Re: CSS 2: priorities in cascading order

to follow up on what Hakon Lie said:
[concerning interaction between a new explicit weight stronger
than !important]

> ... Also, we will need a rule which reservers the new keyword
> for users only.

ASG:: There is one scenario which I would hope we can make work
where "reserved for users only" gives the wrong answer.  This is
the case where the accessibility stylesheet is associated with
the document by an active proxy.  This scenario is why I keep on
talking about being sensitive to the _order_ in which stylsheets
get associated with the document, e.g.
	author, then
	site, then
	search/index service, then
	user-selected proxy, then
	user input to browser (via stylesheets or settings).

One way to ensure user control is to say that only users can
use the new explicit weight.

There is an alternative to say that within the layer of rules
with the new stronger explicit weight that the sorting on the
above axis follows the reverse order from the order followed at
the lower explicit weights.  This is the "tilt" I have been
referring to.  It results in a potato-chip twisted priority
surface which tilts one way at one edge and the other way at the
other edge.

This amounts to "The author gets to speak first, but the reader
gets to speak last."  That is about what we need.

Jason:>  > 
>  > I would favour a key word such as "!required" rather than
>  > "!accessibility".

> No, in many cases the UA will have to approximate values specified in
> CSS declarations. The word "required" implies that if the demand can't
> be met, something drastic will happen -- e.g. the page will not be
> shown. This, in my mind, is incompatible with improving access to
> information.

ASG::  If you can't meet a requirement, you don't necessarily throw up
your hands and stop work.  You go back to whoever laid the requirement
on you to discuss what should be done.  If it it understood that recovery
from inability to meet a "!required" clause involves a dialog where
the user has to make or confirm decisions, this implication might not
be too severe to be what we want.

However, the idea of "approximating" what the style rules say returns
us to the fact that the style language presently says only constructive
things, not the reactive things that user rules want to say [Chris
Wilson point].  The user rules want to say things like "ensure that
one Em is not less than 17 pixels," not "use 17-point type."  With
the rules stated in this way, the User Agent has the room to select
a satisfactory implementation from a range of alternatives and still
claim full compliance with the rule as stated.

There is also a fine point of policy involved in the nomenclature
tradeoff between !accessibility and some other name.  The notion
that the user has a right to ultimate control should be
considered independently from any appeal to special needs for
accessibility.  If we agree on the principle that any user has
this right, the explicit weight to implement this right should
not be named in a fashion which suggests responding to "special"

-- Al Gilman

Received on Monday, 22 December 1997 11:15:33 UTC