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

RE: Label for Top Node of "triangle diagram"

From: David Booth <dbooth@w3.org>
Date: Wed, 02 Oct 2002 16:04:34 -0400
Message-Id: <5.1.0.14.2.20021001110157.033d6860@localhost>
To: www-ws-arch@w3.org

Dave Hollander <dmh@contivo.com> on Fri, 27 Sep 2002 07:55:43 -0700 said:
>. . . I believe that Heather's point about not liking "repository" is 
>because of
>the presumption of storage--and I feel the same about registry. The roles,
>as I understand them, are: 1) a "place" to advertise a service's 
>availability 2) an
>agency that brokers services' descriptions 3) a "place" to discover what 
>services
>are available.  Are these right?

I think this muddies the distinction between nouns (roles) and verbs 
(actions) a little.

The really nice thing about the triangle diagram is that it concisely 
illustrates:
(a) the ROLES of Client and Web Service; and
(b) the sequence of three ACTIONS:
       1. the action of advertising a Web Service;
       2. the action of discovering a Web Service; and
       3. the action of invoking a Web Service.

The actions (verbs)  are the right, left and bottom arcs in the 
diagram.  The roles (nouns) are the bottom two nodes.  And the top cloud is 
in discussion.  (To my mind the top cloud is the Web itself, or in the most 
general case, anything that is electronically accessible.)

It is very helpful to have these two roles and three actions illustrated 
together in one concise diagram.

Why didn't I say THREE roles? If you assume that Client and Web Service 
agree to use some kind of common "registry" to advertise and discover a Web 
Service, then it makes sense to think of the top cloud in the triangle 
diagram as also being a "role".  So if you accustomed to thinking of things 
like UDDI, that would be natural.  But I do not think we should be making 
this assumption.  The Client and Service do NOT need to agree on a common 
registry (again, unless you consider the entire Web to be a registry, but I 
don't think that's a helpful characterization).

All we need to assume is that the Service has some means of advertising its 
description (which could be as minimal as its URL if it provides a 
self-describing interface), and the Client has some means of discovering 
that description.  The Client and Service do NOT need to have any 
previously agreed "meeting place" (again, short of the entire Web) for 
advertising and discovering that description.  If they are using the Web 
then the Service could use one means of advertising and the Client could 
use an entirely different means of discovery.

The reason I am harping on this is that I believe that if we call the top 
cloud "a role" we are implicitly suggesting that there needs to be a 
central or previously-agreed "meeting place" for the Service and Client to 
publish and discover Web Service descriptions.    Certainly, some Services 
might CHOOSE to only advertise themselves using means that DO require a 
previously agreed "meeting place" for their discovery.  That's  fine.  But 
we should not bias our architecture my implying that they MUST do so.  One 
of the most powerful features of the Web is the ability to have this kind 
of decentralization.


-- 
David Booth
W3C Fellow / Hewlett-Packard
Telephone: +1.617.253.1273
Received on Wednesday, 2 October 2002 16:06:38 GMT

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