question on packaging and QA netiquette

[Michael Schwendt]

On Fri, 30 Jul 2004 23:37:38 -0400, Ed Hill wrote:

> I have a few questions about Fedora packaging netiquette.  Just for the
> sake of discussion, lets assume the following circumstances:
> 
>  - you're a strictly part-time packager but nonetheless have put 
>      some effort into your few packages and are prepared to 
>      support them as best you can
>  - you have a small audience (perhaps a few dozen users)
>  - unfortunately, none of your users have the time or skill 
>      to do Fedora QA

Users of a particular piece of software are the proper guinea pigs for
your packages as they should be capable of telling whether the software
works. They might not be able to recompile src.rpms and comment on
low-level technical packaging issues. But there are multiple layers of
QA, and test-driving prebuilt binary packages is one such layer.

>  - you'd like to see your package(s) in a repository somewhere 
>      so that a convenient "yum install $PACKAGE" works for the 
>      new (and often least-skilled) users but you're thinking it 
>      may take a long time--or perhaps never happen...
> 
> So what can you do attract QA attention and maybe get approved? 

An active target group can demonstrate that there is interest in the
package and can declare that the built software works (or is in daily
use). Users can do lobbying and convince the upstream project of providing
signed tarballs. Users can point out of the software is included with a
different big distribution already for a longer time, e.g. Debian
GNU/Linux.

> Is it
> considered "poor form" to ask others to QA your package(s) in return for
> doing QA on theirs?

No. It ought to work like that. And actually, that is recommended on the
fedora.us packaging policy package. With the basic framework at fedora.us,
there are no explicit QA requirements other than what's documented in the
QA checklist and the other Wiki pages. Packagers and potential users are
encouraged to utilize the "testing" and "unstable" repositories much more.
 
> Backing up a bit, is it even reasonable to try to get your package(s)
> included in the first place?

Assuming that the packaging community continues to grow and more packagers
work towards reaching "trusted" level, it is reasonable.

On the other hand, now knowing what will happen to Fedora Extras is not
encouraging.

> When is it or isn't it a good idea to try for inclusion?

The former has too many answers (like the software is popular,
ground-breaking, innovative, not popular yet but a potential replacement
for a Core package, would benefit from an increased userbase and more
testing ...).

If the package content raises questions, e.g. on the proper location and
purpose of included files or in case simple tests don't seem to work, that
is demotivating for the reviewers. It is also demotivating if the package
fails to build, or if the packager has derived the package from an
existing one made by somebody else and does not seem to be familiar with
what's done in the spec file.

> OK, maybe thats enough questions for a friday night...  ;-)
> 
> Ed
>