(j3.2006) Materials for 199

[Van Snyder]

On Mon, 2012-09-24 at 16:14 +0900, Malcolm Cohen wrote:
> >This seems superior to the cumbersome MACRO facility.
> 
> It also does not do anywhere near as much,

It was intended only to address the problem of providing a portable and
automatic way to specify a subprogram having kind type parameter values
for its arguments that correspond to those of objects of the type to
which it is bound.  It has additional funtionality that is easier to
allow than to prohibit.  I.e., the resulting subprogram is not required
ultimately to be bound to a type, and the syntax to declare one can be
used to specify an explicit interface without instantiating a
subprogram. 

> and would appear to take more 
> resources to implement.

Depends upon how one's processor is organized.  Probably fairly close in
any case.

> >  More along the
> >lines of templates, which are requested with some regularity.
> 
> It also does not do anywhere near as much as templates.

Minimalism was intentional.  A compatible future extension would be to
allow types, constant expressions of any type, generic identifiers, ...
as parameters.  The proposal is compatible with macros, in that a
parameterized procedure could be generated by a macro.  But, if we do
macros, why do parameterized procedures?

The goal of the proposal is to get the "generic programming" work item
from the 2008 work plan back onto the agenda.  Hopefully, we won't allow
the perfect to be the mortal enemy of the good enough this time around.

> If we don't start talking about the problems until Delft we won't be in a 
> position to do anything about them.

Absolutely.  J3, J3-USTAG, and every interested national body, ought to
begin discussing the work plan for the next revision.  The more the
proposals are discussed by e-mail, the less time will be wasted on
nonstarters at Delft.

Apparently, we might as well discuss the work plan at 199, since there
are still no papers in the repository.

> Cheers,