W3C home > Mailing lists > Public > www-ws-arch@w3.org > October 2002

Re: Gateways

From: Mark Baker <distobj@acm.org>
Date: Wed, 9 Oct 2002 14:55:52 -0400
To: Jean-Jacques Moreau <moreau@crf.canon.fr>
Cc: www-ws-arch@w3.org
Message-ID: <20021009145552.A4663@www.markbaker.ca>

Hi Jean-Jacques,

Sorry about the hastiness. 8-)

On Wed, Oct 09, 2002 at 09:47:41AM +0200, Jean-Jacques Moreau wrote:
> Mark,
> 
> You may remember the figure at [1]. Although old and outdated, it 
> shows some of the thinking that was later incorporated into the 
> Transport Binding Framework.
> 
> The interesting case if Host III. It is a SOAP node that is a 
> SOAP intermediary. It is also a "SOAP gateway", although that 
> term does not formally exist.  The two rectangles for Host III in 
> the "Underlying Protocol Layer" mean that, AFAIR, the inbound and 
> outbound protocols are different.
>
> However, Host III may not be a "gateway" in the sense you mean. 

Right.  By my definition, a gateway has to terminate messages.

> The SOAP message is interpreted at the application layer before 
> being forwarded. I feel this is different for your typical 
> gateway: the gateway would not, in general, have to understand 
> the contents of the message --- just as a layer 2 router would 
> not have to understand the contents of the packet being forwarded.

Uh oh, it's that layering issue again.  8-O

If the underlying protocol is an application protocol, like HTTP, then
SOAP and HTTP are at the same layer.  So for SOAP bound to HTTP, a SOAP
node may be an HTTP gateway.  Hmm, not sure if that helped clarify
things or not.

In your other message you wrote;
> Is "interface" the widely accepted term here? I can't help but 
> think you're talking about network interfaces; hence I don't get 
> the notion we are converting from one protocol to another.

Well, an application protocol defines an interface, so I think it's
appropriate.

> Re. "gateways that are SOAP nodes are not SOAP intermediaries", I 
> am reading Henrik's answer[1] somewhat differently.
> 
> He says: "gateways are not SOAP intermediaries", meaning SOAP 
> intermediaries do processing at a higher level: they understand 
> and possibly process the message; gateways typically do not.

I'm quite sure that's not what he meant.  He means that gateways,
like HTTP gateways, are not SOAP intermediaries because they
terminate messages (which includes terminating SOAP messages,
because they're at the same layer).

> But he then adds: "One could imagine SOAP intermediaries being 
> underlying protocol gateways", meaning a SOAP intermediary might 
> not process the message at all but simply switch protocols.
> 
> I think this fits perfectly with the figure at [2] and my earlier 
> note [3].

Mostly agreed, but his use of the term "gateway" there is different than
mine.  I agree that both definitions of the term are in common use, but
I believe the definition I use to be more common, especially in the
context of HTTP which has a well-defined meaning for it, consistent with
mine;

http://www.zvon.org/tmRFC/RFC2616/Output/chapter1.html#sub3

I think we need to pick one definition and run with it.  What I proposed
isn't perfect, but I think it's good enough to go with for now.

(P.S. while I'm talking to you, ws-desc issue #65 isn't mine 8-)

MB
-- 
Mark Baker, CTO, Idokorro Mobile.  Ottawa, Ontario, CANADA.
http://www.markbaker.ca             http://www.idokorro.com
Received on Wednesday, 9 October 2002 15:18:10 GMT

This archive was generated by hypermail 2.2.0+W3C-0.50 : Tuesday, 3 July 2007 12:25:09 GMT