Re: Core Component Analysis - SWIFT's Comments

[Martin Bryan <mtbryan@sgml.u-net.com>]

Duane/David

>GUID is intended as internal registry linkage use within registry engines.

>The UID is for human use.   UID has several suggested forms, including

>URN:UID, UID by itself, or URI:UID, depending on need.

Before you get carried away, let me remind you of another important factor,
which has come up in another message. Whilst randomly generated numeric
identifiers are suitable for machine interpretation of the relationship of a
DTD to a registry they are not suitable for human interpretation of a
registry. For this you need something that is both "readable" and
timestamped. A registry has to record all the definitions of an element over
time, not just the current one. Therefore I need to be able to refer to the
Jan 2000 definition of TaxableAmount and compare it with the Jan 2002
definition. I need to be able to distinguish the US definition of
TaxableAmount from the European version of the same thing because they may
require different properties to be associated with them. (For your
information the EU has just issued a statement stating that all taxable
amounts to which tax exemptions apply must contain a reference to the
regulation that defined the exemption, which in turn must be timestamped.)

If people are to be able to distinguish the contexts and times for which IDs
are relevant the ID should include information relating to date and context.
They must also be able to find all IDs that are associated with a particular
model/datatype. This can be done if you have names such as
TaxableAmout-US-20000101 and TaxableAmount-EU-20020202, but not if you have
randomly generated numeric identifiers.

Martin Bryan