In T53294#2191188, @matmarex wrote:

In T53294#1064524, @Dragons_flight wrote:

3. A race condition that causes near simultaneous instances of the AbuseFilter to corrupt each others stats.

The third problem is the dominant one for the highly active wikis and can lead to really silly results.

I don't think this can reasonably account for the reported values being wrong by a factor of ten or more.

I'm pretty sure that the number of conditions is counted globally, for all filters together, then treated it as if it was the per-filter number, which obviously results in off-the-wall values. I wonder what @werdna was smoking when committing rSVN52738. ;)

I believe the initial version of this (now removed) was reading three entries from memcache and then setting three updated entries per filter (as well as a few global keys that were updated per edit rather than per filter).

The underlying technical problem is that uploads proceed in two steps. First the file is uploaded to the server and then the text is processed and saved to the database. We have a hook into the upload process, but at that point in software the uploaded text is not yet available.

The AbuseFilter currently runs all regex in unicode mode. That apparently solves problems for some languages but not all, including the devanagari examples offered here.

One can already write a filter that checks for something like

Change 201171 that I wrote includes this functionality as part of its changes:

@chasemp: You mention changing where such requests are sent. Do I need to do anything else to ensure that this request is seen by the appropriate people? I'm guessing that it is fine here, even if it got started in a suboptimal way, but I just wanted to double check.

This is really three separate bugs that all lead to per filter stats corruption:

Towards getting the details right, I'd like to verify my understanding of how the Captcha functions with regards to URLs on Wikimedia sites.

I've started adding running some simple tests to explore this a bit more. Is there a natural place to discuss link statistics and behavior? Somewhere on Mediawiki.org? I worry that putting too much detail here will be a distraction.