style processing and multi-level semantics

In the XML Plenary discussion of namespaces and modularity, there is much
hope being attached to a new "packaging" working group to get us out of the
woods as far as implementing both loose and precise references to documents
which contribute to the definition of markup.

It was proposed that having a [content] document reference its styling
document was backwards, and someone suggested that if we just had a hub
document that would invoke both abstract content and style rules then we
would be OK.  Here is one of my posts provoked by that proposal.

What I think is important to capture here is the color scenario and the
"naming the classes" scenario as persistent topics that we need to get
written up in the central reference note or WCAT and be able to come back
to them time and again as we apply these concepts to multiple brushes with
the people defining language variants based on XML and/or derived from HTML.

I have laundered this for public disclosure and copies ER IG because one of
the central concepts is a styling adjustment compiler.

Not that we are going to build one tomorrow, but the logic of the format
architecture should not make such a tool impossible.

Al

-- quote begins

This point illustrates why I think we need to include a processing-phases
view in our analysis of what is needed.  

A hub document on the server cannot know the full final styling for the
document.  That is decided in the client with due consideration for
nominations from the author.  Or at least, there is a "author proposes,
user disposes" principle underlying the cascading rules in CSS2 which is
the result of an agreement with the WAI and which logically should be
replicated for the style processing of all Web documents to maximise the
accessibility of web content to users with disabilities.

While the user does want to have the final say in the application of
display properties to rhetorical classes of text, it is also in the
accessibility interest that both the display phenomena and the rhetorical
classes be identified with widely-used and thoroughly defined names, to a
greater extent than is practiced now or formalized in HTML4.

For a person with color weakness, it is relevant to ask what the colors are
that are used in styling before deciding to override the author's styles to
apply non-color distinctions instead.

By the way, this is an example of the case for multi-level semantics.  The
author styles probably only care about color as a point space with a
discrete topology: that the same color code with generate the same color.
The assistive technology that repairs the styling of a document cares about
color as a metric space: it cares how likely two colors are to be confused
with one another.  The usual processing does not need to access the
distance metric information, but the color namespace needs to
deterministically link to it.

Processing the document in accordance with the author's stylesheet needs
only the simple definition of color as a defined way of notating color
values.  Repairing the styling to make it accessible with the minimum
disturbance requires a deeper definition of the color space as a volume
with proximity properties: some pairs of colors are farther apart than
other pairs of colors, and that matters.

Similarly, the fact that the classification used as style rule
discriminants follows the rules in "The Chicago Manual of Style" is
something that can be ignored by most instances of style processing for the
document, but authors of special-needs stylesheets need to know this

information, and namespaces offer a likely candidate for binding to this
schema of categories.


Multi-layered definitions, if we can integrate them in a way so that "what
is easy is easy," do offer benefits in the area of adaptivity of the
documents to client-side conditions including particularly variation in
human performance.

Al

Received on Monday, 4 October 1999 14:37:08 UTC