Lists Home |
Date Index |
On Mon, Apr 22, 2002 at 11:14:03PM -0700, Don Box wrote:
> FWIW, I believe that this last point is exactly what drives you, Mark,
> and Roy nuts about SOAP, in that every developer gets to define his own
> semantics, making it hard for today's HTTP infrastructure to make blind
> assumptions about what it can or cannot do with an arbitrary message.
Don, it surprises me that you can do such a fine job at distilling the
argument down to its essence, but at the same time cannot see the value
in a generic interface.
It is a Bad Thing for interoperability, security, and scalability if
every developer gets to define their own application semantics. This
isn't just a REST argument, it's inherrent in every application
that ever got deployed on the Internet. FTP was deployed because it
defined application semantics that enabled file transfer, and nothing
else. NNTP was deployed because it defined application semantics that
enabled news/article dissemination, and nothing else.
The only time RPC has ever seen success on the Internet, was when it was
deployed together with a set of application semantics. NFS and DNS are
the best examples of this. You don't see wide open, "invoke this
arbitrary method" type services for good reason.
It would be interesting to have the argument with you that HTTP's
application semantics aren't sufficient for what Web services aim to
achieve, but we have to first get past the issue that there cannot be
meaningful application layer interoperability without agreeing on what
the application is a priori.
Mark Baker, Chief Science Officer, Planetfred, Inc.
Ottawa, Ontario, CANADA. email@example.com