[Mageia-dev] Mirror layout, round two

Tue Nov 30 09:15:54 CET 2010

I think this whole question is not done with an easy answer. It can
+also not be ssen in a black/white mode. I see the clear insight of
+Michael's suggestion which is a black/white point of view. Not
+maintained? Kick it out (well, not "out" but into the ante-room). But
+I also see the reality from the user's POV. As for technical skilled
+or experienced users (including server admins) the main question is
+that those packages which are available should work and be maintained.
+Period.
+But there is also the vast group of the "unwashed masses" including
+those we want to attract to Mageia. Many of those do look at the sheer
+number of packages (like, "I'd rather switch to Foo Linux which offers
+2 million packages while Mageia only offers 5,000"). Yes, I know, it's
+rather dumb and those users are the first to complain about some
+missing icon. But they are a large part of the users out there.
+
+So we have to find a middle way between the pure and the ugly. How to
+find that, I don't know, this is far beyond my knowledge. I only
+wanted to comment on the "philosophical" side of the problem. For me
+as a mostly non-technical guy the best solution would be the "flag"
+solution. Forget the main/contrib split and just flag unmaintained
+rpms so that the user sees it in the GUI. How to accomplish that on
+the CLI with urpmi I don't know. Then people who are security- aware
+like server admins can easily avoid unmaintained packages or open a
+request in Bugzilla which **may** inspire somebody to pick up the poor
+unmaintained package.
+
+One comment on the mirror maintainer part of the story:
+I was mentioned by Michael several times as an example of a certain
+kind of mirror maintainers. Yes, ressources are tight but not that
+tight. As I understood the "official mirror" as suggested by Olivier
+was about to fill up to 700 GB during the next 3 years - given that we
+will have 2 releases per year. Most of the official mirrors of
+Mandriva do not provide 6 releases, moreso when the life cycle of a
+release is less than 2 years.
+So, a realistic size woul be more like 450-500 GB at the most which is
+easily done with today's hardware. This is not a problem. Time is not
+a problem either for such people like me. The only problem I still see
+from the mirror maintainer's side is the way to deal with "tainted"
+packages wrt the mirrorlist (as already mentioned).
+
+-- 
+wobo
+