On 22/03/07, Manuel Schneider <manuel.schneider(a)wikimedia.ch> wrote:
Before he wrote this mail he already yelled at me in
the chat.
Not so. I came on to IRC and advised you of the situation. When you
failed to answer the question I posed, and failed to acknowledge the
mistake, I made it explicitly clear what the situation is.
Manuel, we have conventions for working on the code base which are a
lot looser than a vast number of open source projects which are
designed for "release", but we still have a release process, and it
ought to be followed in order to help combat regressions or serious
bugs entering the stable branches. The team at large, and especially
myself, have learned this the hard way.
You claimed that the fix in question was "critical" when it was not,
and gave no indication that you intended to undo the changes. You then
stated your intention to ignore my advice, which is completely
inappropriate and thoroughly obnoxious. [We got off on the wrong foot,
but it doesn't mean we have to continue running on it.]
All new committers should have read
http://www.mediawiki.org/wiki/Commit_access - I would have thought
that making this mistake implied you did not read this information,
which is frankly unacceptable. We are exceptionally grateful for any
and all coding help, but it's all about gauging trust and competence,
and if you're asked to read a document, then it is a basic courtesy to
do so. [If the existence of this page was not made clear to you, then
I apologise.]
I've made a lot of mistakes during my involvement with the development
team, including a nasty merge of broken functionality into a release
branch prior to point-release time. This left a security vulnerability
in an old released version of the software. We no longer have such
gaping and fragile release policies, and so these need to be applied
and enforced consistently.
I'm certainly not calling for anyone to have their privileges revoked,
so don't make that assumption, but what I do think needs to be said is
that people need to stop, read up on how we operate, and appreciate
that when you dabble in the core code, you're no longer exclusively on
your own (extensions, some i18n) turf, and if you break a release,
then there is a problem.
Consistency of released "stable" code is what the release manager's
job is, and we only have one. Don't let's give him any additional
stress.
Rob Church