Hi Jarek, Troy, Connie, Marc, John, Stephen, Axel, Knut, Jan, Urs, Oleg, 
and all those I've missed or who are just watching from the sideline,

On Thu, 1 Mar 2007, Troy Dawson wrote:

> Hello,
> OK, I've had a night to think about it, here is my opinion.
> 
> If I have my way, Scientific Linux 5 continues on as has been planned. We

[Stephan breathing a sigh of relief]

Great!


Some - very - random thoughts on this issue:


We could switch the 400-odd SL systems we're maintaining here to using 
CentOS any day, with very little effort. We could have done it any day, 
for years. Yet, we haven't. There simply was and is no reason to do it.

SL has worked very well. And not just from a purely technical point of 
view.

CentOS has been our "plan B", and a very nice one, since we switched from 
SuSE to SL. Now if SL and CentOS were to merge, what would be the new 
"plan B"?

There's no need for half a dozen TUVEL clones on this planet. But having 
two is fine, especially if they have sligtly different objectives.

I don't share Jarek's opinion on the SL community being weak, at all.

The SL mailing lists are probably those with the best signal-to-noise 
ratio I've ever subscribed to. Not just because they're completely void of 
spam (and could someone please convey my admiration for this to whoever 
deserves it), but because they're populated by experienced admins of 
large linux installations. It's not surprising that most of them mind 
there own business most of the time. But if you have a real problem 
getting serious hardware to work, sl-users IMHO is one the best lists 
on the planet to post your problems on.

Jarek's idea isn't bad at all. CentOS seems a solid, well managed project. 
It's likely to be around for a while. It has a broader user base than SL. 
They've often beaten SL in providing rebuilds of TUVEL errata as timely as 
possible. "Going CentOS" should be something any SL user/developer should 
be considering.

I've never met any of the CentOS developers in person. Unlike *all* 
the main developers of SL. Makes a difference to me.

Someone said that this is the right time for this discussion. I disagree.
The right time for this discussion would have been months ago. That was 
when TUV released EL5beta1, and we discussed about dealing with kernel 
modules and what the repository structure should be, in light of the 
changes TUV is making. Joining forces with CentOS, IMHO, is an SL6 topic.

We started gearing up for SL5 here 7 months (!) ago, by looking into 
FC6T2. While we'll have to provide SL4 for folks working on the LHC 
experiments, probably for years, and we have to provide SL3 for WLCG, and 
it's unknown for how long, we have also have (significant numbers of) 
other applications and user communities that are not locked into an OS 
that's two years old already and would rather work with something 
contemporary.

Having the "very alpha" SL5 was really helpful. It made us confident that 
there are no serious obstacles in creating SL5. We're looking forward to 
the next test (SL5 alpha?) release. And we have a strong interest in 
having an initial SL5 release asap. Unfortunately, I'm not in the position 
to sign up for long term involvement in CentOS QA, which means I don't get 
to see the CentOS5 beta yet, and have no idea when it may happen and what 
it will look like eventually. Please, this is not about, CentOS-bashing. 
They certainly have their reasons. But it shouldn't be hard to understand 
why I do *not* want to see SL vanish.

"Sitting on a release" for years is not a requirement here. Yet, being 
able to apply the minor updates - which have become quite intrusive 
lately - at our own pace (say, after a couple of months of evaluation 
and testing) has been a huge plus. We seem to be in good company here: 
after all, TUV seems to have identified this feature as something that is 
worthwile to provide - and charge a premium for.

SL lifetime of course is an issue. Like others, we'll most likely not be 
able to get rid of all SL3 systems before autumn here. But then I think 
the end of SL3 support is not set in stone, one way or the other. I 
remember the discussion during HEPiX at SLAC: The 4 years we have now
for SL3 is a compromise, to be discussed again when the end of the term is 
coming up. And those who still need it should probably join forces and 
work on the problem, instead of just complaining. Of course, the existence 
of CentOS errata may help...

-- 
Stephan Wiesand
  DESY - DV -
  Platanenallee 6
  15738 Zeuthen, Germany