- From: Walden Mathews <waldenm@optonline.net>
- Date: Thu, 27 Feb 2003 12:18:26 -0500
- To: www-ws-arch@w3.org
> > I read "distributed extensibility" as something different > > from discovery. > > I think it means services can extend their function without breaking > > clients, and also without the designers of the services and > > the designers > > of the clients sitting down ahead of the changes and hammering out > > the new design. > > Hmm, sortof what people usually call "loose coupling?" That makes sense. > But it's different than prior/a priori understanding ... it's saying that > once you come to some out-of-band agreement, you shouldn't break it Sort of. But "loose coupling" is a tremendously broad pond, including stuff like asynchronous exchanges, etc., while "distributed extensibility" is some- what narrower, and it would be a shame to lose that. Over and out. WM > > I'm thinking of proposing that this all comes under a general "loose > coupling" statement to the effect that "cool web services don't force their > clients to know all that much about their details in order to use them." > One aspect of that is that "cool web services" should allow themselves to be > bootstrapped -- GETing a base URI provides a description (we get into the > TAG RDDL discussion here ...) of its semantics and invocation syntax. But I > think this is all about best practices, not deep architecture. And best > practices are hard to come by even on the hypertext Web; AFAIK even > Fielding's thesis is a theory of which practices will turn out to be best > on the hypertext Web rather than an empirical examination of what actually > works best. I could be wrong, and I don't want to open a trout pond here > ... >
Received on Thursday, 27 February 2003 12:18:40 UTC