Re: Some Views on CSS

Thus spake Russell Steven Shawn O'Connor:
 
[tf] As I see it, the requirements
> > for both cascadable style and readily extensible documents/sites lead
> > inexorably to the conclusion that stylesheets must be modular, with each
> > module having a fixed selector structure covering every element in the
> > target DTD, and each complete family of modules providing descriptors for
> > every CSS(1?) property.
>
> Absoultely.  I'm trying to figure out the equivlence classes of
> non-conflicting properties.  Color, background-color, background-image,
> and border-color together form an equivalence class under this relation.
> But the rest aren't so obvious.  Should margins and padding go together?
> I'd like to have the equivalence classes as small as possible.  I'm a
> little surprised because ti seems taht most properties can stand in their
> own equivalence class.

Color, background-color, background-image, and border-color form a
"natural" equivalence class by most anybody's understanding of potential
conflict, as black-on-black is just wrong. But it's simple only as long as
you assume certain other (default) conventions to be in effect throughout
the document. If one does not assume that links will be indicated by
underline, for instance, rather than by color or weight, then typographical
decoration and weight of both links and other phrasal elements need to be
reconciled with the "decorative" color scheme. And so on. The core style
project groups these things into the "affordance" module.

I think other equivalence classes are more a matter of taste and (rational)
typographical convention. I would argue, for instance, that paragraphs
should be separated either by extra vertical space or by in- (or
out-)dents, but not both - superfluous and ugly IMO. Similarly, I think
vertical space modulation on all elements needs to be brought into an
integral system if it is to convey anything coherent about document
heirarchy; similarly the weights of headers, and the treatment of
horizontal rules, and so on. All of these biases are on view in the modular
structure of the core project.

> > As for UA defaults, I think it's fair to guess at this point that any UA
> > whose defaults were to depart so dramatically from the Mosaic defaults[3]
> > as you describe might qualify as a new media type, alongside projector,
> > grid, handheld, TV, etc. Whole new stylesheet ballgame, needing its own
> > modules to accommodate cascading. For the screen media type, "Mosaic won."
>
> I completely disagree.  As Jukka Korpela said in
> comp.infosystems.www.authoring.stylesheets:
>
> <BLOCKQUOTE CITE="news:34e214dc.25115793@news.cs.hut.fi">
> This is going backwards. You seem to give one possible representation
> mentioned in the specs a very special positiona [...] That's more or less
> making HTML a (poor, naturally) physical markup language, with some vague
> options of changing the presentation if one really wants to.
> </BLOCKQUOTE>

Heh? Style is orthogonal to HTML. If style were ever to be implemented
completely, this would be completely true. Implementors have so far shown
themselves unwilling or unable to "cut the umbilical" and support the
display and float properties. I believe this is because they perceive
self-interest in an easy, cheap, inferior display language called HTML,
which, after all, TimBL apparently modelled after RTF. Why mess with a
barely good enough thing, especially when you can jazz it up with scripts?

What if somebody wanted to indicate links by setting their display property
to "block", or strong emphasis by floating phrases into the margin? Or, as
you described earlier - what about a UA that defaulted to displaying
emphasis within divisions within doubly-nested tables with BLINK? That
wouldn't cascade nicely with any stylesheet that assumed more conventional
defaults. That's what I mean about the necessity of declaring all of one's
assumptions. It's possible to make well-informed decisions only in media
where the defaults are known. Jukka would conclude that the whole concept
of authorial style is bankrupt, and I would ask him simply to turn off
authorial styles for the "screen" media type when accessing the Web through
something very unlike Mosaic. The developers of such a UA would likely have
the good sense to disable such support at the factory, however.

I've heard that work is beginning on something called, for lack of better
knowledge, the "Rendering Environment Object Model". This would presumably
make it possible to parameterize authorial style decisions based on the
rendering capabilities of any REOM-compliant UA. I hope such a model
includes extensive user profile information, too, as that's the ultimate
canvas.

__________________
Todd Fahrner
mailto:fahrner@pobox.com
http://www.verso.com/agitprop/

Received on Tuesday, 14 April 1998 12:27:53 UTC