Hi,

I was listening to Roy Fielding’s “A little REST and Relaxation”
presentation as given at the Jazoon ’07 conference:

http://www.parleys.com/display/PARLEYS/A%20little%20REST%20and%20Relaxation
http://jazoon.com/en/conference/day2.html

At 21:50, he goes into a point that Mike Schinkel made here a
while ago:

> Important to REST was the notion of minimizing coupling between
> systems. There’s a lot of talk in object-oriented language
> research about the importance of minimizing coupling and it’s
> interesting for me sometimes because generally what some people
> do in the language research in terms of coupling is so much
> more extensive than the level of requirements that we had for
> minimizing coupling. We needed a system that could be developed
> independently by 500-1000 different companies, and each of the
> things that they added to the web could be extended and
> deployed independently without affecting anyone else on the
> web, again without actually knowing what those extensions will
> be. And the only way you can do that is completely eliminate
> coupling between clients and servers. The only coupling that
> exists in a REST-based architecture is that the first address
> that you access has… ah… basically a bookmark, and you need to
> keep track of that bookmark. So essentially the rationale
> behind the “Cool URIs Don’t Change” is essentially that’s the
> last remaining bit of coupling in the architectures that are
> based on REST.

So yeah. Decoupling via hypermedia-driven app state does not
extend backward past the beginning of a client-server interaction
in a way that rather reminds me of how causality within this
universe does not extend backward past the big bang pinhole.

Regards,
--
Aristotle Pagaltzis // <http://plasmasturm.org/>