W3C home > Mailing lists > Public > www-style@w3.org > February 2005

Re: [css3-background], comments on 'border-radius'

From: Andrew Fedoniouk <news@terrainformatica.com>
Date: Fri, 25 Feb 2005 22:02:19 -0800
Message-ID: <000901c51bc8$bb7037b0$0c01a8c0@TERRA>
To: "Bert Bos" <bert@w3.org>, <www-style@w3.org>

Bert,

the problem is not in 'border-radius' itself.
Problem (IMO) is in the number and complexity of overall
system of CSS attributes. We are adding attributes one by one and finally 
we've
got a system of almost unmanageable set of attributes.

See, 'simple' border-radius immediately introduces questions:
e.g. how to calculate the content box - does radius value 
increases/decreases
padding? If 'no' then it is possible that border will draw on top of the 
content
or underneath. So the second question - order of drawing of border/content?

Next question: What will border conjunction look like for the following:

border-left: 3px double red;
border-top:5px dotted blue;
border-right:2px groove green;
border-right:3px wave gold;
border-radius:1em;


Any brave man to define this?
Just imagine how many combinations this introduces.

|... Using the platform-provided drawing routines gives you
| a better chance of getting high-quality (and fast!) rendering.

I have some doubts about "and fast!".
Drawing of such rounded corners even without antialiasing
is in fact polygon fill which is not faster than bit-blit (in case
of images).

And cascading...

See, definition of shape of
element (its visual style)  consist (now) from
4+4+4+8+2+1 = 23 border attributes
and 9 background attributes. In total = 32 (sic!).

Almost always all shape attributes are tied together.
So if you will want to redefine e.g. border width you should
redefine (or clear) also radius, etc.

IMO, CSS attributes as one big non-structured
rack, as a system, has reached its limits. In many senses.
I don't think that it makes any sense to add anything there without
introducing modularity. I mean real modularity - not
just a splitting spec. onto ten chapters.

Andrew Fedoniouk.
http://terrainformatica.com



From: "Bert Bos" <bert@w3.org>
|
| On Thursday 17 February 2005 08:47, Andrew Fedoniouk wrote:
| > Under the light of border-image: do we need border-radius at all?
| >
| > To be implemented with good quality on media:screen rounded corners
| > should use
| > antialiasing.  No other module in CSS requires AA. Of course if SVG
| > is not a part of renderer.
| > I think that specialized tools like XaraX will produce better bitmaps
| > with rounded corners than any CSS renderer in runtime.
|
| Not necessarily. You don't know what screen the user has. It may be a
| CRT, a vertically striped LCD, a printer with much higher resolution
| than you are prepared to make images for, or a handheld screen with
| limited colors. Using the platform-provided drawing routines gives you
| a better chance of getting high-quality (and fast!) rendering.
|
| But it is not forbidden to provide SVG images for the corners. It's an
| extra download and you may have to make alternative images for UAs that
| don't implement SVG, but if you want to have that extra precision, you
| can.
|
Received on Saturday, 26 February 2005 06:02:42 GMT

This archive was generated by hypermail 2.2.0+W3C-0.50 : Monday, 27 April 2009 13:54:36 GMT