TEI Stand-Off Markup WG 2002-09-26 Conference Call Notes

Short agenda: 1 = continue discussion on linking.

FV & JH explain (in response to SB's question) that it is left to a downstream process that knows the semantics of the document to figure out which attributes are pointers, and to know which attribute it points to.

DD explains that still need to know which attrs are IDs when using XPOinter; and that don't need to do complete parse, but could just parse ATTR decls in internal subset.

JH asks if making people declare attrs in internal subset could they use an external entity and point to it? DD answers yes. [Note: SB is not so sure.]

DD also lose IDREFS, which XPointer doesn't have, although you can have mutliple pointers in a row they are fallbacks or backups of each other, not different targets. FV disagrees, thinks a boolean disjunction returns a set of elements, SB thought it returned only the first element.

Seems feasible to replace ID/IDREF with (URI+)XPointers. Should be feasible for ID/IDREFS, too, but large & ugly.

[?...?]

DD attempts to summarize: could replace ID/IDREF with XPointer without a major problem. Need to look at how much ID/IDREFS to XPointer is confusing.

DD wonders if could use ID/IDREF for relations, and XPointer for navigational links.

FV's suggestion had been XPointer for relations, and XLink (which uses XPointer) for navigational links.

JC? asks (essentially) how we tell the difference — she was thinking we make them all XPointer and let some later user or stage decide the semantics, i.e. whether the link should be followed or not.

FV says there exist link types [target=, e.g.] that are always and only navigational.

JC asks 2-part question. [?...?]? What's advantage of XPointer?

FV encoder will not have the choice to choose, the TEI creators decide a priori which are which. There exists a standard that provides application-related semantics. We should use that.

JC on IDREF v. XPointer [?...?]

DD one remaining technical question: IDREFS. JH suggests using sub-elements as an alternative to large ugly boolean XPointer.

DD would like to divvy up attr list he posted yesterday and have people sign up for them. Suggestion that he just assign them.