SDL-News: SDL/MSC versus UML


Subject: SDL-News: SDL/MSC versus UML
From: Richard Sanders (Richard.Sanders#informatics.sintef.no)
Date: Thu Sep 11 1997 - 08:14:58 GMT


The originator of this message is responsible for its content.
-----From Richard Sanders <Richard.Sanders#informatics.sintef.no> to sdlnews -----

At 16:21 1997-09-10 +0200, Claude AUBRY wrote:
>...
>For the diffusion of SDL, an other question is the position in relation to
UML. I just dowloaded the UML 1.1 documents. Till now SDL owned two main
advantages : SDL is an international standard, and SDL is formalized,
allowing simulation and complete code generation. UML1.1 can be considered
as a standard and the semantics are now defined (only static semantics,
but soon dynamic semantics, I quote UML document "Currently, the dynamic
semantics are not considered essential for the development of tools.
However, this will probably change in the future."). The actual position
of SDL tool vendors (use OMT or UML for analysis and SDL for design) could
be difficult to follow if UML offers the same possibilities (and maybe
more) for simulation and code generation.
>
The potential of UML becoming a formally defined language, implying all
that we mean by formal, can in the future have as a consequence that SDL
and MSC fade out. For this to happen a number of things must be fulfilled:

* UML is completely formalized
* an standardization body governs it (like OMG)
* tools that adhere to the standard and provide the necessary functionality
* an interchange format (CIF for UML) is formally defined
* UML is "at least as good" as SDL / MSC in power of expression

None of this has happened yet, but it may very well happen, and in many
ways we must hope this will be the case. Especially if UML includes things
in its basic definition that we find important for modelling.

Currently our opinion is that UML has lots of potential, that it in some
ways is "overloaded" in expression forms (resulting in more complexity than
necessary), and that it has not reached the maturity of SDL and MSC -
neither in terms of language definition or tools.

Comparing SDL process description to UML StateCharts there are some
advantages and disadvantages with both in terms of expressive power. The
same goes for MSC versus UML Sequence Diagrams. None of them are perfect.
Hence we support two strategies:

* improving SDL and MSC (better standards for the year 2000)
* hope that UML will take the best from MSC and SDL

If UML were to include the things in SDL and MSC that we in the SDL/MSC
user community value, then this would be a major breakthrough in systems
engineering. Currently we are not certain that this will happen, UML being
influenced by many sources.

Our current advice to system designers is to combine UML notations (and
tools) with SDL/MSC, with links from a domain model in UML to the design in
SDL, while MSC or UML Sequence diagrams are used to describe interaction.
Then users can continue to detail system aspects that are not designed in
SDL, by designing (and implementing from) UML.

This is in essence the view we take in TIMe - The Integrated Method - that
we and the SISU project have produced (and which we plan to demonstrate at
the SDL Forum). See http://www.sintef.no/time for more information.

TIMe will very soon be commercially available (we have - finally -
informally been told by the ITU that we may include extracts from the
standards in our electronic textbook).
...................................................................
Richard T. Sanders Richard.Sanders#informatics.sintef.no
SINTEF Telecom and Informatics Telephone: (+47) 73 59 30 06
N-7034 Trondheim, Norway Telefax: (+47) 73 53 25 86
Home: Aasbakken 1, N-7043 Theim Mobile: (+47) 930 58 954
phone 73 52 06 71 http://www.informatics.sintef.no/~richards

-----End text from Richard Sanders <Richard.Sanders#informatics.sintef.no> to sdlnews -----
For help, email "majordomo#sdl-forum.org" with the body of your email as:
    help
or (iff this does not answer your question) email: owner-sdlnews#sdl-forum.org



This archive was generated by hypermail 2a23 : Sun Jun 16 2013 - 10:41:39 GMT