Lists Home |
Date Index |
Mike Champion wrote:
> Let me make sure I understand this "RPC accidentally balkanizes the Web"
On the money! Couldn't have said it better myself.
> c) Thus, the data is no longer portable, and the operations are no longer
> generic, so the Services Web is disjoint from the Web that we know today.
Yes, but Services Webs, plural. How do I point from a UDDI business
reference into .NET My Services or vice versa? I can only do that if
they are integrated point-to-point or through *the real Web.* And, in
fact, even today, the most useful interface to UDDI data is through *the
real Web*. It's just too bad that it isn't available in XML, in addition
> Finally, as I understand it, the practical implication of this is that
> 1 - SOAP as a packaging data into a convenient message format makes sense;
Yes, but you have to use SOAP in a manner that a lot of people will
consider "not really SOAP".
> 2 - An XML IDL perhaps similar to WSDL used as a metadata format for
> describing the packages makes sense
I have to think more about this issue and do some experiments. Obviously
just using XML schemas gets you part of the way. And you want to be able
to strongly-type intra-document links...so you need something more,
maybe Schematron. And then you have to think about describing the
pattern of transmission of documents back and forth so you need a little
> 3 - UDDI is a re-invention of what the Web does better already
I think the discovery conversation is kind of orthogonal. I used UDDI as
an example in my article just because it is so blatantly a simple
document publishing exercise which has been complicated for no reason.
But even if I was the world's biggest fan of RPC I probably still
wouldn't see how UDDI added much. Service discovery is just something
that human beings do, IMHO. Even when it may make sense to automate
service discovery (what printers are available?) it is so trivial to do
it with application-specific vocabularies that I don't see the benefit
of UDDI's generalization. CORBA people should have a lot of experience
with service discovery. Anyone want to pipe up?
> 4- The Way Ahead lies in teaching developers the power of HTTP and XML,
> not hiding them behind Wizards, and RPC protocols.
Agreed. But I think that the real power is in the combination of URIs
and XML. And the major virtue of URIs is their universality. HTTP is
only interesting because it is a resource (the "R" in URI) manipulation
protocol (and thus far, the only one). The REST debates tend to revolve
around HTTP because it is the controversial part. URIs and XML are
motherhood and apple pie.
Maybe if we added a few layers of paint to HTTP and called it URMP it
would be more trendy!