[Mpi3-bwcompat] Notes from ATL Forum meeting

[Quincey Koziol]

On Jan 26, 2010, at 9:38 AM, Jeff Squyres wrote:

> On Jan 26, 2010, at 8:39 AM, Quincey Koziol wrote:
> 
>>        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->...
> 
> I think the single digits would be fine.
> 
> We have a precedent of keeping compatibility between minor MPI revs and only breaking compatibility if something is really broken.  I.e., if we make MPI_Foo3 and then find out that it's broken, I think the gut reaction of the Forum would be to fix MPI_Foo3, not make MPI_Foo31.

        I agree, I was just trying to make the sequences look [more] different. :-)

        The main difference is that the abstract sequence will never skip a value, whereas the version specific sequence could skip values (if a symbol was changed in version 3 of the standard, was the same in version 4 and changed again in version 5, etc.)

                Quincey