Web Services: Its So Crazy, It Just Might Not Work

Clay Shirky clay@shirky.com
Tue, 31 Jul 2001 15:35:52 -0400 (EDT)


> Because if you know a priori whether or not there is a side effect,
> you can do lots of good stuff with that knowledge; [security, cache...]

Yep, but the point I want to insist on is that doing those things does
not violate the http spec anymore than when Web sites do it.  The use
of POST as a catch-all method, in other words, is not specific to
XML-RPC, but is a general (and, judging from your and Adam's reaction,
unobjectionable) aspect of http, and therefore, inasmuch as Dave & Co
wanted to build an RPC method using XML, they made a perfectly valid
choice.

The more interesting conversation, I think, is to take the overall
goal of Web Services, which I take to be roughly "Let's apply the
Web's lessons of loose coupling in document handling to inter-process
communication/remote method invocation", and talk about a) whether and
b) under what circumstances this is a good idea.

-clay