Good thoughts Len. I think the
answer comes down to: how much experience is needed to start a standards effort
for an SOA reference model?
One answer says that we have already
been doing SOA for decades (CORBA/DCOM, N-tier architectures, etc.), so all of
that experience is brought into this effort. And then one can add what's
essentially the newest aspect: use of the Web. Another answer says yes, let's
have more real-world experience in the "newest" form of SOA before we
proceed.
Another viewpoint says that there is
much confusion about what SOA is, and an effort such as this - if done properly
- can help clarify things. If it's not done properly, then we're back to square
one minus 1000.
This TC is taking a very
"minimalist" approach (IMHO) by specifying only those aspects of SOA that are
required for all service-oriented architectures, and being more abstract than
concrete in definition of these aspects. For example, aspects such
as orchestration and security have - so far - been treated as not required
for all service-oriented architectures. Therefore, orchestration will not be a
part of the reference model (the TC has felt that orchestration is really part
of what is being called "process-oriented architectures", and not
service-oriented architectures). Additionally, security is currently being
represented under the overall umbrella of "policy".
We have had discussions about
creating more concrete representations separately from the reference model, but
based on it. Such representations could contain aspects such as
orchestration.
The following publicly available
e-mail contains the latest SOA-RM draft: http://lists.oasis-open.org/archives/soa-rm/200505/msg00009.html
Joe
Booz Allen Hamilton
Visit us online@ http://www.boozallen.com