Mapping between CDIF and EXPRESS for a Study Period on
Mapping Modelling Languages for Analysis & Design

Hugh Davis (Windcrest Consultants)

31st July 1998

1. Summary 

This report is about work done on a mapping between CDIF and STEP/EXPRESS
[CDIF-EXP] in a study 
period on Mapping Modelling Languages for Analysis & Design.  

The mapping between CDIF and EXPRESS has established a proof of concept for
the feasibility of mapping 
between modelling languages by defining a general correspondence between CDIF
and EXPRESS and 
specific transformations from CDIF to EXPRESS and from EXPRESS to CDIF.

These mappings provide a format that can be followed by other mappings and
thus have established the basic 
elements of this mapping process.

2. Background

The problems that give rise to the subject of the study period are as
follows.
- There are many groups (formal standards, industry consortia) defining
models (sometimes referred to as 
metadata). 
- Although there are signs of convergence on some of the modelling languages
(e.g. CDIF-OMG), there 
will never be agreement on a single approach.  
- Therefore the ability to map between different modelling languages is a
firm requirement, and a common 
(potentially standardized) mapping approach needs to be investigated and
developed.  
- The ability to exchange models is a prerequisite for communication between
system designers (and 
others), sharing and exchange of data and interoperability of system
components. 

This problem was studied at a JTC1 workshop (reference [JWKS]) in September
1996.  One of the workshop 
resolutions, later accepted by JTC1, was directed at SC7:

JWS R2.3:	The Joint Workshop requests that JTC1 recommend that JTC1/SC7
conduct a study period 
to establish the technical basis for a program of work leading to the
standardization for the 
mapping of modelling languages. 

Given these terms of reference, SC7/WG11 proposed an initial study period
focussed specifically on 
mapping the following modelling languages:
- CDIF meta-meta-model,
- STEP EXPRESS, and
- OMG Unified Modeling Language (UML)

The objectives of this study period were:
- to establish a proof of concept for the feasibility of mapping between
modelling languages;
- to establish the basic elements of this mapping process;
- to prepare the plan for follow-on work in the relevant JTC1 committees,
e.g. SC7, SC32, SC33, 
TC184/SC4, and liaison groups, e.g. EIA CDIF, OMG.

3. Results

This subsection describes the results of the work on the CDIF-EXPRESS
mapping, particularly the general 
lessons for mapping of modelling languages that have been learned from the
work on the CDIF-EXPRESS 
mapping, concerning e.g. the objectives of the exercise, the different types
of mapping and whether there is a 
need for a reference model.  
 
3.1.1 Scope and purpose of the mapping exercise

Many of the results relate to topics discussed at the first study group
meeting in London.  Various 
judgements are coloured by the objectives of a mapping.  The most important
distinction is whether the work 
is intended 
- to be about modelling languages in general (or all modelling languages),
and to aid understanding of the 
principles of modelling and modelling languages; or 
- to be centred on one modelling language and to aid harmonisation of one
particular modelling language 
with others.  

The study period has taken the latter position, with CDIF as the particular
modelling language, because:
(a) there are insufficient resources and support for a general, long-term
activity;
(b) the results of such an activity would probably become just another
modelling language and would be 
accepted as neither general nor global;
(c) useful results can be achieved in the short term with specific CDIF
mappings;
(d) a few specific CDIF mappings will enable us to make statements about CDIF
mappings in general and 
even about mappings between general pairs of modelling languages.

In London we also noted the distinction between 
- a mapping as a correspondence or comparison of two systems 
- a mapping as a transformation from one system to another system

Those taking a more general position on the aims of the exercise tended to
prefer the former choice; those 
taking the CDIF-centred position preferred the latter.

The work on the CDIF-EXPRESS mapping clarified the distinction and the
respective strengths and 
weaknesses.  This in turn shed light on the question of reference models.

3.1.2 Correspondence versus transformation

It is clear from the work on mappings between CDIF and EXPRESS that both
correspondences and 
transformations are required.  
- A correspondence, i.e. a symmetrical mapping of nearly equivalent systems,
is useful for discussing 
where the equivalence is inexact or where one system lacks an equivalent
feature of the other system.  
- A transformation, an asymmetrical mapping from a source system to a target
system, is necessary for 
generating new features in the target system that are precisely equivalent in
function and meaning to 
features in the source system.  Definition of a transformation encourages
more thorough work on 
difficult differences between the two systems.
Clause 7 of [CDIF-EXP] gives detailed definitions of these terms.

Disagreement over some technical details in Version 1 of the CDIF-EXPRESS
mapping resulted in 
prolonged discussion of the design criteria for transformations.  One
criterion that was generally agreed is 
that the transformation can be automatically applied by a computer program. 
This may produce somewhat 
unfriendly results and so is in conflict with another criterion: that the
results should be in the style of the 
target system and acceptable to its users.  The conflict was resolved in
defining a transformation (called 
mechanical) that meets the first criterion and defining differences for the
transformation (called natural) to 
meet the second criterion.  It was pointed out by EXPRESS experts that there
are different views on what is 
good style and that it may depend on what the EXPRESS schema is to be used
for.

It has been left for subsequent review to determine such questions as whether
it is worth making the 
mechanical/natural distinction, whether the mechanical transformation can be
simplified to make it 
reasonably acceptable to EXPRESS users and so on.

