Re: [xml-dev] MS thinks HTTP Needs Replacing???

Alaric Snell wrote:
> 
>...
> 
> For IIOP, it's documented in the specs... let me see...
> http://www.omg.org/cgi-bin/doc?formal/01-12-45
> 
> Basically, it's something like:
> 
> interface Object {
>         Object invoke (String methodName, Object[] arguments);
> 
>         ...and something that lists the implemented interfaces?
> }
> 
> It doesn't have a flag saying whether the result is cacheable, from memory,
> so there's no difference between GET and POST. Never mind.

When you say "Never mind" are you admitting that what you're presenting
above has none of the features of HTTP and thus is a very good
demonstration that HTTP and RPC are not the same thing? If the
intermediary in question doesn't know whether the goal is to GET
information or to PUT information then it hardly "understands the
message almost as well as the sender and the receiver" does it?

 ***

Anyhow, cacheability is the least interesting thing. Let me say that
several times:

*cacheability is the least interesting thing*
*cacheability is the least interesting thing*
*cacheability is the least interesting thing*
*cacheability is the least interesting thing*
*cacheability is the least interesting thing*

Information addressing is the most interesting thing. Information
addressing is only useful if you can get or put a representation for the
information. HTTP has a raft of features that allows this. CORBA has
standardized information addressing (object identity) but you can't get
the information *out* or put it *in* unless you know a method name. The
method names are not standardized, so you need out of band agreement. 
SOAP doesn't even have standardized information addressing.

HTTP is not "just" RPC precisely because it standardizes the method
names and semantics and thus removes the need for out of band agreement,
like XML is not "just" Unicode because it standardizes the placement of
angle brackets and their interpretation in terms of a tree structure. If
we have this debate again next week can we please start from the point
of view that cacheability is not the point?

 Paul Prescod