Currently openEHR refers to terminologies using a human
readable identifier, e.g. "snomed_ct", which has a version
included, e.g. "snomed_ct(2002)". The first problem is that there
is no standard for these names. There are Oids defined, but using
these in archetypes and real data makes them unreadable without an
Oid-dereferencing service. The second problem is that versions
seem to be needed because we cannot trust terminologies to stick
by the rules of never re-using a term code; ICD-9-CM is known to
be a problem in this regard, although it is being fixed.
Harold Solbrig suggests to keep human readable names inside a
known scope (e.g. openEHR) and define a mapping table from them to
the relevant Oids. He also supports the use of URI style ids for
Stan Huff thinks that the versions could also be dropped from ids,
since we can trust 'most' terminologies, and we can define
distinct oids for distinct versions of ill-behaved terminologies
as if they were separate terminologies (but then there is no easy
way to tell that such terminologies are closely related).
DK: I think I'm with Harold, but not with Stan on this! Too much
experience of version release errors with Read Codes.