[Date Prev][Date Next]   [Thread Prev][Thread Next]   [Thread Index] [Date Index] [Author Index]

Re: repo layout for EPEL (Re: Initial Proposal for doing Enterprise Extras)

Thorsten Leemhuis wrote:

Thorsten Leemhuis schrieb:
But as I said: I thinks it's to complicated. So I put up the following up for discussion - it's just and idea, I'm not really sure
 it it doable (or worth realizing):

Two devel repos for now:

I've thought about this some more; I think it's to complicated as well.
So how about this layout:


that looks good to me, with one change. Move testing to under the {release}/ so it becomes el4/{testing,stable}/{arch}/ - makes life a lot easier for Rsync targets.

testing/el{45}/ (and later 6789...) -- All new packages goto here
normally. Testing will use a rolling release scheme. Repo depends on the
corresponding el{45} branch. Repo gets into a feature freeze round about
two (four?) weeks before each {RHEL|CentOS} dot release update (4.4 ->
4.5 or 5.0 -> 5.1) and packages get copied over to the stable repo when
that dot release gets published.

I have a problem with this wallclock based rollover from testing -> stable. I'd be much happier if all non-security-related updates, sat in testing for 100 downloads and/or 10 days. before moving over.

w.r.t security-related-updates, we'll - lets talk about those :), eg. it should need atleast 1 person *other* than the packager to approve of it ( at the very least, in the absence of any QA process, as Spot has already pointed out )

el{45}/ (and later 6789...) -- stable branch; contains the packages for
{RHEL|CentOS}{45}. Packages are locked down there and only get updated
for security reasons after they were tested in testing for two (?)

I really dont follow your idea of 'locked down' - if it means a version freeze, much like upstream EL base, I have doubts if thats going to ever work - AFAIK, FExtras does not have the knowledge base amongst packagers to get into issues like backporting fix's and bugs. I might be wrong, but I would be surprised if I was wrong.

- KB
Karanbir Singh : http://www.karan.org/ : 2522219 icq

[Date Prev][Date Next]   [Thread Prev][Thread Next]   [Thread Index] [Date Index] [Author Index]