Re: OT: good design vs evolveability (was TAG "Architecture of the W orld Wide Web" new draft published)

On Fri, Nov 08, 2002 at 02:01:24PM -0700, Champion, Mike wrote:
> See http://www.shirky.com/writings/evolve.html

Oh, don't get me started on that one.  8-)  I frequently agree with
Clay, but he fell for the "worse is better" trap there, big time.

Clay wrote;
> Centrally designed protocols start out strong and improve logarithmically.
> Evolvable protocols start out weak and improve exponentially.

Bunk.  Evolable *is* strong.  The Web is not poorly designed, it's the
best designed distributed system ever created.  It was also about as
"centrally designed" (or perhaps even more) as Gopher or WAIS, given
that the design team consisted of just one person early on, and then
two or three for the next couple of years, finally decentralizing
after it was a success sometime around 1995.

> I'm increasingly humble about how little we collectively understand about
> distributed objects, document messaging, service-oriented architectures, and
> how to leverage XML to make them work better.

It's good that you acknowledge that.  I wish more people would do so.
It's how I came to learn about the Web in 1998.

But there are a handful of distributed systems experts who grok those
things you mentioned *very* well, and they all work on the World Wide
Web project now AFAIK, in various capacities, some more direct than
others.  I don't think that's a coincidence.  I know of less than ten of
them.

>  I'm a lot more interested in
> making the web services architecture evolveable to it can improve as we
> experiment and learn than in making it "well designed" by today's criteria.

You know my answer to this one; we already have a both well-designed
*and* evolvable architecture capable of solving the problems Web
services are trying to solve.

I believe the Web will continue to evolve, but I believe it will do it
by adding *new* architectural constraints that induce new, valuable
properties.  I don't believe it will get us anywhere to drop any
existing constraints, because that would likely *remove* the desirable
properties that made it a success in the first place.

MB
-- 
Mark Baker.  Ottawa, Ontario, CANADA.   http://www.markbaker.ca

   Will distribute objects for food

Received on Friday, 8 November 2002 16:36:23 UTC