Web Services: Its So Crazy, It Just Might Not Work
Mon, 6 Aug 2001 18:36:20 -0400
> I'm not against adding new semantics to HTTP, but I am against
> adding new application semantics that don't operate on resources.
> WebDAV, for example, adds a couple useful ones (and some, IMHO,
> stinkers). PEP had some good ones that RFC 2774 absorbed (since
> absorbed into SOAP).
> The key behind these new semantics, is that they be defined as
> operations on resources - existing resources even - so that
> invoking "LOCK" on http://www.markbaker.ca means something,
> whereas invoking "getMyGoat" on it doesn't.
Thought about this for a while, decided it made no sense, then realized that it
makes plenty of sense if you are thinking of HTTP as a shell around tuplespaces.
Your argument, I am guessing, is that binding via RPC leads to versioning
problems, while data-centric binding via tuplespaces doesn't. The reason that
tuplespaces don't lead to versioning problems, in this line of thought, is that
the only instructions that don't run locally are about exchanging passive data.
So LOCK(remoteObject) is legal, but SUM(passedObject) isn't.
Given that SUM(remoteObject) operates on resources but isn't semantically about
passing data, is it a good or evil operation?
Am I anywhere near correct with all this?