William B. Clodius wrote:
>
> Jean:
>
> Be certain you understand what this implies. Genericity is a complicated
> feature, much larger than IF ... THEN .. ELSE ... ENDIF or POINTERs. If it
> is added to the language after the draft is officially submitted to the
> public the delay in the standard will be large. First the detailed form of
> genericity will have to be debated within the committees. Second, large
> sections not directly related to genericity will have to be rewritten.
> Third a new draft will have to be sent out to the public for comment. This
> implies at least a two year delay in publication of the standard. The
> committee would be better off simply noting that the size of the request
> makes it inappropriate for inclusion in the standard at this time and
> putting it on the list of priorities for the next draft of the standard.
> The committee has to respond to comments on the drfts, but being responsive
> does not mean acquiescence.
>
> If this capability is to be in the next standard it is better off being in
> the first draft. If it is in the first draft it will probably have to use
> my proposal as a starting point, otherwise there will be large delays in
> developing the feature. The one person who liked my proposal is a bright
> intelligent knowledgeable user, but like myself does not have detailed
> experience in language design, standardization, or implementation. At
> present I have no idea whether my proposal is a reasonable starting point.
>
> I would not mind others lobbying for inclusion of this capability in the
> first draft, but such lobbyists should at least read my proposal to
> understand what they are likely to get if their lobbying succeeds. It is
> available at X3J3's ftp site as 118-97.
This is true that the inclusion of this feature must not be taken
lightly since it's a major addition to the language. Difficulties
of implementation are also unavoidable. The advantages of providing
this feature must be weighted against the problems it causes for
compiler development.
But the real question is: How problematic is the lack of genericity
in Fortran ? Is this justifies a delay in getting the next standard in
order to get the feature present ?
If genericity is not adopted in 2002, then at least five years will be
needed for developing the following standard (2007). Is this too much ?
It is now up to the users to answer these questions and to do the
necessary lobbying if they feel that the addition of genericity is
critical.
However, as a stop gap measure, it is possible within the current
standard to use workarounds such as the one I have developed using
TRANSFER. For implementing generic linked lists, this method avoids
the need for writing a complete set of routines for each data type.
Thus, for this specific problem, the lack of direct genericity is not
critical.
Regards,
Jean Vezina
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
|