[xtm-wg] XTM-CMS - Draft Minutes of Swindon Meetings

[Daniel Rivers-Moore <daniel.rivers-moore@rivcom.com>]

The following are the draft minutes of the CMS meetings of Swindon. My
apologies for the delay.

Any comments/corrections to these minutes from CMS members to me by email
please (visible to the XTM-WG list). Thanks.

Daniel

= = = = = = = = = = = = = = = = = = = = = = = = = =
XTM CONCEPTUAL MODEL SUBGROUPS
Minutes of Swindon Meeting, 13-15 October 2000

Present:

Jean Delahousse
Luis Martinez
Graham Moore
Daniel Rivers-Moore
Bryan Thompson
Matthew West (first day only)
Ann Wrightson

The participants were brought up to date with the work that had been done
since the Paris  meetings.

Several meetings had taken place between Graham and Daniel, and on one
occasion, Chris  Angus. THe result had been a clarification of some of the
issues that need to be addressed.

These issues come under:
1) Areas of ambiguity
2) unmet functional requirements from implementors and use cases.

Below is the first cut list of issues in these areas.

Areas of ambiguity
	- facets
	- naming and typing
	
Unmet func req
	- reified statements
	- occurrences on assocs
	- identify merging topics
	
	- closure -> rdf, exist
	- name value pairs
	- areas of space time
	- value spaces

In addition, the issues logged in Paris against the 13250 should be carried
forward to serve  as a checklist to ensure that the final model does not
overlook any of these.

Graham and Daniel reported that they have some thoughts on the issues of
naming and typing,  and would like to see associations become a subtype of
topic. This should help to meet the  requirements for reified statements,
and help towards achieving "closure".

In order to address the need to deal with areas of space-time, or other
"value spaces",  without falling into the trap of trying to model every kind
of value space that could exist,  an additional subtype of topic
characteristic has been added. This subtype, called an  Extent, is a
reference to a region in some value space, but makes no assumptions about
how  that value space is codified or structured. It was suggestd that a
syntactic method for  addressing value spaces could be through the use of a
URI/URN.

Graham and  Daniel also brought the following list of comments/issues that
had emerged  during a meeting with Chris Angus (invited expert):

Issues on the 13250 UML Model
- Need clear definitions of all the abstractions, so it can be understood
fully as a  standalone document
- Occurrence has members that are shown as Resources but should be locators
  - Consequence: We need a ResourceLocator in the model
 We need to say what a Resource is in a way that covers resources that are
part of the topic  map, and external resources that are the referents of
occurrences
- need to be topics, so you can say things about them (among other reasons)
- can be of many kinds (e.g., but not limited to, representation, mention,
manifestation)
- Topic occurrence should be modelled as an association between topics, one
of which  contains one or more resource locators
  - Consequence: Topic can contain zero or more ResourceLocators
- The word Context might be better than Scope: it is the context within
which a  characteristic is valid
- is the nature of a reference/association to a topic in another map? (asked
but not  discussed fully because we were out of time.)


After this catch-up report, the work continued throughout the three days on
rationalising  the UML model of XTM. In particular, a UML fragment was
developed showing how Association  can be subtyped from Topic, and
Occurrence subtyped from Assication.

This fragment was shown to the plenary in a reporting session at the start
of the second  day. The plenary requested the CMS to continue work on this,
and elucidate whether  everything would end up being a subtype of Topic, or
whether some things (and if so, which  things) are not Topics.


Work proceeded mainly in the form of progressively refining the UML model
and seeing whether  the constructs used to show occurrence as a sutype of
association and association as a  subtype of topic could be applied
elsewhere in the model to achieve simplification. The  issue of Closure was
also in our minds - is it possible for everything to be of the same
underlying type, so that under the various operations a Topic Map can
undergo (such as  merging, or querying), the output can always be of the
same nature as the input or inpugs,  thus allowing recursive application of
such operations. (in the same way as the sum, product  and difference of
integers, is always an integer, or the result returned by a SQL query on a
set of relational tables is always a relational table).



Consideration was given to whether XLink could be used to model
associations. Is everything  a Link? It began to appear that many things are
Links, but not quite everything. We need to  clarify this as we move forward
towards a fully rationalised model, but if it emerges from  this work that
some things are and some things are not Links, within the Conceptual Model,
this would have an impact on syntax, as it would be desirable to use XLink
syntax for those  things that are Links in the model and not for other
things. This is to be taken forward  betgween now and Dallas, and the
conculsions of this work, and the resulting recommendations  for syntax,
should be communicated to the Interchange Syntax Subgroup.



