Hi Jeff, On Jan 22, 2010, at 12:50 PM, Jeff Squyres wrote:
We had a lively meeting in ATL for the backwards compatibility issues. I took a set of slides from Dave, beefed them up a bit with my own and a bunch of Quincey's thoughts, presented them, and took some of the feedback from the session and put it back in the slides before posting them here (I was rushed a bit when making the slides; they were a little unclear on some points):
http://meetings.mpi-forum.org/secretary/2010/01/slides.php
We had a good 1.5-2 hour discussion on the whole set of issues. The exact discussion isn't important; here's the key takeaways (also on the wiki: https://svn.mpi-forum.org/trac/mpi-forum-web/wiki/BackCompatMeetings):
Looks like a good summary, thanks for taking notes!
* Make new functions with a suffix (e.g., MPI_File_write_L -- exact suffix TBD) that take MPI_Count count parameters.
I think we (our working group) need to press for a "preferred vector" through the "name dimensions" as we make changes to the MPI API routine names, otherwise we are going to end up with a combinatorial explosion of names that should be implemented (or very confusing names). (Not as bad as the combinatorial explosion for the F90 wrappers, but along the same lines) For example, if we use "_L" as the suffix for the MPI_Count change, and then need another suffix (say "_A") for another change, are we going to have all four combinations of routine names? (MPI_Foo, MPI_Foo_L, MPI_Foo_A and MPI_Foo_L_A) What would we do when we made another change and added another suffix, have eight variants of the MPI_Foo routine? I think we need to pick a "preferred vector" through the name combinations for each version of the standard and use a feature-neutral suffix for the changed routines' names. Either a totally abstract sequential suffix like: (nothing)->2->3->4, or a standard version specific suffix: (nothing)->3->31->32->...->4->41->... What you you guys think? Quincey