[osis-core] Proposed works syntax

[Patrick Durusau]

Harry,

I am working on a somewhat longer post to try to summarize some of these 
issues but wanted to be sure I understood what you are proposing.

Harry Plantinga wrote:

>>Seriously, the purpose of <work> is to allow an osisWork attribute to 
>>reference a value, much like castList allows a reference to 
>>particular role.
>>
>>What I was trying to say was:
>>
>><verse work="bible.lxx" osisID="some reference in LXX">blah, blah</verse>
>>
>
>I thought we were considering a collapsing that to a single attribute, 
>with an optional namespace-like prefix for teh work identifier, like this:
>
><verse osisID="bible.lxx:someReference"> ??
>
I think it is doable, but note that while we can specify values for the 
work portion, validation is not a possibility if you want to use datatypes.

Reason? If you want to say, bible.lxx:, bible.nrsva: are OK plus 
Book.chapter.verse, you have to use the patterns in XML schema. That 
means that you can't call another set of values, like an enumerated list 
that a user could extend, to be part of the pattern expression.

If validation of the syntax is limited to its form (in other words, not 
content) then what you suggest (which I think is the latest suggestion) 
is certainly possible.

>
>
>I thought <isbn> <issn> <url> might be a little more straightforward
>than <identifier type="ISBN"> etc, but either is OK with me. I was 
>just concerned that people would use identifier types inconsistently.
>Maybe we should specify a few and use the x-extension syntax.
>
Would not be a problem to put together. Will take a look around and 
suggested types are welcome.

>
>>So what I am trying to do is provide a way to specify what is meant by 
>>"bible.lxx" when it appears as a attribute in <work>. Would this not 
>>give us a way in the initial release to specify several bible.*** 
>>systems of reference? Or any others for that matter? (Extending your 
>>suggestion and may be way off base so please correct.)
>>
>
>What I'm not sure about is how you would use the above system to say:
>
>"I want an american english version of the KJV" -- where do you say
>that it shoudl be a generic "bible"?
>
In the header there is a language element is where I would specify 
language (as well as an optional lang attribute on elements).

>
>"I want any edition of the bible.lxx" where bible.lxx is one of those
>standardized osisID schemes we've discussed. Do I have to know a title,
>publisher, author to fill in?  Or is there somewhere that I can use 
>a generic osisID scheme name such as bible.lxx, which is defined outside
>this document?
>
>What were the other fancy kinds of bible version specifications from
>Rome that I recall someone enumerating on this list some time ago. How
>woudl they work?  E.g. suppose I want a bible (any bible) in American 
>English. I suppose I would use <language>en-us</language> but where 
>to I say "a bible, any bible"?
>
Are you saying that I should be able to say (not in syntax but hopefully 
clear)

osisRef - give me Matt.1.1 but only from any English translation? or 
from a particular translation?

Wouldn't that be defined in the header (with some improvements) as I 
have suggested? In other words, if you have in the header:

<work osisWork="bible.nrsva">etc.

osisRef="Matt.1.1" (assuming work defaults from an attribute on 
osisText) mean any Matt.1.1 that is using bible.nrsva as a reference system.

If you wanted to specify that your references are to a particular edition:

<work osisWork="bible.nrsva.niv">etc.

osisRef="Matt.1.1" is defaulting to pointing only at an NIV translation, 
although the software could fall back to some other NRSVA edition

Or even to a particular language:

<work osisWork="bible.nrsva.niv.spanish:>etc.

I think the problem is and one of the reason I suggested splitting up 
most of this into attributes is what happens if someone wants to skip 
part of the expression? I don't care about NSRVA or NIV but want 
bible.spanish as my osisWork. Not unreasonable but virtually impossible 
to validate with datatyping.

Now, if we abandon datatyping altogether, user has an example syntax and 
if they follow it, other OSIS applications will be able to find their 
materials, don't follow and it won't,   then that is an entirely 
different story.

Maybe the problem is that I keep seeing it as a datatyping problem.

>
>I think we need some way of tying a <work> to a pre-defined osisID 
>scheme. That was the purpose of my proposed additional attribute to 
>specify an externally defined osisID scheme such as bible.lxx or 
>josephus or augustine_confessions.pusey
>
No problem with referencing an outside scheme.

 From an earlier post you made today:

<work refWork="LXX"><osisWork>bible.lxx</osisWork></work> would give us a
way of saying "I want to use the osisID scheme of bible.lxx" without
identifying a title, author, publisher, etc.


I think what is also unclear to me is why user's should be able to or 
able to avoid doing author/title/publisher/etc.?

Either they want to specify a reference system (some of which will be 
predefined with that sort of information) or not. If someone is going 
outside a defined set, I don't understand the reluctance to force them 
to provide some minimal information about it. (I may be mis-reading your 
post completely  so please correct if I have run off into the weeds.)

Not sure what the refWork (which I have renamed in the schema to be 
osisWork  to be consistent with our other usage) gains you in addition 
to <osisWork>bible.lxx</osisWork>

I would just say: <work osisWork="bible.lxx">, etc. and end with 
whatever (if any) other information you wanted to include.

Could even have: <work osisWork="bible.lxx"/>, with osisWorkType (the 
type of the attribute value) being one of our extensible attribute 
lists. The value in osisText of osisWork should match one of those 
values or with the "x=" extension.


I think we are getting a little closer!

Patrick

>
>>Interesting question: If we do specify reference systems using this 
>>method, do we need to pursue the grain/character pointing issue at this 
>>point? In other words, you can use osisRef to point to as fine a level 
>>as specified by a reference system or it can point to an ID, which might 
>>take you to a word for example?
>>
>>Advantage would be that we can simply say that osisWork and osisRef have 
>>to comply with a character (no spaces, commas, periods (other than as 
>>syntax separators) no @, etc.) and period delimited syntax. We could 
>>even specify a couple of common ones in advance.
>>
>>Do need to think about the canRef and osisRef where someone uses a 
>>divergent reference than needs to map back to the canonical set.
>>
>>Have you thought about the Psalms problem if we use NRSVA? I am torn 
>>because if we use BHS, then users will have unfamiliar numbering leading 
>>to errors. If we use NRSVA, then, well I could declare a bible.bhs 
>>system and say that my osisRef in the head for a psalm maps to 
>>bible.bhs.ps.v?  work="bible.bhs" osisRef="Ps.---"
>>
>
>I don't recall what the psalm problem was exactly.  Is it that some 
>psalms (e.g. 42) have a header (For the leader. To "Lilies.") that
>is not a part of any verse?
>
>That's the nice thing about making up your own set of osisIDs. Make
>one up for this part.  So psalm 42 would have
>
>Ps.42.header (or Ps.42.0)
>Ps.42.1
>Ps.42.2
>Ps.42.3...
>
>The mapping could map Ps.42.0 and Ps.42.1 to Ps.42.1 in a greek edition.
>In this regard, I'd rather have the reference version have a separate
>Ps.42.0 and Ps.42.1 -- finer granularity for mapping other versions
>into/out of this one.
>
>Of the dozen or so most common English editions (NRSV, NIV, KJV, etc etc) 
>I think they all coincide exactly on the OT and have only two minor 
>variations in the NT (3 John 15 and a verse in Rev, if I recall correctly.)
>I think we should definitely use something in this class for a standard.
>
>-Harry
> 
>

-- 
Patrick Durusau
Director of Research and Development
Society of Biblical Literature
pdurusau@emory.edu