On Thu, Oct 2, 2008 at 8:57 PM, Brion Vibber <brion(a)wikimedia.org> wrote:
The current theory is we'd like it to be easy to
mark things as needing
*more* review, but hard to mark things as needing *less* review.
So that probably means a split-level permissions model, perhaps with
distinct pre-review and super-review (to use the Mozilla term -- patch
reviews get "super-reviewed" by a core committer, or some such crazy thing).
It is, of course, an evolving system. :)
Yeah, currently I'm only marking things "fixme". How about only you
and Tim get put in the coder group, but that group only gives you the
right to mark resolved/ok, while everyone can mark fixme? Is there a
useful difference between resolved and ok, too?
Maybe the levels should be like:
* unreviewed: default
* fixme: anyone can change a commit's status to fixme, from any other
status (including reviewed/superreviewed)
* reviewed: anyone can change a commit's status from unreviewed or
fixme to reviewed . . . or maybe only from unreviewed? So this would
just mean "someone other than the committer has reviewed this". May
or may not be useful.
* superreviewed: only privileged users (you and Tim) can move to
superreviewed (of course this is a complete bastardization of the
Mozilla terms :) )
And there should be some prominent message saying which commit is the
last one that's scappable, i.e., preceded by only a) superreviewed
commits and maybe b) commits to extensions (or even files, like
DatabasePostgres.php) not used by Wikimedia.
A couple of things to think about:
* Should it be possible for ordinary people to clear "fixme" in some
way? Maybe ordinary people should be able to change fixme to
unreviewed or reviewed, *but* only if they added the fixme status
themselves; and only you/Tim should be allowed to change it to other
statuses. Maybe we could try this way out, and if there's too much
noise, we could change it so anyone can resolve fixmes.
* Should there be some way of indicating who's reviewed the commit?
How prominent should this be in the interface? Should multiple people
be able to mark a commit reviewed independently? Maybe we should
track fixme/fixed status separately from review status altogether?