When dealing with mw-js-deprecate, I noticed that searching for less popular ones with mwgrep that all their hits actually come from closed wikis, such as usability.wikimedia.org, wikimania2008.wikimedia.org and nl.wikinews.org.
This got me thinking whether it makes sense to disable the WikimediaEvents extension on those wikis. And perhaps other extensions as well?
| Subject | Repo | Branch | Lines +/- | |
|---|---|---|---|---|
| Disable WikimediaEvents extension on closed wikis | operations/mediawiki-config | master | +11 -3 |
@Milimetric Does that mean Analytics support and/or is neutral about the proposal to disable the WikimediaEvents extension on closed wikis? The impact will be that the affected schemas will no longer represent "all wikis", which may or may not be important, and may be considered either a good thing or a bad thing.
From my own perspective, it'll avoid false positive warnings logged from wikis that are read-only, as such the warnings should not (or can not) be addressed.
For other schemas, I can similarly imagine it being a good thing to not record new data from closed wikis. For example: Search satisfaction, Navigation Timing, and others. However, I don't know if we're worried about them not matching global numbers for other things. For example, right now the number of edits (as recorded server-side) should roughly match the Edit schema hit count. (Yes, closed wikis can have edits. Rarely, but still. Mostly "global"-type maintenance by stewards and clean up for global user pages). And Navigation Timing should roughly match page views (as recorded by Varnish) divided by 1000. This would no longer be the case if we stop instrumenting closed wikis.
We could potentially try a much wider initiative to simplifying closed wikis (e.g. stopping other metrics as well). So the question is whether it's okay to remove just WikimediaEvents, but keep other EventLogging-sources (at least for now).
We're neutral. It's ok if the global numbers don't match. We monitor EventLogging's health based on how it's able to process what it gets, not on what we expect.
Personally I agree it's good to simplify closed wikis and stop logging from them. By logging we usually trade off a little privacy for hopefully more benefit of improving the site. But in this case, general logging seems to have little benefit and because of the low traffic on closed wikis generates smaller buckets in which it's harder to hide from a privacy perspective. So I support the wider initiative you talk about.
Sorry to untag so quickly, I confess I didn't fully think this through. Thanks for the context.
Change 342148 had a related patch set uploaded (by Krinkle):
[operations/mediawiki-config] Disable WikimediaEvents extension on closed wikis
Change 342148 merged by MaxSem:
[operations/mediawiki-config] Disable WikimediaEvents extension on closed wikis