On Wed, 2008-10-22 at 20:46 +0200, Patrice Dumas wrote:
On Wed, Oct 22, 2008 at 06:25:11PM +0100, David Woodhouse wrote:
>
> Of _course_ it's not about specific cases. That would require actual
> substantiated facts which can be verified or disproved; not just
> handwaving and ranting.
I answered to Rex giving a specific case, regarding ConsoleKit and
various dm breakage. You want the bugzilla entries or you trust me?
/me looks back...
On Wed, 2008-10-22 at 15:06 +0200, Patrice Dumas wrote:
It is not that simple. The distinction between fedora and upstream
in
many cases is rather fuzzy. For example the hal/dbus/consoleKit/*Kit is
put in fedora when it is at early development stages, and many of the
developpers involved in these are also involved in fedora. And the
changes are pushed in fedora without taking seriously the backward
compatibility issues. For example wdm and xdm (and slim) are broken
since consolekit replaced pam_console, and although there is a rather
simple solution to integrate those dm it has never been planned and it
is still not fixed, though a fix exist for months. As long as it worked
in gdm it was fine for fedora. You can tell, hey, xdm, wdm and slim can
copy what gdm does, but when the solution implemented in gdm is specific
and not consistent with the previous designs, it is not so easy. Fedora
controls some upstream so can do anything in these, but doesn't control
all of them.
Yeah, that sucks. We should do better than that.
Although I've always been a little dubious about our 'Feature Process',
it does seem that it addresses this kind of problem. For the feature to
reach 100% completion, it should obviously involve fixes for the other
display managers. And there is a 'reversion plan' in case we don't
manage to complete the feature in time for the release.
From what you say above, it sounds like the ConsoleKit feature should
be
declared incomplete, and we should be reverting it unless the feature
owners finish the job.
--
David Woodhouse Open Source Technology Centre
David.Woodhouse(a)intel.com Intel Corporation