[r6rs-discuss] set-car!

From: Thomas Lord <lord>
Date: Mon, 25 Jun 2007 12:19:20 -0700

> As Anton van Straaten put it:
>
>> Irony is inherent in the design of programming languages. Heisenberg's
>> uncertainty principle applied to programming language design says that
>> the more carefully you specify a feature, the less it resembles what you
>> were originally trying to achieve.
>>


I think that there is something very non-ironic about the draft
document. Namely, if it is regarded as deeply approximate
and suggestive, rather than normative, it does a great job at
prodding quite a bit of "needed stuff" for Scheme.

In some sense, I wish we could skip over the next few months
of diddling around with formal processes and polemical exchanges
and just start to get a lot of this stuff (approximately) implemented
and fed into emerging package systems (snowfort, eggs, etc.)

With experience, all of the details that are contentious today
are likely to become easier to reason about. Meanwhile, even
without settling those issues, a lot of de facto portable code
could be written.

Portability among Scheme implementations doesn't have to be
perfect, anytime soon. One metric by which Scheme is already
ahead in a noteworthy way is in sheer numbers of currently
in-use real-world applications. Evidently, we are already
de facto portable in some minimally pragmatic way. Every
incremental improvement will help. The draft (regardless of
its "legal" status) helps. I think that snowfort, if people will
adopt it, can really accelerate an organic evolution of our understanding
of what kind of portability really matters.

In my own professional work, there are things that keep from
reaching for Scheme for many of the tasks that come up. Interestingly,
the draft document plus a package system give me hope that that
might change. For example, I sorely miss (and am somewhat
surprised to find missing) really efficient, beautifully coded,
conforming, feature rich Scheme libraries for a lot of W3C fundamentals.
Perl and Python libraries included multiple 65% solutions for things
like parsing and printing XML, evaluating XSLT or XQuery transforms,
etc. I'd have thought Schemers would have nailed that kind of thing
by now. Perhaps draft specs for Unicode support and a package framework
for sharing results will be enablers.


-t
Received on Mon Jun 25 2007 - 15:19:20 UTC