[osis-core] reference systems: my thoughts

[Troy A. Griffitts]

Harry,
	I needed a nice summary.  Your thoughts and later proposal are coherent, 
intuitive and have my vote.  Hopefully we can get 3 more.


Harry Plantinga wrote:
>>3. Cf., Todd's latest post on reference systems. I personally 
>>like the 
>>current proposal but realize we need something that is largely 
>>transparent for 95%+ of the users, at least in this release. 
>>I will try 
>>to post something later today outlining (without a lot of 
>>quoted texts 
>>from other posts) where I think we are on this issue.
> 
> 
> OK. I hesitate to post, for fear I'll muddy the water rather than
> clear it, but since there are two different systems in the table
> I will post my thinking on reference systems.
> 
> --------------------------------------------------------------
> First, a general commetn. When one are only thinking about the bible,
> one tends to come up with a solution that is optimal only for the bible.
> But if we also want to encode other works, let's think about the more
> general problem.  Say you've got 100s or 1000s of books to encode, most
> of which will be different, but there will be a few instances of two or
> more editions of the same work, e.g. augustine_confessions.
> 
> 
> Principles.
> 
> 1.  A reference system doesn't exist by itself; it's only a reference
> system by virtue of being used in some particular edition of some book.
> (We've been talking about biblical reference systems by themselves, but
> even in this case I think it's true - we talk about the KJV reference
> system, for example.)
> 
> 2.  When we refer to a work, we may wish to refer in general to all
> editions of the work (Thomas a Kempis' The Imitation of Christ) or to a
> specific edition (New York: Mentor Books, 1957).  
> 
> Some examples of the way we typically make such references: (a) The
> Imitation of Christ, book II, chapter 20. (b) Calvin's Institutes,
> III.xx.4. (c) Berkouwer's Man: the Image of God (Eerdmans, 1962), p.
> 170.  But we don't want to have to write all that stuff for each of
> several references for a document.  Thus, a single work specification
> and some shorthand way of referring to the work. Work identifiers may be
> generic or edition specific.  But we always refer to a book or work, not
> a reference system. When we make a generic reference, it is always to
> different editions of the same work, e.g. different translations of the
> bible.
> 
> 3.  A reference to a work will identify the work (whether generic or
> edition-specific) and give the location reference within the work.  If
> the work is identified generically, there must be an understood default
> reference system, or the reference system must be identified. It doesn't
> make sense to refer to a specific location in a class of works that
> don't have compatible reference systems. (E.g. Chapter1 of any book
> written by Augustine.) 
> 
> Since reference systems are defined by their use in a particular edition
> of a work, one normally specifies a reference system by  the system by
> mentioning the edition of a work  (e.g. not just Augustine's
> Confessions, but Pusey's translation of Augustine's confessions).
> 
> ------------------------------------------------------
> How should this work for OSIS?
> 
> I propose the following.  Actually, as I understand the situation, Steve
> and I both think the group had already decided on the namespace-like
> prefix as work identifier some time ago.  The dropping of the notion of
> "reference system" except as embodied in a particular edition of a
> particular work is new.
> 
> Works are identified in a <work osisWork="conf"> element.  They may be
> identified generically (Augustine's Confessions or bible.lxx.*.en) or
> specifically (Augustine's confessions, SPCK 1912 or ISBN 12-345-67890x
> or OSIS augustine_confessions.spck1912). Later references to the work
> use namespace-like syntax: conf:X.3
> 
> Reference systems (Bible.KJV, Augustine_Confessions.spck1912) are
> defined when a work with that workID is created. Reference systems are
> identified by giving the osisID of the work that embodies it, e.g.
> Bible.KJV. So Bible.KJV is really an external OSIS work identifier,
> which implements a reference system.
> 
> Works can only use the "." in the OSIS work identifier when a
> translation has been defined between the reference systems.  Thus,
> someone will create a "Bible" document that embodies the OSIS default
> bible reference system.  Bible.KJV will be another document, and it must
> contain the Bible.KJV reference system as well as ties back to the Bible
> reference system which would allow transformation between the reference
> systems.
> 
> If one wishes to make a generic reference to augustine's confessions,
> and even if there are two different common reference systems, say
> spck1912 (default) and eerdmans1985, one still doesn't need to say which
> system is in use.  The <work> element may connect osisWork="conf" to
> augustine_confessions or augustine_confessions.spck1912, but in any
> case, the reference system to use will be clear and the transformation
> available if needed.
> - if conf is tied to augustine_confessions.spck1912, then conf:X.3 will
> use the reference system of that osisWork.
> - if conf is tied to augustine_confessions, then conf:X.3 will use the
> reference system of that osisWork, which would be the default edition of
> augustine's confessions.
> 
> In 99% of the cases, no mention need be made of the reference system.
> That is only necessary when one wants to "cross over" reference systems,
> i.e. refer to one edition in the reference system of another. This would
> be done by specifying the osisWork ID of the document embodying the
> reference system.  So, 
> 
> <work workID="conf">
>   <title>Augustine's Confessions</title>
>   <identifier
> type="OSIS">augustine_confessions.eerdmans1985</identifier>
>   <refSystem>augustine_confessions.speck1912</refSystem>
> </work>
> 
> Thus in this proposal we would drop the <refSystems> and <refSystem>
> elements. By this proposal we would not be able to
> augustine_confessions.xyzzy as a reference system unless there is an
> OSIS version of that document. It seems to me that it doesn't make sense
> to use a reference system unless it is defined as an OSIS document.
> Otherwise people may use different syntax, etc.
> 
> What about Todd's objection that people aren't going to want to
> understand the whole reference system thing for simple bible references?
> 
> - in this proposal you only have to mention reference systems when you
> refer to one edition in the reference system of another, so most use
> will never have to deal with reference systems separately.
> - if you want to refer to the bible in the default reference system,
> just say Bible:Matt.1.1. Or make Bible: the default reference system. 
> 
> ------------------------------------
> Specific syntax proposals to follow.
> ------------------------------------
> 
> -Harry
>