Re: Contributed Modules

[Brooks Moses <bmoses at stanford dot edu>]

Jerry DeLisle wrote:
> I was not thinking of modules maintained by the fortran compiler team.  I was 
> thinking of a place for the user community to contribute back in to the effort.
>
> So the wiki would be the right place for that.  I could just create a new page 
> for contributed modules with a disclaimer of "unsupported" and maybe "public 
> domain" or something.  I am not a lawyer but it has to be a no strings attached 
> effort. (Maybe thats a given with a wiki ? ?)

There are, I think, two clear possible models for this, which I'll refer 
to as the "libgfortran" model and the "CTAN" model.

The "libgfortran" model is modules that are in some sense part of the 
compiler package.  They'd be supported and distributed by the compiler 
team alongside the compiler, quite possibly mentioned in the 
documentation, and licensed with gpl+exception.

The "CTAN" model is akin to the Comprehensive TeX Archive Network.  This 
is modules that are provided by the community, not supported by the 
gfortran compiler team or treated as part of gfortran, and licensed with 
whatever the user licenses them under although free-software licenses 
are probably strongly encouraged.

I think these two models serve different purposes, and that having a 
system that fits each of them would be valuable.  It would, I think, be 
nice to have an ISO VARYING STRING module under the "libgfortran" model, 
for instance, and possibly an implementation of some of the POSIX 
functions for environment variables and command-line arguments.  But I 
agree with Jerry that that's not a good model for the bulk of random 
user-contributed modules.

I've referred to this as the "CTAN" model for a purpose -- I think that 
the way that CTAN (and CPAN, the Perl network modeled after it) has done 
this sort of thing is a very effective way to go about distributing 
user-contributed code back to the wider world.  I also think that this 
should to some extent be beside the gfortran umbrella rather than under 
it; most of these Fortran modules will be useful to people who are using 
other compilers, and I think we want to avoid the impression that it's 
gfortran-only.

CTAN, incidentally, handles the licensing arrangement in a fairly simple 
way.  Each bit of code contributed is required to list its licensing 
terms, and then the various code is divided into "free" and "nonfree" 
top-level directories accordingly.

- Brooks