Choice and RDG content model (Re: rend= not global? (was "Re: Fwd: [tei-council] DRAFT Agenda for ..."))

[Wittern Christian]

Dear Council members,

I tried to forward this to tei-council yesterday, but it never appeared 
here.  For the record I am pasting it now again as reply to this topic, 
since I think that makes the point a bit clearer. 


-- pasted message --

Dear Council members,

Daniel has send me the following in reply to my call for agenda items.  
I told him that this would require a discussion on tei-council based on 
a written proposal after which it might appear on the agenda if 
necessary.  So to open the discussion, here is the proposal.

My own reaction is:  I see the problem, have been bitten by it myself a 
couple of times.  However, <floatingText> might be a bit too big for 
most cases and potentially opens up a can of worms. On the other hand, 
this can might already be open with the existence of <floatingText>...

We do have some outstanding actions dealing with stuff from the textcrit 
module, so my suggestion for a course of action would be to consider 
this proposal in that context, if Council feels that it should be added 
as a work item at all.

All the best,

Christian

Daniel O'Donnell wrote:
> Hi Christian,
>
> I don't know if you want me to circulate this to council or not, but
> here is the issue:
>
> I would like us to reconsider the content model of at least two
> elements: choice and rdg. The issues are related.
>
>
> rdg is the easiest one: currently its description is as follows:
>
>
>  
>> element rdg
>> {
>>    att.global.attributes,
>>    att.textCritical.attributes,
>>    (
>>       text
>>     | model.gLike
>>     | model.phrase
>>     | model.inter
>>     | model.global
>>     | model.rdgPart
>>    )*
>> }
>>     
>
>
> The problem with this is illustrated by the recent discussion on tei-l
> concerning witnesses that have additional lines of poetry in them: rdg
> can't handle this case without some kind of work-around because it can't
> contain lines and paragraphs as it is currently formulated.
>
> I can give two examples, in addition to the recent discussion on tei-l
> to show that this is a relatively common case. One (in print) is the
> textual apparatus in Elliott Van Kirk Dobbie's edition of Caedmon's Hymn
> in the Anglo-Saxon poetic records (where several late witnesses are
> missing lines in some places and have additional lines elsewhere. A
> second is Edward Vanhoutte's edition of Teleurgang van 't Waterhoek,
> where he uses a collation unit of the paragraph in comparing the various
> editions of his subject text. In my dissertation I also frequently ran
> into a few cases where I was using the verse line as a collation unit in
> comparing witneses to poems--for example where witnesses had reverse
> lines.
>
> I suspect logically for reasons I'll get to in a minute that in fact
> that rdg like choice should have a content model very similar to
> floating text. Although all the examples I've given are at the chunk
> level, I can easily see a situation arise where it is the presence or
> absence of a text that is the collation unit: e.g. in various
> translations and redactions of the Consolation of Philosophy, perhaps
> were metra are missing, or perhaps in editions of an anthology of the
> Canterbury Tales. While you might argue that these examples are divs, I
> don't see a logical reason why an entire floating text might not be
> missing.
>
> Print apparatus do not handle readings larger than the phrase that well
> either, BTW: In Dobbie's edition of Caedmon's Hymn, he gives no
> indication that the text he reproduces as the reading from some of his
> witnesses represents more than a single line, even though it does. But
> this is in my view an artifact of the medium rather than a logical
> constraint on readings: in electronic form we can do better.
>
> My argument for expanding choice is slightly more logical than based on
> use cases, though there have been some discussions of places where using
> choice to group elements larger than the phrase would also be useful.
> Basically, we started on choice as a way of dealing with the janus tags
> (expan/abbr, etc.). This is reflected in our current definition of the
> tag as something that "groups a number of alternative encodings for the
> same point in a text."
>
> During our initial discussion of the element in the lead up to proposing
> it to council, however, it became clear that the element was actually
> more than an expression of alternate encodings, but rather an expression
> of the existence of alternatives (of content or encoding) for a given
> point in the textual stream. It was this that led to our additional
> comment: "For a specialized version of <choice> for encoding multiple
> witnesses of a single work, see section 15.1 The Apparatus Entry,
> Readings, and Witnesses."
>
> In fact, there is no logical reason why alternative content, views, or
> encoding can't exist at any level of a text. I might have different
> forwards for different audiences. I might have different texts for
> different genders (e.g. I have a book--which I can't find on the shelf
> right now--that comes in Male and Female Editions). I might have summary
> paragraphs and detailed paragraphs.
>
> Finally, I'd like to suggest there is an additional reason why choice
> and rdg should be able to contain a much larger content model: they are
> not really part of a document hierarchy at all: they express logical
> relationships rather than hierarchical. Choice (and app, its more
> specialised cousin) acquires its meaning from the fact that it groups
> children, not from the fact that it has a parent. It is logically
> possible to group the largest elements of any text, and textual
> apparatus in the TEI definition do not need to be attached to part of
> any larger text (you can encode an entire series of witnesses as similar
> a series of collated rdgs rather than as an adjunct to a running
> critical text).
>
> Anyway, in both cases, I believe the content model should be similar to
> that used for floating text.
>
>
>  
>> {
>>    att.global.attributes,
>>    att.declaring.attributes,
>>    att.typed.attributes,
>>    (
>>       model.global*,
>>       ( front, model.global* )?,
>>       ( body | group ),
>>       model.global*,
>>       ( back, model.global* )?
>>    )
>> }
>>     
>
> -dan
>
>   


