IRC log

<wikibugs> RESTBase, Services, Math, Mathoid: Make the Math extension usable without RESTbase or Mathoid installed - https://phabricator.wikimedia.org/T121173#1872277 (Physikerwelt) @Jdforrester-WMF what do you mean by "on our private wikis"?
<physikerwelt_> gwicke: ping
<+gwicke> physikerwelt_: pong
<physikerwelt_> it seems that https://phabricator.wikimedia.org/T121173 is a major problem
<physikerwelt_> I'm unsure what needs to be done
<physikerwelt_> for private wiki's the situation is simpler than ever before
<physikerwelt_> just add two lines of config no binaries, no installation nothing
<+gwicke> my understanding is that people are a) not clear about the long-term math roadmap & the move to mathoid, and b) still have doubts about the feasiblity of third party users installing services
<+gwicke> there is also an element of "do we need to break it just yet?"
<+gwicke> which I think is understandable
<+gwicke> I have recently done some work aimed at b)
<+gwicke> adding mathoid to this is easy
<physikerwelt_> I set up a restbase instance outside of wmflabs
<physikerwelt_> it has good performance about 4ms for formulae that have been rendered before and 50 ms for others
<physikerwelt_> this was form within the local network
<+gwicke> ultimately, I think nobody can seriously ask to be backwards compatible forever if that would stifle progress towards more accessible / semantic & scalable math rendering
<+gwicke> it's more about laying out a path for a smooth transition
<+gwicke> none of these arguments are about the tech itself
<+gwicke> it's about meatspace and third party users
<physikerwelt_> https://gerrit.wikimedia.org/r/#/c/255950/
<physikerwelt_> that's the change... I cold add a config option to go back to texvccheck
<physikerwelt_> but I want to retire texvccheck completly
<+gwicke> it might be nice to have a deprecation period first
<+gwicke> tell everybody clearly
<+gwicke> demonstrate nice packaging including mathoid
<+gwicke> and then retire texvccheck
<physikerwelt_> gwicke I think it's pointless to pack mathoid without restbase
<+gwicke> yeah, I don't think anybody is asking for that
<physikerwelt_> you really want to use mathoid within restbase
<+gwicke> it's more about still having the option to use the texvc rendering
<+gwicke> so that existing installations continue to work for a bit longer
<physikerwelt_> they do work
<physikerwelt_> you just have to change the config to skip texvccheck
<physikerwelt_> which is pointless because texvc does the same checks anyhow
<+gwicke> oh, okay
<+gwicke> is that the default setting?
<physikerwelt_> no
<+gwicke> for backwards compatibility that might make sense
<physikerwelt_> the problem is that the default settings do not make much sense at the current state
<+gwicke> basically, provide defaults that keep things working for users upgrading their old install
<+gwicke> and provide a new shiny install method for the entire thing including mathoid that explicitly configures things to use that instead
<+gwicke> I got pretty far earlier this week: https://phabricator.wikimedia.org/T92826#1866757
<physikerwelt_> gwicke the shiny install method depends on restbase installation
<physikerwelt_> I think it works quite well
<+gwicke> yup, that's already included
<physikerwelt_> the missing bit how to add it as a service
<+gwicke> and mathoid is just another service to include along Parsoid & RB
<+gwicke> basically, a case of adding some lines to the config
<+gwicke> here: https://github.com/gwicke/mediawiki-node-services/blob/master/config.yaml#L43
<physikerwelt_> I do know that
<+gwicke> anyway, this will take a bit of time until people are trusting this
<+gwicke> so it would be good to keep existing users happy until then
<+gwicke> while also clearly saying where math is headed