Several people had a meeting last week to start answering the question:

What do our SOA plans look like for the next fiscal year? How much

of this work will be spent refactoring MW Core, vs. augmenting with new, non-MediaWiki framework(s)?

Here's my summary of what we agreed on last week and what we still need to work out. (Participants included WMF engineers, Wikia engineers, and Markus Glaser.)

Agreed: * Modularity is a good thing * WMF will start inviting a Wikia engineer to the weekly "scrum of scrums" meeting * Wikia will have an all-team meeting in May in SF - will then compare notes with WMF about what toolset they're choosing for WikiFactory & the new mobile article page prototype

(Please also see lines 240-252 of the full notes for SOA risks and deliverables we'll need from service providers: http://etherpad.wikimedia.org/p/soa-kickoff )

Undecided: * what parts of MediaWiki will be refactored to follow in Parsoid's footsteps, who is going to do that, and when * what the working group's concrete targets would be, and who would be in it * what languages to write services in * whether Rashomon will only have a REST API or a PHP one as well * how to communicate between WMF and Wikia (wikitech-l? a different list? lightweight IRC meetings?) * who is Wikia's negotiating partner at WMF, who can authoritatively decide which languages/tools MediaWiki will use

Given this, I'll continue working on architectural guidelines that prescribe modularity in general but don't specifically call for SOA in all new features.

Thanks.