One example I&#39;ve seen of md5&#39;s failing is in common short vandalism phrases. The most obvious of these is removing all text, or single vulgar words.<div><br></div><div>Here&#39;s a scenario:</div><div><ul><li>User A vandalizes an article by replacing it with &quot;poop.&quot;</li>

<li>User B restores the article.</li><li>Some time passes....</li><li>User C vandalizes an article by replacing it with &quot;poop.&quot;</li></ul><div>User C isn&#39;t really reverting B&#39;s edit. You may be able to guess that since A was reverted, C must not be reverting, but this logic can be tricky and wrong.</div>

<div><br></div><div>In practice, I&#39;ve been able to catch most of these instances by a) ignoring md5 reverts outside a certain window of revisions and b) ignoring md5 reverts whose replace lots of text with very little text.</div>

<div><br></div><div>Also, this situation occurs regularly, but nowhere near 37%.</div><div><br></div><div>-Shilad</div><br><div class="gmail_quote">On Fri, Jun 29, 2012 at 11:15 AM, Benj. Mako Hill <span dir="ltr">&lt;<a href="mailto:mako@mit.edu" target="_blank">mako@mit.edu</a>&gt;</span> wrote:<br>

<blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex">&lt;quote who=&quot;Aaron Halfaker&quot; date=&quot;Wed, Jun 27, 2012 at 04:39:30PM -0700&quot;&gt;<br>
<div class="im">&gt; I&#39;m confused by your explanation.<br>
&gt;<br>
&gt; How is it possible that this 37% of revisions that are detected as reverts<br>
&gt; via a md5 hash are not considered reverts by (I presume) humans?  Can you<br>
&gt; give a common example?  By definition, identity revert revisions represent<br>
&gt; an exact replica of a previous revision in an article and, therefore,<br>
&gt; should discard any intermediate changes.  What definition of &quot;revert&quot; are<br>
&gt; you using that the md5 hash method does not satisfy?<br>
<br>
</div>Also, I can&#39;t tell from either the paper or the conversation here: Are<br>
Are you limiting this to edits that are separated by an revisions with<br>
identical hashes by only one edit? When you do that, things become a<br>
bit more complicated.<br>
<br>
And are you sure your human coders aren&#39;t just relying on edit<br>
summaries? Like Aaron, I&#39;m having a hard time imagining a situation<br>
where revisions go HASH-A =&gt; HASH-B =&gt; HASH-A that shouldn&#39;t be<br>
treated like a revert and think tend to think this sounds more like<br>
fallible than broken tools. If the user doesn&#39;t *know* or think they<br>
are reverting an edit, it seems wrong to *not* to call that a revert.<br>
<br>
Later,<br>
Mako<br>
<span class="HOEnZb"><font color="#888888"><br>
<br>
--<br>
Benjamin Mako Hill<br>
<a href="mailto:mako@mit.edu">mako@mit.edu</a><br>
<a href="http://mako.cc/" target="_blank">http://mako.cc/</a><br>
<br>
Creativity can be a social contribution, but only in so far<br>
as society is free to use the results. --GNU Manifesto<br>
</font></span><br>_______________________________________________<br>
Wiki-research-l mailing list<br>
<a href="mailto:Wiki-research-l@lists.wikimedia.org">Wiki-research-l@lists.wikimedia.org</a><br>
<a href="https://lists.wikimedia.org/mailman/listinfo/wiki-research-l" target="_blank">https://lists.wikimedia.org/mailman/listinfo/wiki-research-l</a><br>
<br></blockquote></div><br><br clear="all"><div><br></div>-- <br>Shilad W. Sen<br>Assistant Professor<br>Mathematics, Statistics, and Computer Science Dept.<br>Macalester College<br><a href="mailto:ssen@macalester.edu">ssen@macalester.edu</a><br>

651-696-6273<br>
</div>