Kevin Hawkins wrote:
> Some metadata schemes (like VRA Core and XOBIS) and ontologies (like 
> FRBR, CIDOC-CRM, and FRBRoo) provide methods of describing various 
> levels of abstraction, but they are too generalized -- or specific to 
> library cataloging -- to be useful for this purpose.

We (the NZETC) use an ontology modelled in CIDOC-CRM and realised as an 
ISO topic map. We make significant use of the separation between 
abstract work and text and all or our texts are downloadable in multiple 
formats, for a third level (or two levels of abstraction, depending on 
how you count). But I suspect that the pure-digital level is not of 
interest to you? I think the problem is that everyone cares about 
different levels of abstraction, after the abstract work and concrete item.

A good example of an abstract work is the The Treaty of Waitangi / Te 
Tiriti o Waitangi: http://www.nzetc.org/tm/scholarly/name-122436.html 
Most of the texts are part of compilations of one kind or another, some 
are also translations.

All this information is harvested from a combination of TEI + authority 
information, which is used here mainly as a source of identify, because 
for example there is also a (locally) famous work with the same name 
with the original work as the topic: 
http://www.nzetc.org/tm/scholarly/name-413019.html

If anyone wants a copy of our ontology (core topic map containing only 
abstract concepts and relationships), I'm happy to send out copies, but 
I seem to recall that I can't post attachments to the list.

 > How goofy is this approach?  Is this how others handle the same
 > problem?

  We use: http://www.nzetc.org/tei-source/BIM215Trea.xml

<sourceDesc xml:id="sourceDesc-0001">
   <biblFull>
      <titleStmt>
          <title><name key="name-122436" type="work">The Treaty
             of Waitangi [printed in 1845]</name></title>
      </titleStmt>
      ....
   </biblFull>
</sourceDesc>

 >  Or does everyone else really only care about describing one level
 >  of abstraction and is content with referencing other levels
 > through time-tested means like uniform titles (in cataloging),
 > canonical reference schemes (like Genesis 7:4), and the ed= attribute
 > without providing much metadata on any of them?

I can imagine <name key="bibleverse-Gen7:4">Genesis 7:4</name> working. 
I don't recall ever seeing the ed= attribute in use.

cheers
stuart
-- 
Stuart Yeates
http://www.nzetc.org/       New Zealand Electronic Text Centre
http://researcharchive.vuw.ac.nz/     Institutional Repository