From omg-list-errors@emerald.omg.org  Thu May 23 13:19:06 2002
From: Edwin Seidewitz <eseidewitz@intelidata.com>
To: "'adtf-mda@omg.org'" <adtf-mda@omg.org>
Subject: RE: Surface syntax for the UML action model 
RJLRef: $PH/uml2mda/uml2mdaActionModelSyntax_es.020523

[I am resending this message, because I haven't seen it appear on the list.
My apologies if the rest of you have, indeed, already received it.]

Conrad --
> Apologies, didn't mean to suggest a solution, it's just an RFP.  It
> doesn't need to be related to the diagram interchange RFP.  
> That was an analogy.

Actually, I don't think the diagram interchange RFP is the right analogy
here, and this may be causing some confusion. The goal of the diagram
interchange RFP was to create a means to interchange the already-specified
surface notation for UML diagrams (which happens to be primarily graphical).
The action semantics, by intent of the original RFP, did not include a
normative surface notation at all, graphical or textual. So the real issue
is providing the normative surface notation (or notations).

So, when you mention that the "[Mapping to the action model] is analogous to
the mapping between notation and model that is called for in the diagram
interchange RFP," this IS a relevant analogy. But when you go on to say "In
fact, the RFP for action surface syntax is just the textual version of the
diagram interchange RFP," I think this is the confusing point. 

The analogy to HUTN is more apt. Indeed, the main justification for the HUTN
was exactly to "extend the user base to those who are averse to pictures."
Further, the current HUTN submission claims to cover all of UML (which was
an optional requirement of the HUTN for EDOC RFP). If the proposed textual
notation is too cumbersome for action semantics, then this should be
considered as part of the evaluation of how "human usable" it is.

Finally, before we draft any RFP for a action semantics surface language, I
think we need to consider what the relationship of the action semantics is
to OCL. The UML 2.0 OCL RFP (for which revised submissions are imminent)
calls for a metamodel for OCL and a clearer semantic description that will,
of necessity, overlap much of the action semantics (the part that is
"side-effect free"). It would be truly silly to have, in UML 2.0, t[w]o
different action/expression surface notations with two different underlying
metamodels. This would seem to me to be exactly the kind of overlap and
duplication that UML 2.0 is supposed to clean up.

For UML 1.x, I think any consideration of action surface language needs to
start with mapping OCL (which, in UML 1.x, would only cover a subset of the
action semantics, of course). At this point, I am personally not sure how
far we should go beyond that in UML 1.x. In UML 2.0, however, we definitely
should have a consistent action notation that maps to the action portion of
the UML 2.0 metamodel and that includes OCL as a proper subset (and thus
mapping to a proper subset of the action metamodel). We may very well want
to consider a graphical notation, too.

-- Ed