Broader layering issue [Re: Message exchange patterns in the architecture]

Hugo,

You've hit upon a very important issue for our architecture.  Layering the
various concepts (and their related technologies?) may result in some
clarity in this area and could help us decide what the true architectural
elements are in our architecture.

When considering how the technologies we're considering relate to each
other, we need to be careful when features we'd originally placed in a
lower layer must be exposed higher up.  When I use most HTTP tool kits, I
can ignore DNS, IP and TCP for the most part.  It's possible that when I
(someday) program using a SOAP 1.2 tool kit, I'll need to be mindful of
underlying HTTP, SMTP, MOM features.  While these examples relate to API
layering and not directly to the architecture, the analogy between the two
should be strong.

Our earlier discussion of routing and choreography raised another related
issue for me.  If choreography languages may be used to describe some
routing scenarios and a routing runtime appears to provide (rather limited)
choreography support, does this mean they exist at the same layer?  I
suspect the answer is "no" and our intuition that choreography sits above
routing is correct.  We may however find specific situations in which these
layers result in less performant solutions than required.  This could imply
situations in which we implement tool kits straddling layers that remain
isolated in the architecture.  Let's not collapse the layers for
performance reasons on the first run through.

thanx,
    doug

Hugo Haas wrote:

> As per my action item from yesterday's teleconference, I am
> articulating here the problem about message exchange patterns (MEP) in
> the Web services architecture.
>
> The architecture draft[1] discusses MEPs as an architectural element.
> They currently appear in the document both as a SOAP feature and a
> WSDL description attribute. They are likely to show up in one form or
> another in other places, e.g. in choreography.
>
> An interesting question is to know what relationships MEPs have in
> each of those "layers", if any. I would therefore like to add an issue
> in the document about how MEPs translate from one "layer" to another.
>
> Some background discussions on this:
>
>     Discussion in the WSCG list:
>     http://lists.w3.org/Archives/Public/www-ws-cg/2002Jul/thread.html#1
>
>     A (short) document comparing SOAP's and WSDL's approaches:
>     http://www.w3.org/2002/ws/cg/2/07/meps.html
>
> Regards,
>
> Hugo
>
>   1. http://www.w3.org/2002/ws/arch/2/08/wd-wsa-arch-20020821.html
> --
> Hugo Haas - W3C
> mailto:hugo@w3.org - http://www.w3.org/People/Hugo/

Received on Friday, 30 August 2002 12:17:03 UTC