- From: <Daniel_Austin@grainger.com>
- Date: Mon, 24 Mar 2003 16:22:30 -0600
- To: jdart@tibco.com
- Cc: public-ws-chor@w3.org, public-ws-chor-request@w3.org
Hi Jon,
Good summary. My comments below.
Regards,
D-
"Jon Dart"
<jdart@tibco.com> To: public-ws-chor@w3.org
Sent by: cc:
public-ws-chor-req Subject: requirements summary
uest@w3.org
03/24/2003 03:32
PM
Please respond to
jdart
I'd like to try to summarize some of the recent discussion as it
relates to requirements. I have at least seen some clear preferences
emerge, although all of them may not be shared universally .. in fact
I'd be surprised if they were. But it may help to try to write down some
of the expressed positions, especially those that appear to have wide
support. N.b. I'm not going to define all my terms: that's another
thread.
1. External vs. internal: as Martin mentioned at the end of the F2F,
at least participants seem to agree that modelling an "external" view
of interactions is necessary. Whether an "internal" view is also
needed is TBD.
2. Multi-party vs. bilateral choreography: there is some skepticism
that modelling bilateral interactions is sufficient.
I certainly don't think that is it sufficient to model only bilateral
transactions. Many business transactions have multiple actors, and we want
to build standards that will work for common service transaction models.
3. There also seemed to be agreement that dynamic participation is a
required feature (i.e., not all participants in a choreography may be
known in advance).
This is a must-have - limiting ourselves to predetermined transaction
sequences will not work; this is inconsistent with the WS architecture
model and with the needs of the users of WS technology.
4. There is an expressed preference not to require a dependency on
WSDL (although as Martin noted there is a requirement in the charter
that WSDL 1.2 be supported). (Maybe abstract WSDL is ok?)
Dependency and support are two different animals - we can support
WSDL, without requiring it. This is probably the best approach, IMHO.
5. Declarative vs. executable model. Related to point 1, and also to
the discussion re context-free vs Turing-complete models at the F2F.
This is an area where I see continued disagreement. However, I think
this is not so much a requirements issue as a decision that will result
out of other requirements. I.e. other requirements and use cases will
drive the group to take one or another approach, when it comes to the
specification phase.
I agree - this is a "how" question, not a "what". Let's establish our
ends, and then argue over the means. :)
Received on Monday, 24 March 2003 17:22:22 UTC