Hello,
On Aug 23, 2008, at 12:20 AM, Jean Vézina wrote:
>
> I have noticed the following remarks from Richard Maine about the
> reasons of the current lack of full Fortran 2003 compilers:
> "1. The f2003 standard is a failure. In that case the committee
> should be studying and addressing the reasons for the failure
> instead of just pressing on.
>
> 2. It is just too early to expect full implementations of something
> as large as f2003. In that case, it is also too early to be
> proposing a follow-on standard."
>
> Actually, compilers vendors should answer directly why there are
> still no full implementations of Fortran 2003 processors.
>
> Their representatives in the J3 committee are certainly aware of the
> reasons why it takes so much time to have fully compliant
> compilers. There is no need for guessing here. Complexity of the
> language is a valid reason, but I am afraid that the true reason is
> that Fortran is no longer a high priority item for them and they
> allocate some of their resources elsewhere.
>
In many cases, Fortran compilers are supported by the marketing
department,
not the software development department, of the firm. That is,
Fortran exists because it sells hardware, not because it sells large
numbers
of compilers. This is not altogether bad, because implication is that
requirements in RFPs greatly influence development schedules.
A number of vendors are very close to full f03 compilers;
the difficulty appears to be one feature, parameterized derived types.
The only major feature of f08 is coarrays. There is precious little
in f08 that depends *in any way* upon *anything* in f03. Thus,
the "need experience with f03 before judging f08" argument
is without basis in fact.
While I respect Richard greatly, it is hard to understand why
a lack of one feature from f03 compilers implies the desirability
of delaying standardized parallelism in Fortran.
Indeed, a number of compiler developers are implementing coarrays
before some, or many, f03 features. That, in and of itself,
tells me that f08 should proceed full speed ahead. It can only be
good for Fortran that customer's needs are addressed with higher
priority
than strict adherence to the artificial ordering of revision numbers.
--
Cheers!
Dan Nagle
|