Hello,
On Friday 05 April 2013 08:42:54 Sune Vuorela wrote:
> On 2013-04-05, Martin Steigerwald <
Mar...@lichtvoll.de> wrote:
> > What was the reason for
qt-kde.debian.net then? From the page:
> Some people thinks experimental is hard. Other people dislike doing the
> copyright-documentation that is required for the official archive. And
> qt-kde.d.n also could allow people who isn't a DD/DM to put packages
> there.
Well, qt-kde.d.n as distribution channel has probably outlived its purpose
even if it can still be useful sometimes.
The rest of the mail will be a bit OT for this list, however, in my opinion,
users might still be interested what some of the challenges are with KDE
packaging. Actually, these are the main things which turn (have turned) me off
from KDE packaging these days.
IMO, it is very complicated to maintain anything that is more like 5-10 highly
coupled source packages in Debian. You have to spend so much time on internal
development infrastructure (constantly) that little time (or motivation)
remains to do actual packaging changes. And as far as I know, KDE approaches
100 source packages, so do the math (funtunately, the number of core packages
is low). Especially, it is very expensive (in terms of both time and knowledge
required) to start KDE packaging for the first time or "resume" work after
longer time of inactivity.
I wish there was some "Continuous integration" for KDE packaging which took
the load of:
* Package building and dependency management.
* Package uploading to development repository for testing.
* Automatted Lintian reports and other Q/A.
* Any other repetitive, boring but useful tasks.
experimental is not suitable for that because it's a distribution channel
rather than a development one. Nowadays experimental latency is OK for
distribution channel however it is very slow for development (anything more
than 5-30 minutes is slow because time is expensive). Not to mention the fact
that you don't want to distribute half baked packages for the sake of yourself
and your users since you still need to test before distribution.
Even for distribution alone (no testing), uploading 100 source packages needs
way too much of mantime. Once you do this a couple of times, it is no fun at
all.
P.S. It may sound that monolithic KDE packages were better. Well, no. They
were awful with respect to actual packaging tasks (where is all the fun)
however "development infrastructure" was more manageable with fewer source
packages (even if their size was a huge disadvantage and PITA).