----- Original Message -----
From: "John Cowan" <jcowan@[...].com> 
To: "Michael Champion" <Mike.Champion@[...].com> ; <xml-dev@[...].org>
Sent: Friday, August 04, 2000 11:11 AM
Subject: Re: W3C, responsibility (Re: Why the Infoset?)


> 
>  > Ruthless elimination of complicating
>  > constraints will be much more effective than attempts to eliminate
>  > self-serving behavior, even if such a thing were possible.
> 
>  Unfortunately, discarding Real World constraints produces the beautiful
>  product that nobody uses.  "Worse is better."

Therein lies the rub.  Which real world constraints are truly necessary, and
which are checklist items that everyone thinks they care about but don't
really need?  Obviously I'm not suggesting something like LISP, which was
elegant but simply didn't meet the requirements of the Real World.  I'm more
trying to warn the W3C away from inventing something like Ada, which was
well thought out but had far more than the Real World really needed.

I think the only historically supportable answer is to start from a simple
core, let various innovators add pieces more or less freely, and let natural
selection pick and choose what requirements are truly necessary and which
approaches truly meet them, sometimes adding, sometimes removing.  To use
examples in the area of procedural programming languages, that's how we
progressed from FORTRAN through Algol, Pascal, C, C++, Java, ....

As desireable as it might sound for some large organization or committee to
set standards before the practical knowledge of what is truly useful has
been painfully acquired, that path has led mostly to costly evolutionary
dead ends such as PL/I and Ada.  Or in networking protocols, natural
selection brought us IP, TCP, and ultimately the Internet; a committee
brought us the OSI networking model.  One can of course argue that
SGML/HyTime/DSSSL, LISP, Ada, and the OSI reference model are in some
abstract way superior to what evolution hath wrought ... but I think the
Real World has spoken quite clearly on the subject.

So I think the answer is to start simple (as XML more or less has), and add
additional constraints (or remove unncessary ones) one at a time, via
different activities in parallel rather than as one big blob of
interdependencies, and let the marketplace (of products and ideas) sort out
the winners and losers.  The W3C could clearly play a useful role in such a
world, but probably less as an engine of innovation (at which it is doing
progressively less well lately) and more as a neutral judge blessing
victorious innovations and deprecating the ones that don't truly prove their
worth.