A further criterion that was recognised is that the mapping should be as
simple and uniform as possible and 
omit unnecessary detail, including detail that can be included in separate
general rules.  The aim of this 
criterion is to simplify implementation by tools and use by people.
. 
3.1.3 Use of a reference model

In London we discussed whether the mapping should be structured according to
CDIF concepts or an 
independent reference model.  There were strong reservations about the latter
choice among those making 
the detailed investigation, because:
(a) there is no such independent reference model already available;
(b) a new reference model would be seen, judged and rejected as a proposal
for the basis for a new, 
ultimate, general modelling language;
(c) the reference model would simply add to the variety of alternative terms
and concepts;
(d) CDIF, intended as a neutral architecture and information model, should
serve adequately as a reference 
model.

In fact, most of the CDIF-EXPRESS mapping is structured as corresponding
statements about CDIF and 
EXPRESS (or STEP) concepts, each using the native terms, in adjacent entries
of a table.  CDIF and 
EXPRESS terms are both used in clause headings.  There are also places where
general terms are used, such 
as "module", "type" and "subject area" - some of which may also be special
CDIF or EXPRESS terms - e.g. 
in clause headings and preambles.  All this was done to avoid
"CDIF-centricity" and encourage acceptance 
and joint ownership by both CDIF and STEP communities.  Whilst the mapping is
specific it is not too 
specific to CDIF and the same approach could be used with other pairs of
modelling languages.

The correspondence between CDIF and EXPRESS may be said to use an implicit
reference model: writing it 
required a lot of thought on the separation and classification of concepts. 
So the CDIF-EXPRESS mapping 
provides a format for mappings which can be used (possibly with extensions
and modifications) as an 
implicit reference model.

It would be nice to be able to write similarly that the format of the
transformations from CDIF and from 
EXPRESS provide a suitable structure for defining other transformations from
CDIF, EXPRESS and other 
modelling languages.  It is left to subsequent review of [CDIF-EXP] to decide
if the ordering of the material 
for the transformation from EXPRESS is too much influenced by the ordering
for the transformation from 
CDIF.  To some extent the ordering of such transformations should be based on
the ordering of the source 
language, since the writers of that specification have already worked out an
acceptable way to do this.

In the study period the following mappings of CDIF and EXPRESS have been
carried out:
- a correspondence between the CDIF meta-meta-model and EXPRESS, which is the
basis for
- a transformation from the CDIF meta-meta-model and EXPRESS, and
- a transformation from EXPRESS to the CDIF meta-meta-model.
These mappings are all included in a draft British Standard [CDIF-EXP].

4. Recommendations from study period

4.1 General applicability

The study period has produced specific results that are expected to lead to
useful international standards for 
mappings between CDIF and EXPRESS and between CDIF and UML.  These results
can also be applied in 
other mappings of CDIF and in the mapping of modelling languages more
generally.  Taken together, such 
mappings provide a basis for interoperability between a range of systems.

R1. It is recommended that CDIF mappings be parts of a multi-part standard
CDIF Mappings..
 
 The correspondence between CDIF and EXPRESS provides a suitable basic format
and style, or reference 
model, for defining correspondences between modelling languages.  The
transformations from CDIF and 
from EXPRESS provide a suitable basic format and style for defining
transformations of modelling 
languages.
 
R2. It is recommended that future mappings follow the format and style of the
CDIF-EXPRESS mapping.
 
 The experience from this study period can be used to provide general
guidelines for CDIF mappings.  
 
R3. It is recommended that Part 1 of CDIF Mappings be Guidelines for CDIF
Mappings.
 
 Producing a CDIF-EXPRESS mapping structured like an earlier CDIF-PCTE
mapping has not only provided 
a good starting point for the CDIF-EXPRESS mapping, but also shed light on
both mappings.  
 
R4. It is recommended that future mappings be compared with existing
mappings.
 
4.2 Future standards

The CDIF-EXPRESS mapping is a complete draft of CD quality and has been
received with interest in both 
CDIF and EXPRESS communities. 

R5. It is recommended that the CDIF-EXPRESS mapping be distributed informally
to SC7 for comment and 
that the British Standard, when approved, should be submitted to JTC1 through
the fast-track procedure.

4.3 Quality

The use of compilers for the EXPRESS modelling language has made a vital
contribution.  It is regrettable 
that the primary format for the CDIF modelling language does not lend itself
to automatic parsing of the 
CDIF standards.

R6. It is recommended that proposals be solicited for a simple parsable
syntax for the CDIF meta-meta-
model that can be used in CDIF standards, so as to allow CDIF standards to be
parsed and checked by 
software.

The use of tools for one modelling language (EXPRESS) has shown up
differences between the modelling 
languages.  For complete validation of mappings and other interoperability
specifications it is necessary to 
prototype the whole system.

R7. It is recommended that complete prototyping of the mapping be encouraged.

5. References

[CDIF-EXP]	Mappings of CDIF and EXPRESS, version 3, BSI
[EXPLRM]	ISO 10303-11:1994, Industrial automation systems and integration -
Product data 
representation and exchange  - Part 11: Description methods: The EXPRESS
language 
reference manual.
[JWKS]	Report of the Joint Workshop on Standards for the Use of Models that
Define Data and 
Processes of Information Systems
http://www.nist.gov/workshop/jtc1-96/report.htm
[STEP]	ISO 10303-1:1994, Industrial automation systems and integration -
Product data 
representation and exchange  - Part 1: Overview and fundamental principles.