RE: REST, Conversations and Reliability

Ok, no problem -- thanks!

First, let me say that of course whenever we use generalities and platitudes we are susceptible to second guessing, correction, and finding such things as the exception that proves the rule.

Anyway, now that I understand your question, and if I understood Chris's recent responses correctly, the architecture would seem to allow either.  

This actually gets to the question of "what is a service" which I agree is a central part of the discussion.

It's very tempting, for example, for those wishing to integrate their applications to expose the most generic interfaces possible to minimize the risk of change.  To me it's therefore almost more of a design issue than an architectural issue -- one can use an interface mechanism in either way.

There's no easy answer to this -- it may make sense to establish a common contract at the lowest level of abstraction and specialize it for additional functionality -- this is I believe the SOAP model, in which the basic transport is augmented by headers for specific features.

In any case what I'd *really* like to do is send another email stating my view on the whole topic.  

I have the feeling reading Chris's latest comments that we are starting to get somewhere, and I hope it's not just me -- meaning I hope we are starting to converge.

I really like REST, I think Roy's thesis is a tremendous work, and am an admirer of HTTP with regard to the hypertext Web.  I agree with Jacek Kopecky, and I believe Larry Masinter also raised this point about two years ago, just after the XML Protocols Workshop in Amsterdam, that a native mapping of SOAP onto TCP makes the most sense for a new application type such as Web services represent.

However I also think we may be stuck with reality in two ways -- HTTP is established and so is SOAP/WSDL.

Anyway, let me try to summarize my views for a new thread and let's see if we can move toward convergence.

Thanks,

Eric

-----Original Message-----
From: Mark Baker [mailto:mbaker@myblackberry.net]
Sent: Monday, July 22, 2002 7:21 AM
To: Newcomer, Eric; www-ws-arch@w3.org
Subject: Re: REST, Conversations and Reliability


Sorry Eric, I was stuck in harvesting mode 8-).

By "connector", I'm referring to the architectural element whose role is to bind components so that communication can occur.  By "semantics", I just mean what is the contract in effect when they're bound: does it depend on the type of component, or is it the same for all components?

Or in other words, are the methods in the messages generic or specific?

MB
--
Mark Baker, Ottawa Canada. (613)261-5172
PLEASE RESPOND ONLY TO DISTOBJ@ACM.ORG

Received on Monday, 22 July 2002 09:00:22 UTC