RE: Overhauling the discovery part of the architecture document

[In reply to Roger: Yes, I've been working on these for a few months, 
getting feedback from some individuals, and waiting for an appropriate time 
(after the choreography stuff and the holidays) to bring them to the group 
as a whole.  This looks like the right opportunity.]

I've had similar concerns as Hugo about our existing draft:

1. The existing diagrams are very ambiguous, because the draft doesn't 
clearly say what the things in the diagrams represent.  Some terms (like 
"Service Provider") have multiple, radically different definitions in the 
draft.  This degree of vagueness and imprecision in a specification makes 
me very uncomfortable.

2.  The existing triangle diagram is quite misleading, because it strongly 
implies a particular directionality.  For example, the left arc of the 
triangle diagram is labeled "Find" and the right arc is labeled "Publish", 
which very strongly implies that the WS description must originate at the 
Service Provider and be sent to the Service Requester.  This is misleading, 
because in fact, the only thing that matters is that the two parties (the 
Requester and the Provider) AGREE on the WS description that they will 
use.  Who wrote it (either party, a collaboration, or some third party 
committee) is irrelevant.

3. I believe there is far too much emphasis in our draft on the "Discovery 
Role".  After many hours of further thought and analysis, I do not think 
this role is not inherently needed or relevant in MANY common scenarios, 
even though it is need in SOME scenarios.  (See my below-mentioned document 
to understand why.)

4. People see the triangle diagram with the "Discovery Agency" and 
immediately assume that it means UDDI.  We saw this clearly evidenced at 
the last W3C Advisory Committee meeting.

5. The existing diagrams completely omit any acknowledgement of 
semantics.  We do not necessarily have to define any standard way to define 
semantics, but semantics clearly are a part of the whole picture, and 
therefore should at least be acknowledged.  If nothing else, this at least 
allows us to more clearly state how they relate to the architecture and 
what.  But it also allows us to more clearly describe the role of things 
like choreography language, which allows us to move some of the semantics 
from human-oriented documents into machine-processable documents.

As a result of these concerns, I went back to basics and very thoroughly 
studied several very general scenarios.  I then analyzed, from an 
information point of view, exactly what information is needed by whom in 
each of the scenarios, and diagrammed and explained exactly what was 
happening.  I also tried to give fairly precise definitions of the actors, 
artifacts and actions.

The result of this work is at 
http://www.w3.org/2002/ws/arch/2/10/roles_clean.htm . My hope is that this 
provides a much more precise identification of the actors/roles, artifacts 
and actions that are needed in these various scenarios than we currently 
have in our draft.  I hope we can consider this analysis in re-thinking the 
diagrams and relevant sections of our draft.

At 10:29 AM 1/9/2003 -0600, Cutler, Roger (RogerCutler) wrote:
>Mr. Booth --
>
>Would this be an opportune time for you to disgorge the thoughts and
>constructions you have shown me but have not yet exhibited publicly?  It
>seems to me that they are, at least to some extent, responsive to the
>concerns Hugo raises.  And, as you know, I also think that they are very
>good and that we should be looking seriously at them.
>
>-----Original Message-----
>From: Hugo Haas [mailto:hugo@w3.org]
>Sent: Thursday, January 09, 2003 6:19 AM
>To: www-ws-arch@w3.org
>Subject: Overhauling the discovery part of the architecture document
>
>
>
>Hi all.
>
>I am concerned that the discovery part[2] of the architecture document
>is written from a technology point of view, rather than from the point
>of view of exposing a problem, describing the characteristics of the
>problem, and then starting to talk about how existing technologies
>propose solutions to the problem, or parts of it.
>
>I think that it is because this part has been, to the extreme with all
>the detailed description of UDDI and the mention of WSIL, based on
>exposing existing technologies, without concentrating on what problem
>they are solving. I basically think that we are presenting a solution to
>an undefined, or under-defined, problem, and that doesn't seem like a
>sound approach to me, and gives us a biased view of the situation.
>
>I would therefore like to first define the problem and, see how it can
>be addressed, and how existing technologies cover the proposed
>solutions.
>
>Also, I think that the way I see things will make us make progress
>(hopefully) on what a role (in particular "Discovery agency") is.
>
>The current diagram has IMO a couple of issues:
>
>Issue 1: who publishes the description?
>
>I don't think it must be reserved to the service provider. For example,
>if I am a consumer and I want to decide what the service should look
>like, and the provider had better agree if it wants to get the insane
>amount of money I am offering for it.
>
>The current model does not allow such a scenario to happen, which looks
>like a serious limitation.
>
>One key point is that both actors actually need to agree on a service
>description, and then the service description could be published by
>either, or even both.
>
>Issue 2: The "Find" action.
>
>If a "Discovery agency" is used, one (the requestor or the service) may
>*get* a description from it. Find implies that the agency is searchable,
>which may not always be necessary (e.g. a single description on a sheet
>of paper).
>
>I think that there are three things to do in order to address those
>problems:
>- have a diagram in addition to the triangle one, showing that what is
>   needed between the provider and the requestor is an *agreement* of
>   how to proceed.
>- modify the triangle diagram to:
>   + allow for the information flow to go both ways.
>   + not imply that the agency is searchable.
>- rewrite the discovery section in abstract terms:
>   + what are we agreeing on? syntax, semantics, etc.
>   + what needs to be achieved? common view of what is going to happen.
>   + how is it going to be achieved?
>     . via a description agreed on.
>     . the description needs to reach both actors.
>     . etc.
>
>Only at that point can we start talking about documents, Web resources,
>transfer of representation of those resources, potential indexing and
>research, etc.
>
>Comments?
>
>Regards,
>
>Hugo
>
>   1. http://www.w3.org/TR/2002/WD-ws-arch-20021114/#basic
>   2. http://www.w3.org/TR/2002/WD-ws-arch-20021114/#discstack
>--
>Hugo Haas - W3C
>mailto:hugo@w3.org - http://www.w3.org/People/Hugo/

-- 
David Booth
W3C Fellow / Hewlett-Packard
Telephone: +1.617.253.1273

Received on Thursday, 9 January 2003 14:26:15 UTC