[Fwd: Question: Will Fedora 10 Contain KDE 3.5.10 or Not?]

[Michael Schwendt]

On Wed, 06 Aug 2008 09:05:38 -0400, Jesse Keating wrote:

> The "too many updates" problem is something I've been trying to word so
> that others share my opinion that something is wrong here.  I haven't
> been able to effectively communicate what I perceive to be a problem.

 From the Fedora Wiki:

|    Fedora is a Linux-based operating system that provides users with
|    access to the latest free and open source software,

Always? Always "the latest" as in "a rolling release"? Or "the latest"
as in "what was fresh when Fedora N was being developed"?

|    by sticking close to upstream development teams,
|    Fedora often gets the latest software before anybody else. 

The former can be considered good, the latter is not necessarily
good. Especially when reading it like "Fedora includes snapshots of
unreleased software while other distributors spend more time on
testing". The user wants to know what applies to Fedora N-1. Is it
more stable and less fast-moving?

Fine if the next Fedora release will aim at including the latest
technology. Not so fine if an older still maintained Fedora release is
moved to somewhere between a previously tested/stable release and the
next Fedora, N+1.

Fine if package maintainers release bug-fix updates from time to time,
which include accumulated fixes for bugs filed in Fedora bugzilla. Not so
fine if they release each and every upstream version as a "stable" Fedora
update after a ridiculously short time in updates-testing and possibly
even 0 karma (or +1 self-voted). Minor upstream updates often touch more
code than one might assume. A minor update breaks something or introduces
a new bug. Another bug-fix update is needed, and so on. Version upgrades
sometimes lead to a dead end, if their subsequent bug-fix updates require
upgrades of build requirements. Better spend the energy on Fedora N+1.

Look at software in a Fedora release which is not updated at all! It isn't
bug-free either. The next Fedora is only six months away.

|    in a stable,
|    secure
|    and easy to manage form.

An upgrade a day keeps the user away.

Bodhi lists 3947 "stable" updates for Fedora 8 compared with 260 security
updates. Such a high number of updates and upgrades creates unnecessary
work for the users. Every week they are presented with another large
bunch of updates and need to find out how they are affected. For example,
by .rpmsave/.rpmnew madness, regressions, broken dependencies, the
necessity to "sync" configurations and customisations, being forced
to do manual downgrades, being forced to apply all updates in order to
rule out side-effects during trouble-shooting, silent API-breakage,
silent ABI-breakage (we've had it, too), incompatibilities introduced
by version upgrades *after* the final release.

I'm especially concerned about the wasted effort of doing several Test
releases before a final release and then throwing away all that work
gradually when more and more packages get replaced. Too many updates
receive negative karma. You ship a fix for one user and break something
for two other users.  No reason to believe that untested updates are
safe(r).

I agree with Thorsten in that _some_ updates are nice to have, e.g. a
fresh kernel from time to time (and particularly if it is used for
a respin). A kernel a week is too much, however.