RE: [soa-rm] On UML

[<tmathews@lmi.org>]

Frank - 

I think I see your point.  The target audience for this specification
would be the determining factor.  If the audience of the SOA RM was the
non-technical strategic thinker, then UML certainly would 'not' be the
answer.  I do feel, however, that since this is a SOA RM, denoting both
architecture 'concepts' and RM as the end objective, the deliverable
will ultimately provide better value to all of us 'techies' if using a
common standard (wonder that!) to represent the model.

I would also argue that UML helps to funnel complex mind maps with 'many
relationships' into something more tangible, easier to grasp for many.
We want to avoid abstraction to the point of obfuscation of concrete
relationships wherever possible.

I believe the mind maps to be a viable solution to expand further
'predicate logic' (as you put) in the informal model.  Not sure I see
your number 3 as a disadvantage.

TM

-----Original Message-----
From: frank.mccabe@us.fujitsu.com [mailto:frank.mccabe@us.fujitsu.com] 
Sent: Wednesday, March 23, 2005 12:11 PM
To: soa-rm@lists.oasis-open.org
Subject: [soa-rm] On UML

The issue with UML are
1. A somewhat spurious sense of formality. In fact, the semantics of
even UML 2.0 are less than clear on even the basics such as what does
inheritance mean?
2. The strong bias towards one relationship: isa, when in reality we
will need a great many more relationships: describes, has, owns,
implements, responsible-for, correlates, etc. etc.
3. The strong bias (on the part of certain vendors) to automatically
generating code from UML, resulting in a tendency to 'go concrete' very
quickly.

An alternate suggestion: mind maps. A mind map is simply a collection of
concepts and relationships connecting them. V. simple, and in my
experience v. flexible. It is also possible to develop rigour by mapping
into predicate logic.

Frank