Re: Normative constraints on the WSA

On Sat, May 17, 2003 at 04:15:19PM -0400, Geoff Arnold wrote:
> On Saturday, May 17, 2003, at 09:43  AM, Mark Baker wrote:
> > Not at all.  But you don't see improvement by relaxing constraints and
> > removing the very properties that got us to where we are today.  You
> > see improvement by *adding* new constraints.
> 
> (1) This sounds very principled, but I can't think of a single example
>      of this pattern in other successful standards activities. Could
>      you, for example, describe how it applies to, say, the process
>      of 30 years of evolution in the TCP/IP community?

Consider the slow start algorithm.  It extended a far less constraining
one on retransmission timeouts from the initial RFCs, and realized
improved properties for the system.  Ditto for congestion avoidance.
Ditto for the recent PILC work on tweaking that removes the assumption
that dropped-packet = congestion.

I can't speak to the entire history of TCP/IP, but I expect that most
of it unfolded in a similar manner.  As Walden suggested though, there's
no reason why existing constraints need be axiomatic, so long as their
removal is done with great care.  I'm sure there are examples of this in
TCP/IP given its long history, but I can't think of any cases off the
top of my head (plus my knowledge of it is limited to the last 15 years
or so).

> (2) How would you characterize the addition of support for
>      non-HHTP messaging to SOAP and WSDL?

SOAP is a protocol which extends other protocols, so it makes sense that
it isn't tied to HTTP.  It's more like MIME than IIOP.

> >  I welcome all innovation
> > on the Web that does just that (see KnowNow), and I reject all
> > "innovation" to the contrary; it isn't innovation, it's taking us back
> > between 20 and 30 years in the evolution of large scale distributed
> > systems.
> >
> 
> Specifics, please?

Can you name a system on the Internet which doesn't have any interface
constraints?  Moreover, can you name a system which isn't late bound?
We've been building these things for a long long time.  We know how and
why they work.  There's simply no excuse for ignoring all that and going
back to the bad old days of unconstrained interfaces.

> I happen to think (personally, not a Sun position) that the present web
> architecture made a colossal mistake in not explicitly modelling the
> temporal nature of distributed systems state. Something like Jini
> leases would go a long way to solving many of the synchronization and
> coordination issues that we've wrestled with in various trout-ponds.

Well, I'd say that it just hasn't gotten around to it yet.  You don't
really need it for "Web browsing".  But as the Web is used for more and
more machine-to-machine stuff, we may very well see extensions of this
kind, and I would welcome them; not because I'd know they would work,
as I have reservations there (for the same reasons I have issues with
ACIDity and see hope in forward-progress style solutions such as
compensating transactions), but because I don't know for sure what a
workable solution looks like and I'd like to see a thousand flowers
bloom.

Again, I suggest followups to www-ws@w3.org.  I'd respond there, but I
want folks to know we've done that.

MB
-- 
Mark Baker.   Ottawa, Ontario, CANADA.        http://www.markbaker.ca
Web architecture consulting, technical reports, evaluation & analysis
  Actively seeking contract work or employment

Received on Sunday, 18 May 2003 07:30:49 UTC