Swietanowski Artur said:
> Please, note that V.S. allows "A's implementation to use B's
> specification". You made an example of "A's specification to
> use B's specification". So it's your example that's just wrong.
Thanks for pointing this out - obviously I should have had my second
coffee (and read the message more carefully!) beforehand.
Anyway, to get back on topic: Clearly it is not "impossible" to make
Fortran 90 libraries and sell them (we have existence proofs here!).
The question is whether the gain from implementation/specification
separation outweighs the pain.
Some of the painful aspects are:
- the next revision is still 5 years away
- the standards committees would need to be convinced
- it would almost certainly result in something else
being dropped from the F2002 standard
Note that the standards committees had been out consulting the users
and being lobbied for some time before settling on the content of
F2002. (E.g. We were meant to have decided the content of F2000 in
November 1995, this slipped to July 1996 and again to February 1997.)
Indeed, there are quite a number of individually desirable features
that were not included in the F2002 workplan.
To get this into F2002 would require convincing a lot of people that
it was of *overriding importance*. Important enough to seriously risk
not even making the current schedule which has the standard published
in November 2002! BTW, this schedule is ***NOT*** overly pessimistic.
Personally, I am yet to be convinced that the current module situation
is an insuperable obstacle to program or library design - though it
certainly means one has to think more about the design and not just
throw things together (some might even think this is an advantage!).
Cheers,
--
...........................Malcolm Cohen, NAG Ltd., Oxford, U.K.
([log in to unmask])
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
|