W3C home > Mailing lists > Public > www-style@w3.org > October 2003

Re: [CSS21] Comments on the 2003-09-15 CSS 2.1 Draft

From: Henri Sivonen <hsivonen@iki.fi>
Date: Tue, 21 Oct 2003 16:53:48 +0300
Cc: Boris Zbarsky <bzbarsky@mit.edu>, David Woolley <david@djwhome.demon.co.uk>, <www-style@w3.org>
To: Tantek Çelik <tantek@cs.stanford.edu>
Message-Id: <FEE4ADB2-03CD-11D8-ABD7-003065B8CF0E@iki.fi>

On Tuesday, Oct 21, 2003, at 10:23 Europe/Helsinki, Tantek Çelik wrote:

> On 10/20/03 11:23 PM, "Boris Zbarsky" <bzbarsky@MIT.EDU> wrote:
>
>> David Woolley wrote:
>>
>>> It is not really possible to implement CSS with a true tag soup 
>>> browser,
>>> as CSS requires a well defined parse tree.  (On that basis, I believe
>>> that NS4 is a tag soup browser with CSS bolted on and IE is a 
>>> structured
>>> document browser with extensive error recovery logic
>>
>> See http://ln.hixie.ch/?start=1037910467&count=1 (for Win IE)
>
> Given that tag soup fundamentally doesn't reflect a tree structure, it
> should come as no surprise that different teams of programmers would 
> come up
> with different approximations.

It's no surprise that different teams came up with different 
approximations. I think the surprising thing is that the teams that 
developed Trident and Tasman chose to accept the additional complexity 
that followed from allowing a non-tree internal representation.

> Why bother trying to make CSS and DOM work properly (or predictably or 
> in a
> "well-defined" manner for that matter) with tag soup, when that would 
> simply
> encourage authors to depend more on tag soup?

Because the engine itself might be more maintainable and debuggable if 
it works predictably and there can be code that walks the DOM but is 
neither part of the engine nor supplied by the tag soup author. A 
Mozilla extension can rely on the DOM tree being a tree, so it is safe 
for an extension to walk the DOM of an arbitrary Web page using an 
algorithm that needs a tree in order to terminate.

> In fact, with subsequent versions of Tasman
> (IE5/Mac...IE5.1...IE5.2...MSN/Mac), in some cases we *dropped* some 
> tag
> soup support whenever it got in the way of fixing some sort of 
> standards
> support.

Cool.

> After all, isn't it better to wean web authors off of tag soup rather 
> than
> continue to encourage them to depend on it by working too hard to make 
> it
> work predictably?

Yes.

-- 
Henri Sivonen
hsivonen@iki.fi
http://www.iki.fi/hsivonen/
Received on Tuesday, 21 October 2003 09:55:17 GMT

This archive was generated by hypermail 2.2.0+W3C-0.50 : Monday, 27 April 2009 13:54:24 GMT