[osis-core] osisWork, refSystem - Summary

[Harry Plantinga]

> I thought the ID was arbitrary -- like a namespace *prefix*. The ID 
> is what ties actual references back to the right works and reference 
> systems, which are then defined by their full DC block (which would, 
> for example, have the 1611 as publication date, English as language, 
> etc).

That was also my understanding.  We have talked about IDs such as  
Bible.LXX.charles.en.1880, though, and we talked about using
wildcards to select one from a class, e.g. Bible.LXX.*.en.*.  We may 
have been talking about a refSystem, but some of those attributes, 
such as the language and publication date, may be related only to the 
particular edition of the work, not to the reference system.  (I would
also note that the notion of wildcards seems to be missing from the
schema.)

So, it seems we have two different sorts of external identifiers,
refSystem
identifiers and osisWork identifiers. 

Do we want to say of osisWork identifiers that

  - they must be unique for a particular work/edition/version
  - the first token is the "work", e.g. augustine's confessions. That
would 
    mean using augustine_confessions, not augustine.confessions.
  - if two share a prefix, they are different editions/versions of the
same
    work, e.g. augustine_confessions.pusey and
augustine_confessions.smith
  - wildcards and prefixes can be used in a document-specific way that
is
    not specified in OSIS 1.1.  E.g. Bible.LXX.*.en would have a
Bible-specific
    meaning, but it would match documents such as
Bible.LXX.smith.en.1990

Proposal: we could use an internal-only namespace-like prefix on osisIDs

for works, if we were to specify the external OSIS identifer in the work

element thus:

<works>
  <work workID="lxx_en">
    <title>Septuagint</title>
    <language>en</language>
    <identifier type="OSIS">Bible.LXX.*.en.*</identifier>
  </work>
</works>


....osisRef="lxx:Psalm.20"...


-Harry