragevpn-ncp.html">
    <title>MirageVPN updated (AEAD, NCP)</title>
    <link>https://blog.robur.coop//articles/miragevpn-ncp.html</link>
    <description><![CDATA[How we resurrected MirageVPN from its bitrot state]]></description>
  </item>
  <item rdf:about="https://blog.robur.coop//articles/miragevpn.html">
    <title>MirageVPN &amp; tls-crypt-v2</title>
    <link>https://blog.robur.coop//articles/miragevpn.html</link>
    <description><![CDATA[How we implementated tls-crypt-v2 for miragevpn]]></description>
  </item>
 </rdf:RDF>