Re: Where are we at? Where to from here?



Re: Where are we at? Where to from here?

From: Prof. Dr. Peter H. Schmitt <pschmitt_at_ira.uka.de>
Date: Fri, 09 Jun 2006 10:02:50 +0200
Message-ID: <44892B2A.6000601@ira.uka.de>
Here are - somewhat belated - my comments on the discussion on
formal semantics for UML.
I will be brief, since what I have to say has in one way or another
been said before.

First observation
There still are a number of issues in the semantics of UML class 
diagrams and OCL that are shaky.
I am in particular refering to the issues that have not been resolved in
the OCL draft from June last year.

Second observation
A formal semantics using a heavy formalism is not the way to go in a 
standard. Just writing things in a higher order logic or giving a UML
meta-model for it does not help much. For me this is a second step.
Once the meaning is clear, everybody can produce a formalisation in his
own favorite formal method for what ever purpose s/he needs it.

To Do
Give a precise informal meaning of class diagrams and OCL!
I propose to the use the language of mathematics, that is English.
You will also need a few basic concepts: a naive understanding of sets,
functions and relations. Also the basic set-up, which is of course
there in most publications on the topic, needs to be stressed. In 
particular the notion of a snapshot (instantaneous description, system 
state or what ever you want to call it) and the notion of when an OCL 
expression is true in a given snapshot (or in all possible snapshots) 
and when the constraints of a class diagram are satisfied in a snapshot 
are the main goals. For OCL1.4 the semantics of Richter and Gogolla is a 
good example. But it should be extended to OCL2.0 and could be presented 
in a more pallatable manner.


Cheers

Peter H. Schmitt




Ian.Oliver@nokia.com wrote:
> I though I'd just make a few "industrial" comments regarding this:
> 
> "Is it still a good idea to try to make UML precise?"
> 
> Yes, but what you do mean by precise - see the excellent Harel & Rumpe article at http://lambda-the-ultimate.org/node/1501
> 
> Formality however does tend to scare some industrial people with visions of obscure Z-like symbols :-)
> Can we formalise the extension mechansims too?
> 
> "Can work outside the large IBM funded project have any influence?"
> 
> My take would be: if you can't beat them, join them...
> 
> "Perhaps we should look instead to a new, well defined language, or giving UML-like notation to an existing formalism?"
> 
> Bad, bad, bad :-)   Making SDL look like UML is once instance of this and it has caused all sorts of weird problems, both syntactical and semantic.
> 
> 
> My line would be that UML has reached a place where either it must return to its roots as a language for describing systems using OO concepts or become an unusable syntax for a whole host of domain specific languages, none of which are compatible...personally I prefer the former, which UML, IMHO is better at
> 
> Ian
> 
>>-----Original Message-----
>>From: puml-list-request@cs.york.ac.uk 
>>[mailto:puml-list-request@cs.york.ac.uk] On Behalf Of ext 
>>Pierre-Yves Schobbens
>>Sent: 14 May, 2006 22:59
>>To: puml-list@cs.york.ac.uk
>>Subject: Re: Where are we at? Where to from here?
>>
>>
>>Le 14-mai-06 à 11:00, Greg O'Keefe a écrit :
>>
>>>Precise UMLeers,
>>>
>>>I am interested to know your opinions of the state of play in making 
>>>UML precise.
>>>
>>>
>>>Is it still a good idea to try to make UML precise?
>>I think yes; and also the main players of UML are more 
>>receptive to the idea than before.
>>
>>>Can work outside the large IBM funded project have any influence?
>>If we want to have influence, it is useful to set up large consortia  
>>involving both industrial and academics.
>>
>>>Perhaps we should look instead to a new, well defined language, or
>>>giving UML-like notation to an existing formalism?
>>I don't think so. There are many academic papers in this vein,
>>"I take my favorite language and disguise its syntax into an UML-ish  
>>syntax". However, the results are usually rejcted by industrials .  
>>Further, the unifying aspect of UML is usually lost, since UML is  
>>understood in a very specific way, usually incompatible with the  
>>standard (UML2.0 Superstructure) and certainly with all variants of  
>>this approach.
>>
>>I think we should start afresh from the approved UML2.0  
>>Superstructure, and try to be faithful to it as far as possible.
>>
>>
>>
>>



-- 
Prof.Dr.P.H.Schmitt
Fakultät für Informatik
Universität Karlsruhe
Received on Fri 09 Jun 2006 - 09:03:07 BST