On Thu, 26 Sep 2002, Michael Sperber [Mr.  Preprocessor] wrote:

> What precisely is the problem with doing the splitting
> post-arg-processing, i.e. doing the equivalent of
>
> --long-name="first-arg second-arg"
>
> from the shell?  That's how autoconf/configure operates, for instance.

Yes, in case anyone missed my list of questions/comments about multi
argument options in the "cmdline.ss library in PLT" thread, I suggested
this idiom there too.

At this point, I think the primary argument against the common idiom is
the possibility of implementing the cmdline.ss interface in terms of
args-fold, but it has been revealed that cmdline.ss diverges from
POSIX/GNU guidelines in other ways as well (creating further
incompatibilities), so I don't think it is realistic to pursue multi
argument options for the sake of cmdline.ss. Is there any other good
argument for considering multi arguments options in the face of the
sub-parsing idiom? I'm leaning against extending args-fold to support
them.

  -Anthony Carrico