[tei-council] reviewing SA again;what is standoff?

[Syd Bauman]

Glossary
--------
SA = SA-LinkingSegmentationAlignment.xml
   = Chapter 14, "Linking, Segmentation, and Alignment"
14.9 = Section 14.9 "Stand-off Markup"


SR> I think of "standoff" as being when some markup in document A
SR> says things about markup in document B; thus if document B has 67
SR> paragraphs and 14 images, then document A makes claims about how
SR> the two relate together.

LR> You are perfectly right, this is the way most people (at least)
LR> in the NLP community would think stand-off means.

On rereading, I don't understand what you're saying Sebastian. The
markup in document A makes claims about how the paragraphs and images
in document B relate to each other? IMHO that is one perfectly
reasonable use of stand-off markup. But you imply that the markup in
document A is limited to making claims about the markup in document B
-- why can't it make claims about the content? (Which, IMHO, is
another perfectly reasonable and in fact commonly discussed use of
stand-off markup. TEI discusses this use as a method for overlap
representation in XML, IIRC.)

But maybe I am the one who does not understand what "stand-off"
means.


JC> For the record, I tend to automatically think of link/join/span
JC> when someone talks about stand-off markup.

Reasonable. These are mechanisms for stand-off markup, at least as I
understand reality.


JC> So I think of it not so much the "tags representing elements" in
JC> another location as 'additional interrelations between existing
JC> elements' which I think of first when one mentions stand-off. I
JC> realise this distinction is probably trivial, or based on my own
JC> misunderstandings,

I'm inclined to say your first reaction, while valid, is somewhat
limited, then. But the distinction doesn't seem trivial, and I don't
think what you're describing is *not* stand-off, just that it's not
all stand-off is. See, e.g., Henry Thompson's paper on stand-off:
http://www.ltg.ed.ac.uk/~ht/sgmleu97.html.


JC> ... and discussed earlier in the same chapter. 

Well, such uses for stand-off are not described anywhere generically,
I don't think. They are discussed when particular issues are
discussed, like the use of <join> or <link>.

Very well may be a good idea to have a more generic discussion.


JC> The inclusion of virtual content with XInclude (as you suggest) to
JC> include boilerplate text or managing chapters somehow seems like
JC> something slightly different to me.

Not just slightly different. :-)


JC> (But as you say this is a more common use, so is it discussed
JC> elsewhere?) 

No, and IMHO it almost certainly should be. Lou still holds that
entity references are the way to do these things, though. (I do not
claim TEI should *not* discuss entities, only that we should discuss
XInclude, whether instead of or in addition to entities.)


JC> I don't think most people are going to use XInclude in the manner
JC> shown in 14.9.5., but in much more simplistic ways.

I agree completely. But this is true of a lot of the material in SA
(and I daresay other places). Very few people will use XPointer in
the various ways described in SA ... they'll just use bare name
identifiers and be happy. Heck, that's what I do!


SB> Stand-off is only one use-case for XInclude, and I daresay it is
SB> one of the less common ones, at that. Boilerplate text or
SB> managing chapters is probably much more common.

SR> indeed. which is why the harsh strictures on XInclude are
SR> misleading in 14.9?

I'm not sure what you mean -- the idea that validation is performed
after XInclude processing not before? That's a problem? It seems to
me to be a liberating concept, not a restricting one. And I don't see
why it should be any different based on what your goal for use of
XInclude is.

Remember that what it means is that you do not have to worry about
getting your ODD just right so that your schema permits <xi:include>
where you need it. It does not prevent you from doing this, but gives
you permission to skip this step, and just rely on your processor's
capability to do the XInclude processing before validation. (You
still have to declare the namespace, of course.)


SR> personally, I think we should drop 14.9 entirely, and
SR> remain agnostic on the subject. It gives an odd view
SR> of XInclude, and talks about standoff in a way which
SR> I claim is not consonant with the way the rest of the
SR> Guidelines sells it (tho not under that name).

This seems like a step backward to me, and one I would be
extraordinarily hesitant to take without serious reconsideration by
Council of not just the issue, but what we want from, and how we
interact with, our WGs. Remember that 14.9 was conceived and written
by a stand-off markup theorist (David Durand) and a stand-off markup
practitioner (Fabio Vitali), and was at least approved of, if not
thought highly of and actively encouraged by an implementer (Jessica
Hekman), two computational linguists (Nancy Ide and Jean Carletta),
and one TEI editor (me). Oh yes, and then approved by Council.