[CentOS-devel] Octave for CentOS5
mail-lists at karan.org
Wed Mar 7 16:36:52 UTC 2007
Philip Ray Schaffner wrote:
> On Wed, 2007-03-07 at 00:37 +0000, Karanbir Singh wrote:
>> Hi Phil,
>> we are working on a process for contributing rpms into centos-extras..
>> I'd expect that process will pickup steam once c-5 is out of the door.
> So have you core guys thought about how EPEL will impact extra,
> centosplus, kbs-extras, ...? Seems like joining that effort might be
> more productive than separate thrusts to port/rebuild packages that are
> in Fedora Extras.
EPEL is interesting and it has a lot of potential, but the fact that
they only care about, build for and expect usage on RHEL tends to sort
of exclude a lot of external participation in the project. So much so
that a @redhat person said that the only aim they have in pushing epel
is to they can go tell their customers about it!!
Added to that the overhead of needing to sign papers with Redhat and the
need to become a Fedora contributor first, only further increases the
bar to entry and creates really un-necessary issues for people who are
unable to, for various reasons to sign such papers etc. And given the
fact that Fedora packaging dynamics are drastically different from
packaging on *EL, whether the general Fedora guidelines and
infrastructure is even usable long term for EPEL is itself doubt.
So in effect what epel creates today is yet-another-repo. Perhaps the
best repo of them all. But it is effectively, just another repo.
EPEL does not really solve the big problem, about being able to present
a single repository for the EL user base ( whatever variant they might
be on ). Not sure how many people are following the rpmfusion
discussions that some of the fedora-unity guys have going on at the
moment, but i think thats a brilliant effort, if we could stretch it to
*EL, that would be a result.
The real, short to medium term, wins for CentOS and other *EL distros -
is to have an infrastructure that allows packagers to maintain their
spec's in whatever system / buildprocess / version control system they
prefer and yet be able to expose the resulting binaries in a single
repository ( or, well, a single place for the repos - split by stability
and disto friendliness rather than role ). The mechanics and policy for
such an effort to happen are things that need working out, but based on
the conversations that took place at Fosdem this year - its achievable.
Finally, I just want to point out that this represents my personal
viewpoint, and is not a 'CentOS perspective' on the issue. How and what
CentOS as a project can do, should do and is able to do with EPEL is
something that still needs to be worked out. But for now, epel will be
just another repo, at par with anything else / everything else out there.
Which is why I think that the special interest groups that care about
specific vertical markets and deployment roles should be organised. And
one thing that I know definitely needs to happen, asap, is to expand on
the 'involved contributors' numbers. CentOS today has a few million
users out there, but the number of people actively involved is still
just a few dozen at best. That needs to change, and ideas on how such a
change might come about and what we are doing wrong that needs to be
fixed, are *very* welcome.
PS: I am not being negative about epel, just sharing what the picture
looks like from this side of the fence. I know almost all the guys who
are branching for epel at this time, and I think they are all great guys
with excellent packaging skills.
Karanbir Singh : http://www.karan.org/ : 2522219 at icq
More information about the CentOS-devel