Ralf Corsepius wrote:
* Many (most) packages get pushed without testing. I consider people
who
believe package to see tested in "testing", to be in error.
To me, "testing" isn't much more but a delay queue.
Good point.
* Some maintainers ignore feedback on "packages in
testing".
Indeed, and the proposed policy would do absolutely nothing to stop that!
* Some packages will never be "100% stable" (e.g. the
kernel, gcc, kde).
Right, and especially for those, being able to rush out a regression fix
without delays is important!
* The bugs with the highest visibilty are packages adding broken
package
dependencies. The fact these make it into the repos is a defect of
Fedora's infrastructure.
In fact the fact they happen at all is often due to infrastructure/procedure
issues, e.g. buildroot overrides are a mess. (Rel-eng now recommends using
special build tags for the big updates, we'll see whether that'll help for
KDE or just cause more chaos.)
But yes, the lack of automatic depchecking is also a big source of trouble.
* Banning direct pushes does not prevent Fedora from being hit by bad
SW.
+1
I think this FESCO is heading the wrong way.
The real problem is not "broken updates", the problem is "badly
maintained packages", "careless maintainers", "overzealous
developerss"
and defects of Fedora's infrastructure.
Indeed. This is a social problem (and yes, the lack of automated QA is also
an issue, but people are the main factor), it needs a social solution, not a
technobureaucratic one!
Particularly delicate are cases when "upstream" and
"maintainer" are
identical.
Well, I'm not sure this is a big factor in this particular case. The
conflict of interest is more apparent in other situations, e.g. good luck
getting a broken upstream default fixed if upstream is also the Fedora
packager. (See e.g. spatial Nautilus, for which it took years for the
maintainer to realize it sucks, it's now finally getting disabled by default
in F13 according to the GNOME 2.30 feature page.)
Kevin Kofler