Re: Error Parsing

On Saturday 2004-10-23 17:58 +0000, Ian Hickson wrote:
> Right -- it's definitely intended for CSS to be parsable with no more than 
> one character look-ahead, so anything that implies otherwise is most 
> probably incorrect.

That's not really true -- it requires one-token lookahead, not
one-character lookahead.  (With the namespaces draft, a case that
requires that is distinguishing the selectors:

[lang|foo]   /* an element with the lang:foo attribute */
[lang|=foo]  /* an element whose lang attribute starts with the
                dash-separated word foo */

However, currently comments are themselves tokens.  So I guess, in the
spec as currently written, an unterminated comment is not a comment, but
I don't think that's the right answer.  I think the tokenization rules
should be clearer that it's an untokenizable stream, which should in
turn require that the tokenizer use the largest tokenizable initial
segment of that stream.

-David

-- 
L. David Baron                                <URL: http://dbaron.org/ >