Re: XHTML Considered Harmful

On Mon, 25 Jun 2001, Arjun Ray wrote:
> [snip replies to some of my comments which I have since retracted
> because I realised I was wrong]
> [...]
>> In a non-validating UA, this means you can pass well-formed junk
>> to the UA and it will render as per the CSS rules.
> What rules?  What if there aren't any rules?

I meant the rules as in the spec itself, not as in CSS rules as in
selectors and declarations.

In the absence of ANY rulesets, the CSS specification results in an XML
document being rendered as one long line of text wrapped at the viewport
boundary. You can see what I mean by passing any long, stylesheet-less XML
file to Netscape 6.1 PR1.

> What are the "rules" for text content found in elements *defined* to
> have element content only?

(Assuming you mean rules as in the guidelines in the spec:) CSS doesn't
care either way, so long as you have a DOM you can render it.

XML 1.0 brought us the idea of official non-valid-but-well-formed
documents; in this world vocabularies can't give rules for what elements
go where. I can only hope we leave this world soon (maybe by moving to UAs
that, in the presence of Schema declarations, always validate?).

>> The rendering rules are well defined (by the styling language).

Incidentally, this is an important, if maybe subtle, step forward. We
finally are in a world where given a specific set of inputs, one can state
categorically whether a particular output is correct or not. With tag soup
and "quirks mode renderings", it was never clear exactly what should
happen. By moving into a world where, modulo browser bugs, one can
actually write to the specs and get what one expects, we are going to
improve human electronic communications no end. Of course, we are far, far
away from actually living in this world -- XHTML (as text/xml) is very
much a bleeding-edge technology right now, and probably will remain so for
a good 5 to 10 years.

>> The processing rules are well defined (if it is invalid, then
>> semantics are void).
> You did examine the two examples I gave?  (a <foo> child of <head> and
> a <foo> child of <ul>)

Yep. The exact results depend (assuming CSS) on the stylesheets present in
the cascade.

>> What more do you want?
> I don't want anything.  I'm just aware [of what (?)] non-geeks are
> likely to expect, and therefore what willing vendors will eventually
> provide.

What do you think non-geeks are likely to expect? (This is not a
rhetorical question -- I am genuinely curious as to your answer.)

Ian Hickson                                            )\     _. - ._.)   fL
Invited Expert, CSS Working Group                     /. `- '  (  `--'
The views expressed in this message are strictly      `- , ) -  > ) \
personal and not those of Netscape or Mozilla. ________ (.' \) (.' -' ______

Received on Monday, 25 June 2001 02:21:12 UTC