[ofa-general] RE: [ewg] RFC: Do we wish to take MPI out of OFED?

[Jason Gunthorpe]

On Fri, Jun 05, 2009 at 09:44:09AM -0400, Jeff Squyres wrote:

> 3. As Doug described, packaging MPI and OFED together actually makes it 
> *harder* for distros.  Remember that RHEL and SUSE don't end up using any 
> of the OFED packaging; they essentially use the individual SRPMs.

I would almost say the entire OFED process is making it harder for the
distros, and I mean that in a very general sense.

A working IB stack has been in the kernel for a long time now. Can I
install Fedora Core, Ubuntu, Debian, etc and have all the necessary IB
tools and diagostics available? Nope.

IMHO, the entire process would be better served if OFA focused on
producing a bleeding edge environment within the community
distributions (Fedora, Debian, OpenSuse, etc) and testing that entire
stack, including MPIs included in the distro. If this had been done
from the start I would have a complete IB stack available in every
single distribution TODAY.

The truth is, within the Linux world, if you want bleeding edge stuff
you use something like FC or Debian Unstable. Otherwise you settle for
older, hopefully more tested stuff.

The OFED distribution ontop of a distribution is just weird and
painfull... For selectivly moving an old distro forward I would *much*
rather have true backport packages that exactly match in form and
function the native packages in my distro with new versions. That is a
far safer path if I actually care about not disrupting my core
distribution.

Jason