`scheme-script' and multiple Scheme installations David Rush (12 Mar 2001 08:27 UTC)
Re: `scheme-script' and multiple Scheme installations sperber@xxxxxx (20 Mar 2001 10:44 UTC)
Re: `scheme-script' and multiple Scheme installations David Rush (20 Mar 2001 11:36 UTC)
Re: `scheme-script' and multiple Scheme installations sperber@xxxxxx (20 Mar 2001 12:47 UTC)

Re: `scheme-script' and multiple Scheme installations David Rush 20 Mar 2001 11:36 UTC

xxxxxx@informatik.uni-tuebingen.de (Michael Sperber [Mr. Preprocessor]) writes:
> >>>>> "David" == David Rush <xxxxxx@bellsouth.net> writes:
> David> I just don't see how forcing them all to use a single name in 'exec'
> David> space will help anything.

And I still don't. Say I've already installed Scsh 0.6, which comes
with the brand new SRFI-22 support and poerted my GWZ application to
take advantage of the SRFI-22 compliant installation
capabilities. I've wrapped eveything up in cond-expands, and I'm
pretty confident that it's "portable".

Joe Bloggs now D/Ls GWZ and installs it on his SRFI-22 system, which
is Gambit. How likely is it to work? Well, if I have *really* done my
homework, maybe pretty good. A more likely scenario is that I haven't
actually run it under Gambit because there are too many
implementations to have covered all of them in my testing. Whoops!
there is an incompatibility.  So I have to go and put the Scheme
implementation specific stuff back into my configure.in, and I have
gained nothing out of the SRFI-22 exercise except the standard
parameter passing.

OK, the example is a bit contrived and arguably GWZ has packaging
bugs, but my point is that SRFI-22 didn't help me deal with them. I've
got no guarantees beyond what I can acquire from inspecting the system
through autoconf, and I'm still pretty much going to need to use
autoconf to do the inspection (if only to verify that I've got a
Scheme that I've actually tested).

> David> none of R5RS, SRFI-0, or SRFI-7 provides
> David> enough functionality to do significant scripting.
>
> I disagree with that from practical experience.

To be fair, you *can* write highly portable code with just SRFI-0, I
can't speak to SRFI-7, I don't use it and I haven't seen
widespread support for it (although I haven't looked hard since it is
even more of a meta-language than SRFI-0, and out-of-band to the
executable code, to boot).

> Moreover, SRFI 7
> gives you conditional access to the rest.  The "single name" (several
> in the next revision) is a central aspect of the SRFI, I'd say.

I just feel that single-name doesn't provide enough, and in fact loses
valuable information. Perhaps if 'scheme-script' maintained a registry
of compliant installations on the system along with introspection
facilities at the command line. e.g.

        scheme-script-srfi-0 --on Scsh --on Bigloo

or perhaps something autoconf can use like:

        $scheme-script --substrate
        Scsh 0.6

Otherwise, It doesn't seem to add much value. OTOH, Perhaps I'm just
falling in love with autoconf. Gripping hand is probably that  this is
a good opportunity to shake out non-SRFI compliant systems.

Just my $0.02

david rush
--
The Torah is written in black fire inscribed upon white fire - fire
mixed with fire, hewn out of fire and given from fire
	-- 3rd Century Palestinian Merkavah Haggadah