[Date Prev] | [Thread Prev] | [Thread Next] | [Date Next] -- [Date Index] | [Thread Index] | [List Home]
Subject: RE: [Fwd: Re: [wsbpel] abstract process strawman]
Monica,
I think conformance to the specification is a separate issue -- it is
more like validation beyond just schema and this is what I would like to
see us use Issue 84 to solve.
As we have discussed before, there are multiple notions of conformance
between abstract and executable. We are not going to forbid people
inventing new notions of conformance, including even notions of
conformance between two abstract processes in a use case of successive
refinement.
What we should do is define some salient notions of conformance in the
specification, especially those where we believe we can contribute
sufficient technical content to make it worthwhile. I believe that
behavioral conformance definitely fits in the latter category since it
is non-trivial to define, to say the least.
I do not believe we will mandate any implementation level requirements
for abstract-executable conformance verification, such as monitored
conformance, but I am open to being educated.
Satish
-----Original Message-----
From: Monica J. Martin [mailto:Monica.Martin@Sun.COM]
Sent: Tuesday, September 28, 2004 10:44 AM
To: Satish Thatte
Cc: Danny van der Rijn; rkhalaf@watson.ibm.com;
wsbpel@lists.oasis-open.org
Subject: Re: [Fwd: Re: [wsbpel] abstract process strawman]
Satish Thatte wrote:
>I don't understand the distinction you are drawing when you write
>"guidance or compatibility between the abstract and one of the
adjoining
>executable processes rather than conformance".
>
>
mm1: The latter infers more rigor and requirements on the
implementation. Guidance is only that - this is a best practice and
recommended that an executable process be conformant to an abstract one.
However, it's not dictated by the specification. Conformance, although
voluntary, infers that, for example, the mandatory functions of a
specification can be tested against in a verifiable way. With the case
of abstract-executable, the conformance is two-fold - to the
specification and the executable to the abstract. Another discriminator
would be:
* Monitored or passive: An executable process could be compatible
with an abstract process. This could still support conformance of
an abstract or executable process to the specification. It however
does not require an executable process to be conformant to an
abstract process. It could be recommended.
* Active or directed: An executable process is evaluated to be
conformant with an abstract process. The use of executable process
could be constrained by the conformance to the abstract process.
It also infers that the abstract and executable are conformant to
the specification.
I am not supporting one over the other, but making the distinction as it
is relevant to this discussion and what it infers on implementations.
Thank you.
[Date Prev] | [Thread Prev] | [Thread Next] | [Date Next] -- [Date Index] | [Thread Index] | [List Home]