Re: INT: Re: Intermediary Discussion

On Thu, Feb 08, 2001 at 09:48:07AM -0500, Noah_Mendelsohn@lotus.com wrote:
> Mark Nottingham writes;
> 
> >> Are you suggesting that there should be a divorce between the nature
> >> of the transport binding (in HTTP's case, request/response) and the
> >> XP message exchange pattern?
> 
> No, there should be synergy, BUT I am suggesting that the XP level 
> abstractions (exchange patterns, intermediaries, etc.) should form a 
> coherent model, and will the level at which most applications are framed. 
> There is synergy, for example, between SMTP and the underlying binding to 
> TCP, but most email-enabled applications are coded to SMTP or some 
> abstraction of it.   If it flows well on TCP, so much the better.
> 
> I expect most apps (or the libraries they call) will implement XP at the 
> level of "Create an envelope for request message, add add body, add header 
> to mark transacted, add digital signature header, etc."  With proper 
> bindings and implementations of those bindings, this may cause all kinds 
> of nicely optimized http or even https magic to happen...the same 
> connection will be used for request and response, etc.  Just as in the 
> email case, the application is working at the higher level.  Therefore, 
> the higher level must stand on its own as a coherent model for use by 
> applications.  Does that make sense?

I think so. This is much more ambitious than what I see in the
charter, though. I almost want to say that it feels like an API,
rather than protocol, issue.

Some applications may need this invisibility, but many will not. I
dispair of trying to define a taxonomy of all protocol and message
path characteristics that XMLP can use; this is a very serious
undertaking, and I don't see how leaving it out of scope prevents it
being done separately (perhaps in relation to Web Services). 


-- 
Mark Nottingham, Research Scientist
Akamai Technologies (San Mateo, CA)

Received on Monday, 12 February 2001 17:52:51 UTC