-- 
Christian Wittern Institute for Research in Humanities, Kyoto University
47 Higashiogura-cho, Kitashirakawa, Sakyo-ku, Kyoto 606-8265, JAPAN

-- end pasted message --




Daniel O'Donnell wrote:
> Since we are raising this kind of thing, let me ask one too:
>
> Should we not seriously consider changing the content model of choice
> and of rdg (which is a related issue) to something similar to that for
> floating text? Currently both can contain only phrase level and global
> elements.
>
> There are two reasons for this:
>
> 1) choice is a logical relationship rather than a natural part of the
> document hierarchy. That is to say, the function that choice
> performs--grouping alternate or encodings (I would say alternate or
> simultaneous text orencodings)--is not something that logically must
> occur at a given point in a document hierarchy. I can have paragraphs or
> even divs that are alternates for each other (e.g. prose and verse
> translations of Boethian metra spring to mind), there is no reason why I
> could not have alternates for entire texts (the Dictionary of the
> Khazars [sp?] comes in male and female versions).
>
> 2) rdg is a child of app which is a specialised version of choice (as we
> say in the element definition. In a recent discussion on tei-l, we had a
> question about encoding variants where one witness has an extra line (or
> is missing one). This is quite a common case, even though our current
> definition of rdg does not allow you to indicate that one or more of the
> collated rdg's consists of more than one line.
>
> Some use cases: Edward van Houtte's Electronic Teleurgang van het
> Waterhoek collates witnesses on a paragraph by paragraph basis. My
> Caedmon's Hymn has some witnesses that are missing lines at some points
> and have extra lines at others. In the Anglo-Saxon poetic records, some
> variants involve variation in the order of lines.
>
> This is not something that print apparatus have handled very well in the
> past--Dobbie's edition of Caedmon's Hymn just prints the additional
> lines but doesn't indicate that the variants involve the addition of a
> metrical line. But we can do better. The information is useful and ought
> to be encoded.
>
> What say all ye merry councilors?
>
> -dan
>
> On Thu, 2007-06-14 at 15:28 +0100, Sebastian Rahtz wrote:
>   
>> James Cummings wrote:
>>     
>>> If there was a good way to have an
>>> att.global-for-everything-under-text which elements only got when they were
>>> used as a descendant of <text>, then that would be fine.  (i.e. <p> in the
>>> header didn't have @rend but <p> inside a <div> inside <text> did.)
>>>       
>> er, thanks but no thanks. implementing that is a project
>> for the future....
>>
>>
>>
>>     


-- 
 Christian Wittern 
 Institute for Research in Humanities, Kyoto University
 47 Higashiogura-cho, Kitashirakawa, Sakyo-ku, Kyoto 606-8265, JAPAN