> Early in MPI-2, Jim Cownie
> said (I'm paraphrasing) that MPI-Forum is not here to standardize early
> research. I couldn't agree more. The wide uneasiness reflects the
> newness and immaturity of the put/get programming methodology.
I don't think that the get/put programming methodology is new or immature. For
example, there are many succesful get/put aplications developed on the Cray T3D.
Another example are applications that use Global Arrays (GA), which implements
get/put model portably on most MPPs and networks of workstations. There are
nearly one million lines of code in the GA applications, with one of them
being larger that 250,000 lines. However, GA does not support any
of the features that make people "uncomfortable", including the items
on the Marc's list...
It is really strange how some of these items appeared in the 1-sided chapter
and became impossible to get rid of, even though nobody was able to name
any motivating applications that would need them.
Jarek