Hi All,

On Wed, 28 Feb 2007, Jaroslaw Polok wrote:
> Just investigating for the future: or shall we (we = all
> of us using SL(X) go Scientific Linux 5 .. or maybe re-base
> to CentOS 5 ?

since Alex asked for a suit's opinion, I'll throw in what came to my mind:

DESY is a lab supporting several distinct user communities from HEP/GRID 
and photon sciences running experiments and accelerators both on- and 
off-site. We are currently providing "plain" SL3 and SL4 in both 32 and 64 
bit flavours as well as their "value-added" localized counterparts SLD3 
and SLD4 to our users using GRID- as well a custom deployment tools.

SLD (compared to plain SL) adds our in-house configuration and lifecycle 
management (wboom/SALAD), a few hundred application software packages 
(e.g. gnu stuff, commercial compilers, MatLab, Oracle client, ...) and the 
integration into our local infrastructure (PAM, NIS, LDAP, and Kerberos 
configs for identity and access management, mail configs, printer setup 
for 450+ devices, monitoring, central logging, ...).

One of our concerns is the distribution support life. We need to maintain 
some of the systems over many years without upgrading to newer 
distributions, and we'd like to have security updates as long as we can 
get while these systems are alive. This basically rules out Fedora. As far 
as I can see, CentOS plans to deliver updates as long as RedHat does for 
3.0x and 4.x. IMHO, this would be an improvement compared to the current 
SLx roadmaps.

A second major concern is the availability of a binary compatible 
Enterprise distribution, in this case RHEL. This will allow us to mix in 
"fully supported" servers for commercial applications like Oracle or SAP 
while being compatible with the "free" CentOS or SL server farms and 
their configurations. This requirement basically rules out Debian, but a 
SLES rebuild would do as well.

Migrating to a different package format (RPM to .deb) wouldn't be too 
hard, though we wouldn't want to maintain both over a long period of 
time. We already have Debian deployment support in our framework, but no 
application software or user consulting.

Having said that, I (personally) see
- no strong arguments against re-basing to CentOS, but a few in favour
- no available short term alternative to a RHEL-based distro

Just my two cents.

Cheers,
Knut

-- 
   Knut Woller          -o)
   DESY -IT-      (o-   /\\
   Hamburg        //\  _\_V
   Germany        V_/_