[MPI3 Fortran] MPI Fortran bindings

[Aleksandar Donev]

On Friday 05 June 2009 13:17, Jeff Squyres wrote:

> the functionality *does* exist, even if  
> it is not in the language standard. 
Except that, it is slightly different in every compiler, in ways that 
you won't get even if you read the documentation. If you pass an 
integer/real contiguous buffer, it will likely work as expected. 
Otherwise, all bets are off.

If you really want to specify that what you want is equivalent to what 
we have proposed to the Fortran committee (TYPE(*)), which requires 
more than one sentence to fully specify or explain (so simply 
saying "you know what I mean" will not do), then you can simply cut and 
paste our latest description of TYPE(*) (Bill can send that to this 
list) and put it as part of the MPI specification.

> While we (MPI) clearly don't 
> want   to use a pre-standard form that may or may not be accepted,
> *some* notation is needed to denote this (potentially
> vendor-specific) behavior in the bindings. 
Again, denoting a non-existing or unspecified thing does not make it 
clear.

You can give all the "succinct" sentences you want...just because they 
are short does not mean they are clear. You are simply glossing over 
the complexity!

> 2. A second issue is that there is some debate as to whether INTENT
> is   useful.  I think that most people are saying yes, but Aleks is
> saying it's not that important.
Don't listen to Aleks on that one and add INTENTs if it makes someone 
happier.

Best,
Aleks