Minutes of the Amsterdam meeting/March 27-28

[astes@disi.unige.it]

Minutes of the CoFI-reactive Task group meeting
Amsterdam,March 27-28

Participants
(To all specific sessions)
Egidio Astesiano, Heinrich Hussmann,Christine Choppy, Gianna Reggio,
Francoise Tort,Martin Wirsing
To some sessions
Maura Cerioli, Hubert Baumeister, Eva Coscia, Markus Roggenbach
===================================
 There have been 4 presentations  for the general audience

CoFI-Reactive TG:current status
by Egidio Astesiano (see summary below)
UML/CASL and related:the present status
by Heinrich Hussmann
CASL/Chart
by Gianna Reggio
Labelled Transition Specifications in CASL notation
by Gianna Reggio

*Note: the slides of a UML tutorial given at the previous meeting in Cachan
by Heinrich Hussmann are available at
http://www-st.inf.tu-dresde.de
where also can be found references to available tools such as Argo/UML
----------------
CoFI-Reactive TG: current status (March 27, 1999)
by Egidio Astesiano

AIM AND SCOPE IN BRIEF

The final goal of the group is a formal support for the development of
concurrent, reactive, parallel and  distributed systems; in other word those
systems for which   a purely functional approach is not natural , if not
impossible; hybrid systems are not addressed.
Since in the real development the object- oriented methods have a dominant
role, the group will of course address and use OO aspects as long as they
are needed for reaching the above target, but Object Orientation is not
addressed  as an independent topics in itself.

To achieve that goal the group  shall coordinate the development of
suitable 'extensions' of the basic CoFI/CASL  formalism ,where by
extensions we include methods whose formal basis   includes a significant
part of CoFI/CASL in a conservative way.

CURRENT STATUS

The work is organized in two tracks: autonomous  and coordinated extensions.

A)
AUTONOMOUS EXTENSIONS
They are   autonomously submitted to the Group; they  are required to
follow a suggested pattern, available on the web, and are subject to
approval.

Currently three proposals have been announced (but not yet submitted), all
by the Genova group, on the basis of some previous and also more recent
work:
-CoFI-LTL/Design
It is a method for the design specification of concurrent systems, based on
a formalism of  (structured) conditional specifications defining processes
in terms of labelled transition  systems;the method  also includes a purely
graphical version
-CoFI-JTN
JTN is the Java Targeted Notation,essentially a restricted subcase of   LTL
with shortcut notations,automatically translatable into Java
-CoFI-LTL/Requirements
It is essentially a many-sorted first- order temporal logic.
All the previous announced extensions only need an easy "translation" into
the CASL syntax for the CASL   part.
They are accompanied by  illustrative and signifcant case studies.
Also there is some ongoing  work  on  a CoFI version of
Statecharts(CASLCHARTS).

B)
COORDINATED  EFFORT

It has been decided   to start an effort centered around UML,the Unified
Modelling Language.

We are pursuing two tracks.

CoFI annotated UML
The basic idea is to adopt CoFI/CASL or an extension of it, for annotating
UML, thus possibly replacing OCL  (work led by Heinrich Hussmann,
Dresden).

Precise UML
We have joined the activity of the free group Precise UML; the target is to
provide a formal/rigorous undepinning of UML,possibly exploiting CoFI
related techniques.
 A   sketchy porposal for a general approach to the problem has been
presented at a Workshop at OOPSLA'98 in Vancouver.
Two other  draft papers ar available , one on  relating the ADT approach to
UML and the other on proposing an underlying model for the UML State
Machines.

The work about UML will likely require an extension of CoFI/CASL to handle
state-based systems. This work will have to be coordinated with the people
in the Design group also interested to that topics.
=======================================
The specific sessions have addressed the following issues

A) Workplan revision
B) Report  and disussion on Precise UML
C) Technical issues

 A) Workplan revision

A proposal   presented  after Cachan by Heinrich Hussmann has been taken as
a basis and updated, resulting in the following.

Task   T1
Choice of appropriate UML subset of interest
Responsible:  Tort;(Static part :Tort;  Dynamic part : Reggio )
Task T2
Modelling and Semantics
Responsible: Reggio (Static part :Hussmann;  Dynamic part : Reggio )
Task T3
Constraint language (OCL and CoFI/CASL)
Responsible: Hussmann (Static part :Hussmann;  Dynamic part : Reggio )
Task T3
Case studies
Responsible: Choppy

The work schedule has been organized around deadlines and meetings:
-planned presentations for UML'99 (deadline May 5)
-session at WADT'99 (  May  26for abstracts)
-participation to ETAPS 2000 (October 18 )
-meeting in Paris 18-20 June
-meeting in Genova,January-February 2000

B)Report and discussion on Precise UML
Christine has reported on a meeting she attended in Paris,where it was
explained the adoption of Z as a formal notation in support of a subset of
UML.

It was agreed to follow the work and discussion done within the precise UML
open group.

C) Technical issues

T1
The choice of an appropriate subset of interest of UML has been discussed,
as reported in the following minutes by Francoise Tort.

A discussion on the first task scheduled for the CASL-UML working group,
"choosing a subset of UML notation", has ended at following choices :

Some notations or concepts of the UML class diagram (static analysis),
will be omitted in a first approach, or replaced by others. "Omitted in
a first approach" means that their formal foundation in CASL is not
trivial and will be invastigated at a second stage.

1. parameterized class or template
A template is the descriptor of a class with one or more parameter.
-> omitted for first approach.
Typical parameterized data types, like collections, can be expressed in
UML by a class "collection",  a class "elements", and a composition
association between both classes. The class "elements" is a super-class
of all potential element classes.

2. interface
An interface is a specifier for the externally-visible operations of a
class.
It specifies a part of the behaviour of a class.
-> omitted

3. association-class
An association-class is an association that also has class properties,
as attributes.
-> omitted. When specifying, one must replaced it by a class.

4. qualifier
A qualifier is an attribute whose values serve to partition the set of
objects associated with an object across an association.
Generally used to obtain a unique target object, as keys in data bases.
-> omitted in a first approach

5. dependency
A dependency indicates a semantic relationship between two model
elements. A change to the target element may require a change to the
source element.
-> omitted in a first approach

6. multiple inheritance
Multiple inheritance, polymorphism of operations are allowed in UML.
-> multiple inheritance will be allowed with suitable restrictions (to
be defined).

7. properties of model elements
The properties "isActive" for an attribute, and "concurrency" for an
operation,have to be treated.
All other properties of model elements have to be investigated, how easy
they are to map.

8. Type information
Attributes and operations may not have type information.

The definition of the subset of the class diagram of UML is assigned to
Francoise TORT.

T2
 Heinrich Hussmann started a lively discussion on the static
(non-dynamic/non-active classes) part of UML.
It has been immediately realized that there are,perhaps questionable but
already well  rooted,  features of UML which are not amenable to CASL as it
stands.
I was decided to investigate further this issue.

Egidio Astesiano