Delays in package processing

[Ralf Corsepius]

On Thu, 2007-12-20 at 10:19 -0900, Jeff Spaleta wrote:
> On Dec 20, 2007 10:05 AM, Michael Schwendt
> <mschwendt.tmp0701.nospam at arcor.de> wrote:
> > That's easy. The spec %changelog entry already ought to explain why the
> > update is important. It's the many hundreds small/minor/unimportant
> > updates which fill the updates repository and drift away from the original
> > release of the distribution. It leads to a scenario where after firstboot
> > you are offered so many packages that you're annoyed.
> 
> The spoken question here is... are we doing releases the 'right' way
> for our target users?
Which target user audience?

As far as I am concerned (power user, developer): No, Fedora doesn't.

> Is our release and updates policy inconsistent?
Yes. I feel Fedora is applying rel-eng strategies which do not fit into
a "forward looking" distro aiming at early adoption of "leading edge
technology".

IMO, it's natural for a leading edge/early technology adopter distro to
see frequent updates. But also note that "leading edge" shouldn't mean
"instable" or "premature". Unfortunately I also feel this is what some
people (esp. some people at RH) seem to be wanting to treat Fedora as.

> If lots and lots of updates cause annoyance for fresh installed users,
> should we be making a bigger deal about point people to re-spins
> as an option?
Yes. I think, there should be regular respins of the disk-images.

Just consider the current situation: The FC8 *.isos already have become
obsolete and are not really worth downloading anymore. 

More generally speaking, /me thinks the concept of "gold image" has
reached its technical limits and won't lead much further. 
A bit far fetching, I think, Fedora needs bootimages supporting
networked installs directly from Everything/ and some means to cut isos
to for local reuse after installation.

>  Also assuming we can have client
> side tools which could be told to "update only package updates flagged
> as security related" on a daily basis from the network, people could
> grab a snapshot of the updates tree when they want to for anything
> non-critical.
No, it don't think this is useful. To low-bandwith users it's not the
number of packages, which are causing problems, it's the mass (Which
happen to originate from a handful of packages which normally are marked
security update) and from unreliability of the technology underneath
(such as mirrors out of sync, broken metadata, yum issues).

Ralf