> I think Ousterhout's notion of a glue language is essentially bogus.
> N years from now, I think we will look back and say, "How's that
> again?" Here's what the Scheme and Scsh communities can really take
> out of that paper: by devising the right kinds of interfaces and
> libraries, Scheme can be an outstanding glue language _and_ a systems
> programming language.
I read something slightly different from this paper (which Ousterhout
doesn't really state). I believe the reason for the success of Perl,
and Tcl is their comprehensive nature. You can use them for
1) building GUI
2) database access
3) CGI
4) tons of other specialize things.
There are scheme implementations that have some of these elements,
but NONE that have them all. It's the synergy between all the
pieces that's important. The great thing about Perl is that you
can do just about anything with it. There are modules/extentions
to do just about any specialize task you can think of. And the
Perl community has taken the time to organize it all (CPAN).
Part of the problem is that scheme developers have been developing
their own scheme implementations rather than rallying around a
single distribution. This divides the energy of the group and
results in a dozen (slightly incompatible) scheme implementations.
So if you want scsh, you use scheme48. If you want Tk, you
use STk. If you want database access you use SIOD. Etc...
Hopefully when Guile arrives on the scene, many of these things
will be eventually available for it (primarily due to the higher
profile of the FSF and GNU project).
Also I think the scheme community has been tackling the wrong
problem. They want to replace C/C++/Java, when they should be
trying to replace Perl/Tcl/Python.
Richard Coleman
coleman@math.gatech.edu
|