W3C home > Mailing lists > Public > www-style@w3.org > January 2008

Re: Re[16]: css with attribues

From: David Dorward <david@dorward.me.uk>
Date: Wed, 16 Jan 2008 15:44:33 +0000
Message-Id: <B45A6AC5-DDD4-445C-9BC3-5DE3C12C5D60@dorward.me.uk>
To: CSS Style <www-style@w3.org>

On 16 Jan 2008, at 15:20, Dmitry Turin wrote:
>>> (a) Other variant is to create cas-file (file with only attributes),
>>> and to put
>>> <link href="a.css" type="text/css">
>>> <link href="a.cas" type="text/cas">

> DD> Making this an entirely different language would be a more  
> reasonable
> DD> approach.

> User gets benefit (unnecessary to keep separation in brain),

Who is the user? The author or the visitor? It sounds like you are  
talking about the author.

As mentioned previously, separation of concerns is a good thing. "I  
am writing HTML, therefore I am entering content and describing its  
semantics". "I am writing CSS, therefore I am describing how the  
semantics should be represented to the user". It isn't difficult, it  
lets authors think about one thing at a time, and it lets multiple  
people contribute to a document (with a CSS expert working on the  
look while HTML experts work on the content).

If your goal is to put the presentation and semantics into the same  
place, then just go back to HTML 3.2. Otherwise you'll end up trying  
to squeeze content into CSS, and the structure of the language  
doesn't lend itself to that.

> machine gets loss (necessity to update tree by attributes from css).
> Total: benefits of user is more important, than uncomfort of machine.

I'd say that this would make life more complicated for authors and  
for machines. So both sides would have a net loss.

>>> But man will be still duty to remember,
>>> what characteristic is for CAS ('attribute' in old terms),
>>> and what characteristic is for CSS ('property' in old terms).

> DD> That's no different from today
> Yes, today is bad.
>   Possibility to put some characteristics into comman file (css- 
> file) is good,
> but order to make this - is bad: any characteristics can be not  
> only common,
> but unique for concrete set of documents (redundant 'style=').

>   Possibility to put some characteristics directly in document is  
> good,
> but order to make this - is bad: any characteristics can be not  
> only unique,
> but common for concrete set of documents (this topic in www- 
> style@w3.org).
>   Author must make line between unique and common (depending of own  
> site).
> Now some people obtrude (impose) line.

It sounds like you are advocating a really complex solution that will  
just reduce bandwidth use and typing for a few edge cases. In  
general, having presentation in CSS and semantics in HTML works just  
fine, with little duplication of effort.

>>>     Robot will download 1 css-file in addition to 100 html-files
>>> of each site. What are you economize !!
> DD> I don't see a great many attributes duplicated across pages
> Size of CAS will be less, than size of CCS.
> But CAS
> 1) economize human efforts

Not given the statement you responded to.

> 2) stricturize thinking.

I have no idea what this is supposed to mean.

>>>>> non-realtime UA can parse after downloading of csas-file.
>>> DD> That still means they have to built an in memory DOM tree and
>>> modify it based on the stylesheet.
>>> Yes, and what ?
> DD> This makes them harder to write,
> +


> DD> require more memory and CPU time
> See first answer.

A plus sign is not an answer.

> DD> slower,
> Not actual for non-realtime UA.

(a) Garbage. If something needs more processing, then it will take  
more time given the same hardware.

(b) What about "realtime" UAs?

David Dorward
Received on Wednesday, 16 January 2008 15:45:14 UTC

This archive was generated by hypermail 2.3.1 : Monday, 2 May 2016 14:27:33 UTC