Consideration was also given to the following questions:
- what is a resource and what role does it play?
- can a resource be considered to be a topic (perhaps an anonymous topic -
i.e. one that has  no topic names)
- can identity be considered as an association of a topic with such an
anonymous topic that  contains the resource that is the subject descriptor
of the topic
- what is the relationship between the sense in which everything in a topic
map is a  "topic", and everything in the real world is a "thing", and the
topic map in some sense  represents the real world and makes a view of it
communicable from one person (mind) to  another

Arising from these considerations, we began developing a picture in which
there are two  "chasms", not just one

On the left of the picture are the Subjects. These are in a human mind.
In the middle of the picture is the Topic Map. This is in the computer.
On the right of the picture is the Real World and the Things it contains.

Among the things the real world contains are things that Subjects "mean",
and that Topics  "represent". Among the Things the Real World contains are
also the Resources that are the  Occurrences of the Topics in the Topic Map.

This drawing needs to be refined and adequate prose developed to describe it
and link it to  the constructs in the UML model.


Ann Wrightson developed a set-theoretic view of these concepts, which she
will share with  the group (since the meeting, she has posted this to the
xtm-wg list).


Some time was spent considering the concept of "type" and the various forms
it can take.  Rather than start from the various attributes that bear the
name "type" in ISO 13250, we  considered typing in general, and drew up a
list of different kinds of typing relationship.  We then considered what
their distinguishing features were, and which ones should be  considered to
be in scope for the XTM 1.0 specification itself, which would be in scope
for  public subjects to be published along with XTM 1.0, and what should be
left for user-defined  public subjects and not standardised.

The conclusion was that class-subclass and class-instance were strong
candidates for  inclusion within XTM 1.0.

It was agreed that an instance be an instance of multiple classes.

It was suggested that for the XTM 1.0 specification, a class would only be
allowed to be a  subclass of one class, with support for multiple
inheritance being a matter for XTM 1.x or  2.0, or for Public Subjects
outside the core specification.

However, it emerged from discussion that different people had  different
things in mind when  they used these terms. For some a class-instance is
like an instantiation of a class in  software implementation terms; for
others, class-instance means what others mean by  set-member. Similarly, for
some class-subclass was about specialisation of class  characcteristics
while for others, it meant set-subset. It was agreed that these issues and
differences of interpretation need to be clarified, and a clear semantic
needs to be  provided for whatever concept of class is included in XTM.
Matthew West has considerable  expertise in this area but was not present
during this discussion. It was agreed that his  assistance will be sought
between now and the Dallas meetings, to elucidate these points.

Aspects of typing that were listed to be dealt with through public subjects
were:
- extent
- part-whole
- configuration (e.g. next-to)
- contained-by
- multiple-inheritance for classes
- inheritance of attributes
- strength of a property
- traversal properties (e.g. transitive)



Facets and RDF were both discussed. I was suggested that it might be
possible to do in RDF  what facets attempted to do but did not do
satisfactorily. It was posited that if this were  indeed the case and if RDF
was able to be expressed using Topic Maps constructs (and  syntax), this
could give us a route towards a rigorous topic-based approach to what facets
attempted to do outside the strict topic framework. This, if successful,
would give us the  kind of closure we are looking for while addressing the
perceived weaknesses of the facets  construct. This is work to be taken
forward between now and Dallas.

The Extent model needs to be reviewed. We did not get on to this.

We need to develop adeqauate prose to aid understanding of the UML diagrams
and the diagram  of the two "chasms" alluded to above.

We need to addrss the BOS and where it appears (or not) in the UML model

It was decided that we needed to continue as a subgroup in order to complete
our work in  time for the Dallas meeting. This would happen through postings
to the XTM-WG list, and  through at least one face-to-face meeting, to be
held in the UK, hopefully with Matthew West  present.

We need to complete the XTM work, even if some aspects of the 13250 work
remain incomplete  within the Dallas and XML 2000 timeframe


It was suggested that to help communication between the with conceptual
model subgroup and  the interchange syntax group, a formal liaison should be
established. It was agreed that we  would invite Steve Pepper to be this
liaison from ISS, and we would offer Ann or Jean to be  the liaison from
CMS.

	 		
It was agreed that we would comment on syntax group work that had been
presented to us,  through on-line discussion and expression of our views (of
approbation or concern) about the  various decisions proposed by the ISS. A
formal communication of the consensus of the group  on syntax issues would
be communicated to the ISS before the Dallas meeting


Daniel

= = = = = = = = = = = = = = = = = =
Daniel Rivers-Moore
Director of New Technologies
RivCom
Tel: +44 (0) 1793 792004
Mobile: +44 (0) 7970 893847
Email: daniel.rivers-moore@rivcom.com



-------------------------- eGroups Sponsor -------------------------~-~>
eLerts
It's Easy. It's Fun. Best of All, it's Free!
http://click.egroups.com/1/9699/4/_/337252/_/972430485/
---------------------------------------------------------------------_->

To Post a message, send it to:   xtm-wg@eGroups.com

To Unsubscribe, send a blank message to: xtm-wg-unsubscribe@eGroups.com