I hear from Jon and Cathy Laws "we don't want to talk about heuristics."

I think that we need to have more discussion of this point.

Deconstruct the concept 'heuristics' into a richer model of the processing
and its specification and the life-cycle of the content properties that
are sensed in the processing.  And have a fresh discussion of how far
down the road of laying work on the UA we ought to go as opposed to
asking for high quality content from the authors.  In addition, the
techniques that seem too burdensome to deploy in Customer Premises
Equipment software AT, may still be relevant to apply within authoring
tools as drivers for hints that aid the author in producing the markup that
the client-side processors need.

I am perhaps inclined to propose solutions that are part UA-algorithm,
such as inspecting the ancestors of an element for orientation cues
before saying that the element is inadequately explained in the content
for the purpose of orienting AT users.

This sort of a processing rule, a search-path directive, could be in a grey
area that some would call a heuristic and others would not.

The state of practice suggests that currently it takes heuristics to
support users with current web content.

My star example in this regard is the command that says "seek forward
until you find more than N consecutive characters of non-link text."

This is very low-level, and hence robust and omnivorous.  It presumes
nothing about page structure other than that links are recognizable, so
text can be classified into link text and non-link text.  And it takes a
heuristic test, the length of text that is present between two adjacent
links, as a prognostic for where there is main content, or at least
allows skipping material which is a very dense sequence of links.

So one possible test for success on our proposals could be stated
as:  "What combination of easier processing in the assistive technology,
coupled with what level of uptake in the web content production
world, would mean that users would stop using the heuristic-rule
keystroke and use the more formalized methods?"


Received on Friday, 15 April 2005 16:17:40 UTC