What do we think of this?

[Michael Schwendt]

On Tue, 27 Mar 2007 10:47:05 -0400, seth vidal wrote:

> On Tue, 2007-03-27 at 10:32 -0400, Greg Dekoenigsberg wrote:
> > http://www.oreillynet.com/linux/blog/2007/03/where_fedora_went_wrong.html
> > 
> > Michael Schwendt has made a spirited and passionate response (thanks, 
> > Michael) to Caitlyn's comments.
> > 
> > Is there any substance to these comments?
> 
> Well, rawhide is frequently broken/out of sync. I don't think we expect
> but so much there. 
> 
> I think Michael has a point about deps not being broken in any given
> repo. However, due to the different mgmt/deployment practices/behavior
> b/t extras and core I know there have been cases where one is out of
> sync with the other for a little while and then it takes the mirrors X
> amount of time to catch up - but in the mean time the user has a broken
> dep dangling. 

The term "broken dependency" is used differently here, though, which
creates a blurred picture of the primary complaints.

It is not [just] referred to the plain unresolved RPM dependencies that
made it impossible for ESR to install something.

The term is also used to refer to all sorts of bugs in packages for a
stable dist, especially regression found after an update or after a
dist-upgrade.

The message in short: no bug whatsoever ought to find its way into the
packages in the public repositories. Bugs in the packages <=> sloppy
repository management <=> insufficient testing.

It is accumulated grief, expressed as general criticism of Fedora's
testing. Personal experience, a friend's experience, everyone of those who
complain loudly could tell of unpleasant memories with one release of
Fedora, even if it's just single bug in something as old as FC5 or even
FC2. For one it's an update that broke Exchange Connector, for another one
it's a kernel panic after a kernel update (and with poor trouble-shooting,
e.g. no attempt at booting the previous kernel). Even another one has
built up a lot of hate for Yum (its slowness or GUIs), RPM (database
corruption anyone? [1]), trouble with proprietary drivers, or the
FOSS/proprietary/non-free separation of repositories.

[...]

With regard to the mirroring problems, I don't understand yet why we push
FE-6 and FE-5 daily (on average) even when the updates don't contain any
important fixes, such as security fixes. Some packages apparently are
updated daily with either snapshots from VCS or with minor updates. IMO
this is over-ambitious.

-- 
[1] Unstable RAM chips or unstable hardware are one way to reproduce it,
but else?