On Jun 14, 2005, at 5:04 PM, Jonathan Day wrote:
The mailing list has fallen virtually silent, there's
no beta yet out and (although I may have missed it)
there doesn't seem to be much in the way of updates on
the website.
We're a quiet bunch. :-)
Seriously, we didn't really anticipate much activity on this list until
the software becomes available. Then we plan for this list to be the
main "user questions" kind of list for Open MPI. A corresponding
"devel" list will also be made available for development questions
(i.e., tinkering around in the code of Open MPI itself).
The beta is not out yet, but barring any unforeseen problems, it should
be out "real soon now" (we're adhering to the same policy as LAM/MPI:
no absolute predictions on release dates, because they will inevitably
end up being wrong). It will still definitely be beta quality -- a bit
rough around the edges (e.g., we have taken just about zero effort so
far with making clear user error messages) and still in need of
additional work, but the intent is to get the code out in the real
world and a) find interesting ways for people to break it, and b) give
3rd parties direct insight into how our component system works, and c)
encourage others to join the process.
We actually branched in our subversion repository for the beta about a
month ago. So the beta is actually a bit behind the main trunk HEAD of
development, but it'll represent a relatively stable snapshot of the
code base.
As for the web site, we intend to eventually make it quite full of
information -- similar to (and someday replacing) the LAM/MPI web site.
It has been our experience that the more self-serve information that
is up on the web site, the quicker users are able to help themselves
when/if problems occur. But again, the first thing that has to happen
is a software release.
Can someone on the project just confirm that you are,
indeed, all alive and well, not kidnapped by aliens,
or been delayed for anything other than the usual
round of unpredictable last-minute code glitches?
I can neither confirm nor deny that we have (or have not) been
kidnapped by aliens.
Admittedly, this is taking us longer than we anticipated. :-( But we
are making steady progress in making what we believe to be a
world-class MPI. As you probably know, Open MPI is the result of
merging several different MPI projects, so we're taking the best ideas
from each and putting them in Open MPI. While not all the features
from each of the prior MPI implementations will be in Open MPI's first
release, they will eventually migrate into Open MPI over time.
Also, is anyone in the project doing a Linus Torvalds
and using the time to add bunches of new features?
(And if so, what are they? :)
We actually implemented a lot of MPI-3 in Open MPI... which mainly
consists of removing all the useless parts of MPI-2. ;-)
I'm kidding, of course. Open MPI's first releases will have all of
MPI-2 except the one-sided functionality (this was the plan from the
beginning, actually -- there was enough else to do and not much demand
for the one-sided stuff to justify putting it in the first versions).
What features were you thinking of?
--
{+} Jeff Squyres
{+} The Open MPI Project
{+} http://www.open-mpi.org/
