[Date Prev] | [Thread Prev] | [Thread Next] | [Date Next] -- [Date Index] | [Thread Index] | [List Home]
Subject: Service Interface Model (graphic attached)
Colleagues,
OK. Here's my take on a visual model representation for
Service Interface (see attached). I also uploaded a copy to our Kavi
site.
It's consistent with the RM and the RA and should serve as a
candidate replacement for the current Fig. 20 on pg. 41 of the WD 0.3 baseline
dated Mar. 4, 2008. The text of Section 4.1.1.2.1 needs to introduce the
notion that visibility has a role to play in the Service Interface (via service
Reachability). This is consistent with the original intent of Fig. 20 but
the bullet points currently do not reflect this fact.
I chose to use the Class classifier in MagicDraw with the
<<interface>> stereotype rather than the MD Interface
classifier, which is non normative against the UML 2.1.2 standard. Very
frustrating at some of these tool embellishments and also its limitations
(compartments for example).
Also, in an attempt to be consistent with Ken's modeling of
the Service Description artifact at a higher level of abstraction, I chose to
model Service Interface as a Class Diagram rather than a Component/Artifact
Diagram. We can debate the aggregate associations as to whether or
not they are loosely coupled aggregates or tightly coupled composite
relationships. I'm open to that discussion but we shouldn't dwell on
it.
I believe Michael is going to take a crack at updating
elements of Sections 4.1.2.2.2 and 4.1.2.2.3; however, we need to be
very careful not to introduce new concepts that are not consistent with the
Interacting with Services Model or even the RM in what we mean by Action (and
Events). Also, we need to be crystal clear that we use the concept of
Service Description (an artifact) to as COMPLETELY AS POSSIBLE to describe a
service; we do not use other terms floating around the cyberether like "Service
Contract" as synonymous with Service Description. The latter term is
pretty common in the various SOA textbooks out there and one can argue
a holdover concept from the older distributed object computing
standards out there like CORBA IDL. Our notion of Service Contract is
very, very specific and is clearly articulated in the RM. In the RA, it is
modeled as part of the Policies and Contracts Model element of the Realizing
SOAs View.
It would be nice to zero in on some consensus to a reasonable
model for Service Interface before our tutorial at the OASIS Composability
symposium.
Cheers...
- Jeff E., JPL
|
[Date Prev] | [Thread Prev] | [Thread Next] | [Date Next] -- [Date Index] | [Thread Index] | [List Home]