Re: A Conclusion Re: SchemeQL

[MJ Ray <markj@cloaked.freeserve.co.uk>]

Noel Welsh <noelwelsh@yahoo.com> writes:

> When code = data, SSI is very easy.  

It's more a case of using include virtual directives to put certain
objects (such as talk-backs or votes) on pages which are designed in
the wide range of other tools and platforms used by the section teams
to produce their sites.  After a number of attempts to change that and
failing to find a single development platform that they all are happy
with, I'm almost resigned to it.

> Since I don't know what you're doing I can't offer advice (and heck,
> I haven't done much web dev anyway!)

The site is tsw.org.uk but things are under heavy review.  Please
don't imagine that the half-broken shell currently there is in any way
representative of what it once was or what it will be.

> However the PLT team have indicated a desire to postion PLT Scheme
> as a web dev scripting language in the same domain as Perl & Python

Great!  This world needs good systems backing up the web and PLT
Scheme is definitely a good platform.  I think, long-term, running
something persistent to handle a subdirectory on the web site, or on
another port, is the way to go.  If that can be a scheme server sat
behind apache's forwarding and caching, then I'll be well happy.  I
know that the paper says that it's as fast as apache, but I think we
still need apache's feature set for such a large and organic site,
really.

> so I'd imagine they'd be quite supportive of your efforts.  

I'm likely to test their support to the limit.  And that of the very
helpful people on this list.

> More power to you if you want to start a project along the lines of
> OpenACS in Scheme.  I imagine there would be others interested in
> such a project.

Well, "web-site-engine" is registered on http://savannah.gnu.org/ at
the moment, but there's no code to speak of yet.  In fact, the only
piece of code which I'm really happy with yet is the following:

(define nntp:head ; server num -> list-strings
  (let ((cache '()))
    (lambda (serv num)
      (if (not (assoc num cache))
	  (set! cache (append cache 
			      (list (cons num (nntp:head-of-message serv num))))))
      (cdr (assoc num cache)))))

I'm sure someone will now put holes through that for me, but I don't
know many languages where I can achieve that in so little programming.
Of course, it needs restructuring to run inside a persistent process
without consuming memory endlessly.  Sucks to be me.

I must say that I'm amazed.  I think I've just seen my first post ever
where Shriram is posting something I vehemently and violently disagree
with.  Expecting templates to be written in scheme is totally
wrong-headed.  The designers just won't go for it.  Mostly you get
HTML from them, or XSLT if you're very lucky.  You will never get that
sort of mark-up from them, because they have no tools to develop it.
What is needed is some way to turn nice neat scheme trees into output
formatted according to their templates.  The world does not need
another mark-up language, even if it is a more beautiful one than that
used.

And now if *that* doesn't get shot down in flames, nothing will.