The meeting discussed most of the topics on the published agenda.
It did not have time to address forward planning, but
had full and productive discussions on
Semantics in Enterprise Architecture and UDEF as a Global Interoperability
Enabler. The first of these topics included a report on the deliberations
of the Information Architecture project, which had met earlier in the week.
The meeting was chaired by Arnold van Overeem of Capgemini.
He used a set
of presentation slides to structure and guide the discussions.
After Arnold had opened the meeting, Bob Weisman of
Build the Vision, leader of the Architecture Forum's
Information Architecture Work Group, presented a summary
of the discussions of that work group earlier in the week.
The focus of that project is on information quality,
and it has a vision of an interoperable collaboration environment.
The information is typically used to support decision-making,
and this is important,
but the project is concentrating on the information itself.
Arnold then described a draft Interoperability Framework
(slides 5-15 of his presentation), and this was the basis of an extensive
discussion. Points from that discussion included:
- The framework is related to the EU Interoperability Framework,
but is more sophisticated, including separate ICT, administrative,
and cultural planes.
- Security and identity management are very important,
but are not shown explicitly. Together with governance,
they belong principally in the administrative plane.
In view of their importance, they could be made explicit.
- Good security arrangements improve interoperability
and information sharing.
- The architect has to solve procedural interoperability problems
as well as other kinds of interoperability problem.
There was not, however, agreement that the decision tree
on slide 10 applies.
- Interoperability should not just be point-to-point;
it should be possible to enable interoperability through a pool of shared information.
- The Boundaryless Information Flow™
Integrated Information Infrastructure Reference Model
is an important reference point for interoperability.
Further information can be found on the
TOGAF Interoperability Work Group wiki.
There was strong consensus on the Definition of Interoperability
(slide 6 of Arnold's presentation), and general agreement
on the high-level interoperability framework diagram
(slides 8 and 9 - the diagram is also shown on
the TOGAF Interoperability Work Group wiki).
Further discussion is needed to reach consensus
on other parts of the framework.
The discussion then moved on to the role of the UDEF
as a global interoperability enabler.
Ron Schuldt of UDEF-IT explained the basis of the UDEF,
and Arnold described the proposals for distributed UDEF
(slide 16 and subsequent slides of his presentation).
The following points were made in discussion:
- Governance is essential.
Coordination of the development of trees by multiple authorities
is a major challenge. It is possible because the roots of the descendent trees
are in the main UDEF trees, which are controlled by The Open Group.
- It is possible to link a descendent tree to
more than one place in the main trees. Questions were raised as to
whether this might impair semantic integrity.
- The child nodes of a UDEF object class node
do not represent disjoint subclasses. Indeed,
they appear to represent a mixture of subclasses
and subsidiary roles. This is because the UDEF object classes
are not defined as absolute object classes; they are defined in relation to
the user enterprise, and in fact correspond to relations rather than to classes.
- The concept of Satellite Data Element Tree was broadly agreed,
but there was considerable discussion on the concept of
Private Data Element Tree. There is a trade-off to be made
between the need to enforce consistency to preserve the integrity of the UDEF,
and the need to allow flexibility to encourage adoption.
- Priority should be given to the definition of a concept
of temporary tree space to allow new concepts to be added quickly.
(They would be added initially in the temporary space,
and moved to a final part of the UDEF space after discussion and use-experience.)
- The detailed syntax of distributed UDEF should be transparent to users.
- The names chosen for the different kinds of distributed UDEF tree
may not be the best possible. For example, "Enterprise Data Element Tree"
might be better than "Private Data Element Tree".
- The Open Group should consider developing accreditation or certification programs
for UDEF registrars.
In general, there was agreement on the proposals put forward by Arnold,
except for those relating to detailed definition and numbering of
Private Data Element Trees.
