Re: Debian desktop -situation, proposals for discussion and change. Users point of view.
Testing is usable. I used it through the whole development cycle of etch.
Bugs are unavoidable, you said it yourself. It's a matter of how many
problems you can accept.
Yes, bugs are unavoidable. However, testing is often in situation "whole
system broken" or "nearly useless". I see difference here; occassional
bug in desktop app is acceptable. Whole system unreliable is not acceptable.
In the Debian context, "Gnome" is a platform. It's not only software
that runs on top of libc6. Gnome represent dozens of libraries that are used
by hundreds of applications.
You can't just get the latest version and hope that it won't break
That should be verified in light of broad experience (I don't have any).
Does it happen often that GNOME version change breaks many things? The
only my try was to put GNOME 2.0 to Debian Woody (ugly GNOME 1.2), and I
If You mean to use the software from testing -You must first make it run
on stable without need for library upgrades. That is more similar to
backports job, than to testing.
You can't backport everything if you don't want to upgrade libraries. It's
simply not doable without rewriting the application.
I think majority of software _should_ build w/o problem with ordinary
libraries of maximum 2 years age. In my experience, apps are generally
happy if libraries are not older than that. Of course, shorter release
cycle could remove remaining problems in this order.
Next stable release of Debian will of course upgrade the whole platform,
including the versions, thus software would be happy for next 18 months.
Testing should simply be the place where _platform_ changes are shaken
out, not the "input buffer for the new software".
Actually sid is where the platform changes are done. And once they're OK,
they get moved to testing in a coherent manner. So testing should stay
This is just a wish, not a common experience.