Hypertext links and formatting

basic problem

mismatch between processing phases

LInks are made to documents before presentation

Link presentations are made afterwards

presentations are arbitrarily complex transformations that insert, delete and rearrange data

link anchors are not necessarily hierarchical

The transformation models we've examined so far don't really deal with that

XML and its ilk are also inherently hierarchical

we may need different processing and representation models...

Some possible data models

Stick with hierarchy, and make it easy to push overlapping elements into lower, hierarchically acceptable places

adopt a model that looks like overlapping spans -- think multicolored highlighters

use a hybrid of these approaches

hack around the problem, by using things like empty tags as delimiters for formats that aren't hierarchical

Processing model implications

Data dependencies must be tracked

This requires similar techniques to those used by Lilac

In presenting the output of a transformation, the source of the data must be known to know if it's a link

It is also possible to represent links to output documents directly

advantages

simplicity

allows linking to generated text, or other presentation-derived aspects of a text

disadvantage

Links can't easily persist across changes to the source document

links won't apply to other presentations (without a 2-way tracking operation, similar to the one we're trying to avoid

Structure and processing model interactions

The processing models we've seen are basically tree-rewriting systems (as we saw at the beginning)

These models won't apply to non-tree-structured items

Event-based processing models may be simpler

HIerarchies can map to sets of boundaries, but explicit boundaries may be more expressive