Core vs. Extras (was: Re: rawhide report: 20050405 changes)

[Michael Schwendt]

On Tue, 5 Apr 2005 16:14:41 -0400, Gregory Maxwell wrote:

> Here are some reasons that if I were a package, I might not want to be
> in extras:
> 
> 1. Much smaller audience (lots of people do install everything in
> core, but not so with extras today)

Moot point, as laziness is the primary reason why Joe User chooses an
everything-install in the fear that manual selection of packages would be
to complicated or time consuming. Despite the availability of tools like
Yum, it's still considered too inconvenient to add missing pieces after
installation (and system-config-packages is a dead end with regard to
adding software to an up-to-date FC). And do those people only install
everything, or do they also use everything? Where is the benefit of users
who install everything but use only a fraction of the packages? For most
of the Extras users it makes no sense to install every package from
Extras.

> 2. More difficult distribution model (if someone burns me the DVD ISO
> redhat provides it has all of core, but none of extras)

Does the same "someone" also burn complete Fedora Core Updates onto a
separate DVD? That could also be done with a mirrored snapshot of Fedora
Extras today. But as pointed out above, Joe User most likely doesn't want
another DVD of packages from which to use only a few.

> 3. 'Less reliable'source: When I get a distro from redhat, I'm getting
> it from people I trust.. When I get extras it's coming from a bunch of
> people on the internet I dont know. Now it's probably true that core
> doesn't have that much more redhat oversight on low profile packages,
> but it's a perception issue.

That's why fedora.us started with a mandatory QA process which included
reviews and approvals by two different persons after verification of
source tarball checksums and the recommendation that they are to be
compared with packages from big/well-known distributions, too. But I
agree, unless a packager follows the source code changes done by the
upstream project with every release, there is the risk that an upstream
developer goes wild and tries to introduce malicious code if it's not done
by a hacker who breaks into a download server and manages to fool the
packagers and the community. If you believe in threats like that, you
belong to the target group of Fedora Extras commits-list, where you
can observe CVS commits.

> 4. If a package is broken in core is known broken it has some
> potential for holding up a release. Not so for extras, so there is
> more incentive to fix a package.

Well, in a community project, anyone who has strong interest in a package,
which is found to be broken, can volunteer as the one who provides a quick
fix.

> 4. Almost no build synchronization.  When is a package in extras
> guaranteed to build on a new version of FC?  ... Never. 

Bad. Unless I'm misinformed, hopefully we still plan to have FC4 Extras
ready together with the release of FC4 or shortly after. For that we would
need a mass rebuild at least for FC4 Test3, though, to catch remaining
packages which don't compile or fail at run-time and which have not been
checked by their owners due to lack of a Rawhide or FC4 Test installation.
I also think there are a very few package owners still missing.

> It's possible
> that a package will effectively drop out of core without any conscious
> decision.  What if a user depends on that package, upgrades to FCn+1,
> and finds that it no longer works?  This is far less likely to happen
> 'on accident' for packages in core.

Fedora Extras Development is used to prepare packages for FCn+1,
currently. And once more, a user, who _depends on a package_, should
invest some time in making sure the package does work in FCn and continues
to work in FCn+1.

-snip-

> All of these issues only matter for packages that are coming out of
> core.. For a package that never was in core, it is less of an issue...
> being in Extras is better than nothing at all.

Packages, which come out of Core and effectively are "dropped on the
floor", need to be picked up by the Fedora community.  And Fedora Extras
development is open enough to allow for participation, and monitoring
of development done by others.
 
> Perhaps a solution would be to define a subset of extras that belong
> to a set of more stable/more important.

Start with a set of packages which you think are not stable or not good
enough. File a bug report for each package.