[EPEL] EPEL -- the way forward

[Stephen John Smoogen]

On 2/23/07, Mike McGrath <mmcgrath at redhat.com> wrote:
> Josh Boyer wrote:
> > On Fri, 2007-02-23 at 08:47 -0800, Karsten Wade wrote:
> >
> >> One idea we want to promote is making the packaging into a job role
> >> instead of a personal side-project.  This is possible and useful where
> >> your employer or other outside organization benefits from your package
> >> in EPEL.  For example, your infrastructure requires package Foo be
> >> rebuilt in EPEL.  If you make package maintenance part of your job role,
> >> you gain:
> >>
> >
> > Is your intention to preclude people that _are_ doing it as a side
> > project?
> >
> not preclude but join them together.  I suspect the climate for EPEL
> will be a lot different from Extras.  It'll be interesting to see if
> someone like IBM comes in and starts working with Joe Blow on package B
> because IBM needs to roll it out to 20,000 customers.
>

One of the things I have run into for needs for Extras for Enterprise
at various sites is that there are three different camps you need to
be able to satisfy.

Camp1 wants the same release for the lifetime of the product until it
can no longer be patched. So if clamav-0.88.1 was what was released
then they want patches backported until the end of the 7 years of
support for say RHEL-4. So when 4.5.1 comes out, they want only things
updated that have security updates and not API/ABI changes. Currently
they will take FCL-3 for say RHEL-4 and use whatever is in that repo
til time ends.

Camp2 wants general updates to match the quarterly release cycle. They
dont want to upgrade every 4 days to the latest, but they want
technology upgrades at regular times. So say clamav is the same for
RHEL-4.5.0 but want 4.6.0 to have whatever is considered stable at
that time. Currently they are taking a src.rpm from say FCL-5 and when
FCL-6 comes out upgrading to what was in there. They will upgrade
other stuff when it is needed.

Camp3 wants to get the latest stuff when it is available. They need it
for whatever project and are basically wanting a 'barely-qa'd
rawhide'. Currently they are taking Fedora rawhide and compiling it to
meet daily/weekly needs.

One thing we need to figure out what we can afford to do. I think
Camp3 is the easiest for volunteers to do.. and Camp2 has the largest
number of people. Camp1 should be left to people who are going to be
paid for it. It takes the most work and has the least 'reward'.


-- 
Stephen J Smoogen. -- CSIRT/Linux System Administrator
How far that little candle throws his beams! So shines a good deed
in a naughty world. = Shakespeare. "The Merchant of Venice"