Cluster Poll Results (tangent into OS choices)

John Hearns john.hearns at
Tue Nov 4 10:06:16 EST 2003

On Tue, 4 Nov 2003, Robert G. Brown wrote:

> As I've noted and will continue to note again, I think that we are on
> the verge of a paradigm shift in linux anyway, and that this is going to
> likely kick it over the edge.  The existence of yum AND apt-tools
> Instead of upgrading systems on a timescale with granularity of years,
> we may be entering a universe where systems are microincrementally
> updated nightly, with immediate feedback and repair, and with a
> user/admin determined lag relative to the primary repositories to insure
> a level of institutional stability deemed locally acceptable.  In a way


A couple of days ago you made a post re. updating mechanisms, where
you talked about yum, or yum-like mechanisms which would recompile 
packages to suit the target node, or something along those lines.

How about we think of applications along the same lines - but not
necessarily being recompiled.
In the context of a grid world, it still worries me that people are
saying that application Z will run anywhere - on some machine maintained
on another campus - who knows if it is even running the same Linux
distribution? (OK - maybe everything needs to be statically linked then.)

If I may, I'll join Bob in a blue sky thought. How about applications
being installed as RPMs? Then the RPM would have dependencies - 
application Z needs library-b between x and y. 
Could you then get a message back saying 'sorry - this cluster won't
support this application: it needs x-y'

Sorry - a huge bly sky here. But we do keep hearing about grid and
infrastructure on demand.

Beowulf mailing list, Beowulf at
To change your subscription (digest mode or unsubscribe) visit

More information about the Beowulf mailing list