(j3.2006) (SC22WG5.3786) [ukfortran] Ballot on the technical content of the TR

N.M. Maclaren nmm1
Tue Dec 9 06:40:30 EST 2008


On Dec 9 2008, Aleksandar Donev wrote:

>Robert Corbett wrote:
>
>> I understand that.  I also understand the advantage the proposal will
>> give to vendors who already use the knew capabilities.  I don't like
>> Sun being put at a disadvantage relative to those vendors.
>
>Sun (and NAG, and whoever else does not store rank/type info in their 
>descriptors), were not singled out to be "disadvantaged" intentionally 
>nor even by accident (there were lengthy discussions over the rank/type 
>issue). Neither was IBM for passing OPTIONAL arguments differently from 
>others. Is it a better alternative to put everyone at a disadvantage, 
>i.e., force everyone to implement something utterly new, unused by 
>anyone at present, thing (as an example, read Nick's proposal), that is 
>actually less handy for users???

This is ridiculous.  I am still trying to make time to respond to your
last messages to me, but may not manage to.  However, I can't let you get
away with such unjustified assertions about N1761.  Some of these points are
responding to assertions in previous postings.

N1761 is simple to use only if you want to do exactly what it assumes. You 
have provided no evidence that its assumptions of what users want are 
correct, let alone representative. They certainly aren't correct for any of 
the dozens of users (mostly MPI users, too) I have talked to over the past 
decade or so.

N1761 matches existing practice only if your implementation matches N1761, 
and we have had at least two vendors say that it doesn't match theirs. You 
have provided no evidence that it matches the majority of implementations.

The vast bulk of N1761 doesn't help even existing MPI, let alone help MPI to
provide a real Fortran 2003 interface.  It is pretty clear that MPI will
continue to say (more-or-less explicitly, as has been done on its mailing
list) "Don't use assumed-shape arrays, because they aren't portable - stick
to explicit-size ones."  I.e. don't use Fortran 90 - stick to Fortran 77!

N1761 is a well-defined specification only if you close your eyes to the 
problem areas which occur with the interaction of TYPE(*) and other 
facilities (old and new). I will try to post an example of that if I do 
nothing else. As I said right at the beginning, it is virtually impossible 
to comment on N1761's details, because it is so unclear precisely what it 
is proposing.


Regards,
Nick Maclaren,
University of Cambridge Computing Service,
New Museums Site, Pembroke Street, Cambridge CB2 3QH, England.
Email:  nmm1 at cam.ac.uk
Tel.:  +44 1223 334761    Fax:  +44 1223 334679






More information about the J3 mailing list