[TIH WG3: Data Dictionary & Registry] Registry data types

[Cornwell, Ian H]

TIH Working Group on Data Dictionary, Data Registry & Data Models

I'd like to draw your attention to a technical policy change in the ITS
metadata registry (www.itsregistry.org.uk). Up to now we have allowed
some basic data types to exist outside the regular registry process, so
they don't have any specific status. We have now decided to consider ALL
datatypes to be registered items with status. This has a number of
advantages: the datatypes will not escape the quality controls present
elsewhere in the registry; it would be inconsistent for a Recorded-level
attribute to refer to an undocumented root-level datatype for example,
but now the general rule excluding dependencies to lower status items
will apply; the eventual establishment of Preferred datatypes will
provide a clear signal to modellers, and will help in promotion of other
derived datatypes in models already submitted. The approach is arguably
closer to ISO 14817 which has "Value Domain" as a registered item
(although it also recognises ASN.1 basic types outside the registry
process, which is more like our previous scheme).

There are now a collection of datatypes under Recorded and three
datatypes in Draft. These are not new datatypes: they are the types that
have been referenced in various submitted models but for which no
definitions exist within those models, i.e. the modeller considered them
as basic types, declared a priori. These now have status as registered
items, joining the various data types which are defined (and remain)
inside specific models. I have also been able to rationalise to remove
some redundant datatypes without losing any information.

I believe that despite superficial differences between names and
locations of basic types in the registry, there are actually common
intentions behind many of these types, and therefore we can actually
make progress towards a single set. It may be that there are good
reasons for naming differences in specific technology implementations,
but at the platform-independent level we can agree on a single set and
then deal with the variations in our mapping to specific
implementations. For example, we may be able to agree that the "String"
referenced in model A is the same underlying type as the "string"
referenced in model B and also the same as the "String" defined within
the structure of model C. It is my intention to negotiate with
submitters, and, if there is agreement from the submitter, to go as far
as transforming the submitted models to refer to a single version of the
basic types! As the models are abstract PIMS this can be done without
harming implementations, but will be of benefit for the clarity of the
registry, for the harmonisation process, and for further progression of
derived types in submitted models.

For the datatypes under Recorded level I have added documentation with
the aim of defining the type in a platform-independent way. I have
generally followed the XML Schema basic type definitions, but avoided an
explicit dependency on XML Schema since the registry types are abstract
and may be used in non-XML implementations. So the registry datatypes
are not actually identical with the XML Schema data types as they are
more abstract, but they are compatible with the XML Schema types, i.e.
the XML Schema types would be an obvious, direct and suitable PSM
mapping. I am reassured that these new descriptions are also valid
descriptions of the types as used in submitted models *not* derived from
XML Schema.

I request that the Working Group as Steward considers the datatypes
under the Recorded status package for progression to Qualified or
Preferred status. The decision for each datatype can be made
independently. I recommend that these types become Preferred, but as
Submitter my vote doesn't count - the WG should make the decision.

I'm looking forward to hearing your views.

Regards

Ian Cornwell

-------------------
Mott MacDonald Ltd.
1 Atlantic Quay
Broomielaw
Glasgow G2 8JB

Direct tel:  +44 (0)141 222 4576

This message is from Mott MacDonald Limited, registered in England
number 1243967. Registered office: St Anne House, 20-26 Wellesley Road,
Croydon, Surrey, CR9 2UL, England.