FW: [osis-core] osisCore_Candiate_1.1_003 - 12 - splitID

[Steve DeRose]

At 12:53 PM -0400 08/26/02, Patrick Durusau wrote:
>Todd,
>
>Thanks for the repost!
>
>Taking your post into account, then the syntax should read:
>
>refsys[work]:canonical@grain-canonical@grain
>
>with a separate splitID, which should be xs:string, that users could 
>use either to indicate order (your case 2) or portion of segmented 
>element (your case 1). If it is xs:string, doesn't that get you 
>numbers and letters?
>

I was thinking in terms of Todd's case 2:

>>Case 2:
>><div>
>><p>
>>	<verse osisID="X" splitID=" 1"/>
>></p>
>><p>
>>	<verse osisID="X" splitID=" 2"/>
>></p>
>></div>
>>

If I absorbed the hard case right, the key issue is that there are 
things (like <q>) that can need to be split, and that don't even 
*have* an osisID -- thus having it just be an integer doesn't do it.

To which I reply: Oops. I wasn't thinking of that at all; just plain missed it.

The "quick hack" portion of my brain suggests that we should indeed 
allow arbitrary strings for splitIDs; and specify that the 
combination of osisID and splitID must be uniqute; that would allow 
the simplicity/non-redundancy in the splite-verse cases (you could 
just use numbers or letters), but require longer strings in the 
non-osisID cases (since the osisID there doesn't contribute to 
uniqueness much, being nil).

>>
>>I think it would be better to have:
>>A) A splitID that has a string value so that I would have used some
>>thing like splitID="Q-X.1-001" for the quote identifier, but DOES NOT
>>IMPLY ANY SORT OF ORDERING, with each element that is a part of the
>>split having the same splitID AND having NOTHING to do with the osisID.
>>OR B) (A) but also having a splitSeq value that is a positive integer that
>>specifies sequence.
>>C) Have a unique identifier that is the same for all of the element in
>>the split AND is independent of the osisID AND is postpended with a
>>value that indicates the sequence within the split.
>>D) Use the current splitID as positive integer which is independent of
>>the osisID and has not indication of sequence.
>>E) Use the current splitID as appositive integer with is dependant on
>>the osisID and NOT handle splits for elements that don't have an osisID.
>>
>>I think options A and B are good options.
>>

I guess the ordering isn't especially important; it seemed kind of 
nice, but I'm not all that sure it matters. I could go with (A) 
except that it seems kind of redundant to not make use of the osisID 
when it's there. You'd end up with a lot of <v osisID="Matt.1.1" 
splitID="Matt.1.1.a"> kind of stuff, which might annoy users quite a 
lot (plus being non-normalized).

Of course, since I'm still catching up maybe y'all have solved this 
already a few mails down the queue....

Gotta run off for an errand, back soon....

-- 

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