Bad package selection practices in Fedora packages

[Joel Rees]

On Wed, Jan 4, 2012 at 10:35 AM, Adam Williamson <awilliam at redhat.com> wrote:
> On Wed, 2012-01-04 at 09:52 +0900, Joel Rees wrote:
>
>> I think the error here is less in the coding in packages than in the
>> design of the default system specifications, specifically the package
>> selection.
>>
>> The errors gnome has committed would seem to be off-topic here, unless
>> the Fedora community still needs more evidence of how far gnome has
>> gone evil, or still needs more fuel for the debate about
>> whether/when/where said evil should be considered necessary.
>>
>> The Fedora community should be able to decide which packages to group
>> together for which installs.
>
> The desktop team decides what packages go into the desktop spin, and
> that's the same group of people as maintain GNOME, and many of them are
> upstream GNOME developers.
>
> Their stated position on tracker - it's been brought up on the desktop
> list before, see
> https://lists.fedoraproject.org/pipermail/desktop/2011-September/007377.html - is:
>
> "Yeah, tracker is a required dependency of gnome-documents now. As
> Bastien says, we should try to identify and fix possible bugs and
> resource issues upstream."

Had to go down a few posts to see that. But thanks for the broader thread.

> i.e., Tracker is now part of GNOME and they don't intend to disable it
> by default, but its default configuration could be adjusted (there was
> some discussion in the thread of doing this), and resource consumption
> bugs should be reported upstream and fixed.

https://lists.fedoraproject.org/pipermail/desktop/2011-September/007395.html

"Hmm, so investigating this further with strace it appears to me that
tracker is trying to make use of the kernel in a way it shouldn't. Or to
put this another way: our infrastructure (the Linux kernel) isn't ready
for tracker yet."

And yet Poettering proceeds with talking about how to shim Tracker in,
and the thread gets lost in wondering how to push the development of
fanotify ahead so it do what inotify can't. No evidence that anyone is
considering whether there is a fundamental conflict in requirements --
security vs. convenience, expecting context-free performance when
executing an unrestricted grammar on a very large finite state
machine.

I suppose I have to go to the gnome lists and raise Cain about this
kind of fundamental mis-engineering?

If I had any influence here, which I don't, I'd be pushing for Fedora
to kick gnome3 back to the curb and put its weight behind the gnome2
fork.

--
Joel Rees