On Thu, 19 Jul 2007, Brett Viren wrote:

> Hi Michael,
>
> Michael Hannon <[log in to unmask]> writes:
>
>> I'd like to know how others are dealing with this.  Is anybody using
>> Ubuntu clients with SL servers for instance?  Any other words of
>> wisdom on this topic?
>
> In our group we run Debian on our cluster, workstations and some
> laptops.  We are collaborators on two experiments that have picked
> some flavor (MINOS with SLF and Daya Bay with SLC) of SL as the
> dominant platform.  For the most part there are no problems in terms
> of client/server communication.
>
> The two real issues I have experienced are:
>
> 1) Fermi Kerberos (not really an SL issue, per se).  Debian's
> openssh-client package does't have kerberos support and if you use the
> kerberized ones and place your workstation inside FNAL.GOV realm you
> will suffer long timeouts when connecting to systems not known to the
> realm servers.  For this reason I run the standard openssh client and
> keep an "ssh-krb" client for when logging in to FNAL.

Since Ubuntu is based on Debian I assume the openssh in Ubuntu 
is not kerberized?

-connie sieh
>
> 2) Reliance on binary libraries produced in a monolithic culture.  Be
> aware Debian is marginalized or completely ignored by the big high
> energy / nulcear physics software developers (LHC and RHIC groups to
> name some).  Binary libraries built on SL, particularly C++ based
> ones, will very likely not work on Debian, even when used with a
> matching GCC.  As physicists are want to do, these package are
> sometimes built with baroque (broke) build systems and may be very
> difficult to build by non experts.  If you need such software you
> should take this into consideration.
>
>
> But, imo, these difficulties are insignificant when compared to the
> benefits of using Debian relative to SL.
>
> -Brett.
>