[Modeling] Document structure

Radovan Cervenka rce@whitestein.com
Fri, 21 Feb 2003 11:37:57 +0100

Dear Marc-Philippe and all,

> > > b. Description: do we speak about the description of the class or the
> > > semantics of the class, the former is usually a natural language
> > > document, the latter a formal definition
> >
> > I was thinking about two sections: description and semantics. But, they
> > more-or-less about the same, because semantics is defined in natural
> > language as well. Therefore I put them into one. But no problem for me
> > people decide to have them separated.
> Not at all, never say to some formal people that semantics is defined as a
> natural language document, you will be killed at that moment, if you
> semantics, define it formally or do not define it, OK, UML missed this
> but not formal methods

:-))) I see... I was talking about OMG documents. E.g. UML and all related
specifications define semantics in natural language - see e.g. UML 1.4.
Semantics - semantics defined in English and some constraints of abstract
syntax in OCL. BTW how to define syntax (meaning) of modeling elements
formally? AUML is intended to be a language taht is usually marked as
semi-formal. Any opinion?

> > > Once again, we do not feel obliged to follow strictly UML and maybe we
> > > can have several other fields in the class description
> >
> > I agree. Do you have some ideas about further fields?
> Actually not really, I just think about what Bernhard did in agent class
> the class is no longer a UML class with just attributes and operations but
> services, protocols, organizations, etc

UML defines compartments for classes that may contain "everything", not just
attributes and operations. People usually define also other compartments
like: responsibilities, exceptions (see Figure 3-23 of UML 1.4.
specification), tags and constraints (Figure 3-32 of UML 1.4.
specification), etc. I have seen also CASE tools supporting more

What you put into class compartments depends on what meta
properties/features (attributes, roles) you specify for your metaclasses. We
will see what further info will be applicable for different types
(stereotypes) of AUML classes...

> > > XMI might be a good idea as long as we share some strong links with
> > > else it would be more efficient to define our specific language (based
> > > on XML?)
> >
> > I'm afraid we will have not enough time to define our format...
> It depends, as far as I am concerned, I found XMI specs really difficult
> read and if you don't have a tool to play with the format, you have no
> idea what is it exactly, reinventing the wheel is something computer
> are good and maybe a new format won't be difficult to get, based on XML
> some ideas of XMI...

Sure XMI cannot be used without tool. Any interchange format cannot be used
without tool easily. BTW we are talking about metamodel written in UML -
visual aspect is important as well. Everybody likes picture much more than
some "stupid" XML document. A tool is necessary.

> > > Tool? Well, I strongly agree that Rational Rose is a good solution, my
> > > concern is money, what to do if your team does not consider this
> > > expensive tool as essential for the team? Does it mean I am out for
> > > several things?
> >
> > We will see... In the worst case we can exchange just pictures... But
> > is winner for now :-)
> I know, I know but that would be really frustrating for "poor" people like

I see... There are several free UML tools. E.g. Umbrello, Argo/UML, etc.
support also XMI.


Radovan Cervenka | rce@whitestein.com
Whitestein Technologies | www.whitestein.com
Panenska 28 | SK-81103 Bratislava | Slovak Republic
Tel +421(2)5443-5502 | Fax +421(2)5443-5512
If you are not the intended recipient of this email,
you are not authorized to make any use of it;
please delete it and notify us by return email.
Thank you.