On Tue, Jul 29, 2014 at 4:29 PM, Steven Walling swalling@wikimedia.org wrote:
I'm glad we merged this sooner rather than later, since it means there are fewer Flow-specific overrides on top of mediawiki.ui and it forces us to have a cross-team discussion.
Flow doesn't use overrides on top of mediawiki.ui (except for mw-ui-button, which only started recently). We wrote all mw-ui components independently of Core, using a flow-ui prefix -- including a rewrite of mw-ui-button, which I felt had become too messy and confusing. When I felt that the flow-ui versions had reached a certain level of maturity, I'd put them for review into Core's gerrit as mw-ui components.
On Tue, Jul 29, 2014 at 11:16 PM, Erik Moeller erik@wikimedia.org wrote:
I agree we have to be wary about consistency. I'm not sure putting these kinds of changes on the release train incrementally is the way to go -- perhaps having a test instance in Labs run a WIP changeset or branch til we're happy with it, including things like RTL testing?
This is specifically the reason I was using Flow as the testbed to develop and iterate upon the mw-ui components. However, I think my team jumped the gun on putting this into Core without fully testing all of the use cases. I take responsibility for this, as I should have been more careful to a) include i18n people in the design mocks, and b) thoroughly test any current uses of affected mw-ui components being moved to Core.
Going forward, I think we will need to shift from a single +2 in gerrit, to a +1 from several teams (perhaps Flow, Growth, Mobile, and Language), before approving a merge to Core. For these mediawiki.ui design goals to succeed, we'll need to make sure we have our primary bases covered via these teams.
--Shahyar