Re: comments on CSS1

lilley (
Tue, 5 Dec 1995 13:31:29 +0000 (GMT)

David Seibert said:

> <h3>Warnings and notifications</h3>

> <p><strong>
> The job of the author is to provide documents that can be interpreted by
> the UAs that their readers use, and the job of the UA is to make
> these documents as clear as possible within the prescriptions given by the
> author.

Fine so far

> When the UA receives a document that it cannot render as
> prescribed, it should not warn the reader of this except as a last resort.

Hmm. I would say instead that the warnings should be configurable.

> In most cases, a warning to the reader will not help him to read the
> document, but will merely inform him that either the author or the UA has
> failed in their job.

Some people might like to know this. In particular, the rather common
case where the author and the reader are the same person; switching on
all possible warnings is then desirable.

> If a reader receives too many warnings, his most
> likely response is either to find a new UA or to stop reading documents
> from the author; this is bad for the reader, the author, and the UA
> vendor, so the use of warnings should be minimized.

Nope. The use of warnings should be configurable (at a finer level of
granularity than on/off).

If a reader chooses to stop reading documents from an author who is
not up to scratch; if a reader switches to a more capable UA; they
are perfectly entitled to do so. This is good for the overall standards
of authoring and the overall standards of Web software.

I believe the term for this type of evolutionary pressure is
"market advantage"

> A warning when the reader or UA is overriding "legal" or "important"
> declarations may be a good idea.

More than that, it may be a legal requirement ;-)

> However, a warning is only necessary
> when a font size is declared to be legal or important and the UA is going
> to use a smaller font size. Substitution of a larger font may be done

David, you are confusing common sense with legal procedure. Don't; it can
be expensive.

> When a reader turns style sheet downloading off, the UA should warn the
> reader that he may miss legal or important declarations. This should occur
> at the time that the UA is instructed not to download style sheets.


> However, if style sheet downloading is off, the UA should not be
> expected to check the (non-imported) sheets for important or legal
> declarations and warn the reader of their presence, as this can
> significantly degrade the performance of the UA.

Good call. But for this reason, any stylesheets with legal importance
are likely to be in the head of the document rather than referenced

> The reader may wish to
> be notified whenever he reads a document with an attached style sheet that
> has not been downloaded, but this is really at the reader's discretion.

I think that is the same thing as saying it should be reader configurable.
Good, so we agree then.

> When specified colors are not available, the UA should attempt to provide
> contrast so that all text is legible. However, it is not necessary to
> warn the reader that a certain color was requested but unavailable.

Not necessary in all cases. By the way, why assume the intent was
legibility? A nice bit of subliminal text, or some neon pink-on green
grunge typography might be anything but legible ... but I take your point.

> The main purpose of style sheets is to enable authors to improve the
> appearance of their work. If a reader would rather reduce the quality of
> what he sees by not downloading style sheets, so that he can read it
> faster, this should in general be allowed without warnings.

Except, as you note, an initial warning on switching off stylesheet download.

> If an author
> is concerned with this, he may warn his readers that the quality of what
> they see may be reduced if they do not use his style sheets,

How? With messages in the body text?

> <h3>Caching style sheets</h3>
> It will be important for authors who use complex style sheets that their
> readers have a way to save the style sheets, so that they do not need
> to download them every time they read a new document

I expect setting a long expiry time will solve this. Most browsers now
manage their local caches in the same way that proxy cacheing servers do,
so the same techniques of explicit expiry, expiry hueristics, conditional
GET and probabilistic prefetch still apply.

> Probably a distinct style sheet cache would be the best

I don't see the need, particularly. What about the existing cache
mechanism is unsuitable?

> The "on-off" approach to style sheets in CSS1 is also too restrictive.
> There should be at least three options for standard UAs, in order of
> text presentation speed:
> <ol>
> <li>Download (and use) all style sheets,
> <li>Use only cached style sheets (don't download new style sheets), and
> <li>Basic html (don't use any style sheets).

I take it you do not mean, display the raw HTML. You mean, render the
HTML somehow? This is, at least conceptually, using an internal stylesheet.

> <h3>Reader-author conflicts</h3>
> The reader will (and should) always have the power to override the
> wishes of the author with respect to style - it's his UA.

I agree with this position although I am surprised to see David
argue in favour of it.

> [...] Thus, issues of priority will naturally evolve in
> accordance with the draft specifications of CSS1, so that the author's
> style choices will dominate.

Could well be. We will have to see.

Chris Lilley, Technical Author and JISC representative to W3C 
|       Manchester and North HPC Training & Education Centre        |
| Computer Graphics Unit,             Email: |
| Manchester Computing Centre,        Voice: +44 161 275 6045       |
| Oxford Road, Manchester, UK.          Fax: +44 161 275 6040       |
| M13 9PL                            BioMOO: ChrisL                 |
| Timezone: UTC        URI: |