RE: CSS discussion

 

________________________________

From: Sean Owen [mailto:srowen@google.com] 
Sent: 07 August 2007 15:03
To: Jo Rabin
Cc: Abel Rionda; public-mobileok-checker@w3.org
Subject: Re: CSS discussion

 

 

On 8/7/07, Jo Rabin <jrabin@mtld.mobi> wrote:

Hi Abel

 

I think it makes sense to do what you suggest - the location of the CSS
and where it is referenced from needs to be noted I think - per the
earlier discussion on this list (and suddenly realizes that he didn't
update the example moki document, as he was having fun being on vacation
:-( ).

 

As to a full serialization - I think I support this idea, though like
Sean I am worried about "where it will all end" and do we really need
it?

 

I think that it probably is worth while if only because we need to
figure out at least two particular cases:

 

1. Valid CSS Level 2 selectors which are not valid CSS level 1 e.g.

body > h1#potter + p.hogwarts a:visited.broomstick


We are not concerned with selectors right? how does the serialization
help this?

 

Well I think we _are_ concerned with selectors as a CSS Level 1 parser
will ignore things it doesn't understand. So if the selector syntax is
post Level 1 then we should fail.

 

I agree that serialization is not so much the issue as having a way to
process things.  

 

 

	 

	2. Values that are inappropriate to the property in question

	e.g. width: -1px

	e.g. width: italic

This is more a question of validation, which is indeed important, rather
than the two CSS tests themselves. But again what does building a tree
representation out of this do to make things easier, per se?

I don't think the tree representation is important so much as having the
information in a processable form. From the point of view of this
project processable form means pretty much in XML I think. Or at least
it is not inconsistent with the rest of the project for it to be so.

 

	It may be that we want to check and warn for other things too.
E.g. we might want to check that the selectors actually refer to valid
HTML elements. All of which says to me that it would be better if we got
on with a full serliazation.


Nothing about validation requires a DOM-like model. I think it's well
out of scope right now to check elements vs. selectors, but anyway, CSS
selectors don't actually have to be used right? I can define a foo class
that's never referenced? 

sure, but it would just be wasteful to do so. Not "in the spirit" of
minimize


This one I feel strongly about. This is creeping far out of scope -- for
version 1.0. It is all good, but I do not believe it's a good idea to
put energy into this while tests remain unwritten. This is a whole
project unto itself. 



I agree that we should think carefully before doing it. Fools step in
where angels fear to tread. But then again, we have both feet over the
threshold already, don't we?



Jo