Re: configuration improvements for the PLT webserver?

[MichaelL@frogware.com]

> I have not been paying close attention, but for both this and the log
> file it would be nice to have modules that are imported into the server.
>  This would allow a sort of mod-rewrite on steroids that was actualy
> comprehensible to humans.  The architecture of the URL rewriter seems
> somewhat clear, URL and client info (request header, authorization) in,
>   URL out.  I had played around with hacking web-dispatcher-body to do
> this, but it seems possible to extract out into a unit.

I second the motion. I've only read through the web server code, I haven't
used it, but I do remember in reading through it that it looked like there
was a great opportunity to use units to break out implementations so that
new implementations--logs, cgi, security, error handling, etc--could be
plugged in by linking together new compound units.

In a sense, I see a tremendous opportunity for the web server to be a real
"showcase" piece that represents what you can do with Scheme--and mzScheme
in particular. Along those lines you get into things like remote admin,
dynamically updating modules in a running server, etc. Lots of work, but
interesting from a design perspective and an mzScheme perspective.

I read in the mailing list archive the "Peasant Revolt" thread. To me, the
single best way to get people to talk about and use Scheme is to provide
great solutions in Scheme. If the solutions are great, the people will
follow. (Right now mzScheme is a pretty cool environment for creating
solutions, but there aren't a lot of "solutions"--like the web
server--built on top of it, other than the teaching stuff.)

And that's where the myriad Scheme implementations hurt the effort. There
_is_ a lot of interesting Scheme stuff out there, but it isn't very
portable. Well, damn, I guess that's a fact of life--but what a loss.