[osis-core] osisID as List

[Steve DeRose]

At 04:44 PM -0400 08/14/02, Harry Plantinga wrote:
>It was also my understanding of the matter (and my understanding
>of Steve's position) that osisIDs are never ranges.  I think that's
>a direct quote from Steve.

That's how I've been looking at it; though I've been know to be 
proven wrong.... from time to time (as C3PO said).

>
>However, that's a slightly different matter from what I understand
>Todd to be proposing -- that a single entity such as a paragraph
>could have several osisIDs. 
>
>Take the Matt.1.6-Matt.1.11 case. If a single element corresponds
>to several osisIDs, how would you mark that up? 
><p osisID="Matt.1.6">
><p osisID="Matt.1.7">
><p osisID="Matt.1.8">
>...
>[text of paragraph]
>...
></p>
></p>
></p> ?
>
>However, I'm not sure this should really occur. At least in this
>example, I would suggest that the CEV is using a different versification
>scheme, and it would use some other ID to identify that paragraph,
>perhaps Matt.1.6_11 (which is NOT a range, just a single osisID).
>A translation would be defined between the two schemes.

There seem to be two basic options: either to treat Matt.1.6_11 (or 
some syntactic variant) as a primitive canonical reference in an odd 
v. scheme, or to  treat it as shorthand for a list of canonical 
references in a (more-or-less) normal v. scheme.

I tend to look at it as the latter, because it uses the same syntax 
as people would use for the latter (say, when referring to that same 
range of verses from a commentary), and because it has the right 
compositional semantics -- for example, you would create an 
identifier like Matt.1.11-Z or Matt.1.11-3 or Matt.1.zorch-foo, and 
the reason you wouldn't appears to me to be that they don't decompose 
properly into the more primitive/basic v. scheme.

Another reason I favor listing the applicable IDs, is that you can 
then do a relatively brain-dead match to find a given verse. Or if 
evaluating a reference to a a range, you only have to expand the 
range, not do a double-ended range intersection (the algorithm for 
which is not terribly hard, but also a but much for the "desperate 
Perl hacker" of XML fame).

Oh, I just realized that if we're allowing things like Matt.1.11.b in 
IDs, then we need a slightly more sophisticated match than XPath 
typically does anyway, so it doesn't seem we would by a lot by the 
bracketing to ensure token boundaries.

>
>But could there ever be cases in which a single elemetn (e.g. paragraph)
>corresponds to two or more osisIDs in the chosen reference scheme?
>I'm not sure. Perhaps. So maybe we SHOULD allow a list of osisIDs.
>

-- 

Steve DeRose -- http://www.stg.brown.edu/~sjd
Chair, Bible Technologies Group -- http://www.bibletechnologies.net
Email: sderose@speakeasy.net
Backup email: sjd@stg.brown.edu