(j3.2006) (SC22WG5.4550) [ukfortran] Comments on the technical content of the coarray TS

[John Reid]

David Muxworthy wrote:
> On 28 Sep 2011, at 11:37, John Reid wrote:
>
>> Would anyone else like to comment? I will be talking about this at the
>> BCS Fortran Group AGM tomorrow and hope to solicit some comments to
>> add.
>
>
> and on 28 Sep 2011, at 12:10, N.M. Maclaren wrote:
>
>> I have rechecked the WG5 Garching minutes and, while we did not
>> formally agree to Reinhold's point (2), I recall there being a
>> consensus that it was a necessary step within J3.

I think you mean point 2(a). The main thrust (2) is a request for a 30% 
increase in complexity, which is contrary to WG5's agreed aim of 
avoiding such an increase.

> As author of the minutes and resolutions, I have to admit that they
> are pretty minimal. However my recollection of the discussion at the
> WG5 meeting was that N1858 was to be shredded, annihilated, trashed,
> 100% forgotten about and that we would start again from scratch by
> first asking anew for user requirements, then producing an Objectives
> and Rationale document and only thirdly specifying syntax and
> semantics. I was surprised then at the BCS Fortran Group AGM when the
> discussion seemed to imply that the coarray TS would essentially be
> N1858 with variations.
>
> Perhaps others at the meeting could confirm or contradict my
> impression.

David, you were in the room. Why did you not speak up? I should have 
emphasized that J3 has been asked to produce an Objectives and Rationale 
document. But we are still at the stage of seeking user requirements. I 
have not seen the roughest first draft of an Objectives and Rationale 
document. I only had half an hour and I was trying to get some views on 
requirements.

I explained the history, which is that four features were removed in 
2008 with an informal promise that they would be put into a TR. These 
must be candidates. I was not positive about keeping any of them as they 
now are. I mentioned the new set of collectives from Bill, which seem to 
have consensus. I mentioned doubts over teams. Whether or not teams are 
included will be a big issue. Is there a requirement for parallel I/O? 
Is there a requirement for notify/query? I was trying to get opinions on 
these two. I asked if there is anything else for which people saw a need.

Cheers,

John.