On Thu, Jul 24, 2014 at 7:00 AM, Moritz Schubotz physik@physikerwelt.de wrote:
Hi Bryan,
thank you very much for your contribution to the Math extension and the associated vagrant role.
:) You're welcome. Thanks for being dedicated enough to this problem to work on it as a volunteer.
I have tried to write a patch for operations/puppet https://gerrit.wikimedia.org/r/#/c/148836/ but that's far away from finished. A lot of configuration and server assignment has to be done.
It looks like your patch is getting some feedback this morning. \o/
I searched for the string parsoid in the config files and found a lot of entries, that seem to specify how parsoid is deployed to individual servers. It thought it makes little sense to copy that.
I'm not sure what sort of config Mathoid needs. In general, cluster config uses feature flags like $wmgUseParsoid that are set in InitialiseSettings.php. Then the feature flags are used in CommonSettings.php to guard setting "LocalSettings.php" config values for wikis where the feature flag is enabled. Both prod and beta use the settings from this file. CommonSettings-labs.php can be used to change settings from the defaults used in production to match beta specific needs.
I had the impression that prior to this change http://git.wikimedia.org/blobdiff/operations%2Fmediawiki-config/173851ef6283... it was possible to enable the MathML rendering mode on betalabs. The comment "These variables point to non-WMF servers by default." is not correct. $wgMathMathMLUrl = "http://mathoid.testme.wmflabs.org" is hosted in the eqiad cluster.
For the purposes of production configuration, something hosted in wmflabs.org should be treated the same as something hosted at nsa.gov. We don't want production to talk to labs ever. First there are privacy concerns and second Labs will melt if you turn even a tiny fraction of enwiki's traffic towards it.
But it showed that there things on the beta cluster are different from vagrant https://bugzilla.wikimedia.org/show_bug.cgi?id=66516 describes that the MathML tags are filtered out somehow.
I get MathML rendering of the formula on http://deployment.wikimedia.beta.wmflabs.org/wiki/User:Anomie/Sandbox, but I'm not sure what that proves or disproves in relation to your bug reports.
Bryan