Re: Anti-aliasing border corners [spin-off of Style confirmation descriptors]

I fundamentally disagree about one key thing in your message.

> border-radius is pretty much useless without anti-aliasing

This is simply not true. It is a /desirable/ feature for a user agent
to posess, since it improves the aesthetics, just like anti-aliasing
on fonts when operating systems started shipping with it. That does
not, however, make the border-radius property useless without. Heck,
enough people have prettied up their sites using the incomplete
Mozilla -moz-border-radius implementation to show that (the Mozilla
implementation currently has no AA).

Whether corners, fonts, images or whatever get anti-aliased is
entirely the user agent's business. If a UA is being developed for a
low-resource device (old PC, handheld, mobile phone) then
anti-aliasing is not something that's going to be added in. For a full
powered desktop browser then maybe it is.

But it's a perk. It's something that might leave you preferring one
browser over another. CSS is not a pixel perfect system and never will
be. This kind of control is the business of the UA and whether or not
a UA implements it is the business of those developers.

It's not a feature that has any particular need to be controlled. If a
UA implements smoothing on border-radius (or any other border) then
it's up to the UA to ensure that it doesn't alter the way the page is
rendered (in terms of positioning, compared to the the rendering if it
was not anti-aliased). I can't possibly see how there's anything more
to it than that.

Sorry, this reply was too long for the subject matter.

On Apr 10, 2005 11:52 PM, Barry <wassercrats@hotmail.com> wrote:
> 
> Laurens Holst wrote:
> > Besides, even when you would have a need for anti-aliased horizontal or
> > vertical lines (this could happen when drawing borders on non-whole-pixel
> > boundaries, which I don't think is a good idea), it wouldn't make anything
> > larger, the borders would still be exactly 1 pixel wide.
> 
> > In any case, there's nothing much mysterious about anti-aliasing, and it
> > can be entirely handled by the user agent and CSS doesn't need to worry
> > about it.
> 
> I looked at a <span style = "font-family: verdana; font-size:
> 160pt;">P</span> with and without Windows XP font smoothing. With font
> smoothing, the straight top part of the "P" gained an extra pixel in width
> for a length of five pixels. This reduced the straight, equal-widthed
> section from 31 pixels to 26 pixels in length. With more aggressive
> anti-aliasing, it would be reduced even more.
> 
> If you specify a two-pixel top margin for text contained within a border
> with rounded corners, there might be only one pixel above the text for a
> length of several pixels, or maybe for the entire length of the top border,
> depending on the extent of the anti-aliasing. I don't know enough about
> anti-aliasing to know how aggressive the anti-aliasing would need to be
> before an entire top border would be an extra pixel thick. I'm just raising
> this issue incase it really could be an issue.
> 
> It might have to be made clear to the web developer whether the margins
> being set will be measured from the border as though there was no
> anti-aliasing, or if the margins will be measured from the anti-aliasing
> pixels. If there are no standards for this, it could be done either way,
> depending on the browser. There are three ways this issue can be addressed:
> 
> 1. Create standards for how margins are handled in the event of anti-aliased
> corners.
> 
> 2. Have text be anti-aliased whenever border corners are anti-aliased. Not a
> good option.
> 
> 3. Make sure the straight parts of borders aren't thickened by
> anti-aliasing. I don't like this either.
> 
> If a web developer chooses rounded corners, he'll know about the obvious
> effect of there being less corner space, but extra pixels that thicken the
> "straight" sides of the border might pose a problem--without standards.
> 
> And as long as you're creating these standards, and since border-radius is
> pretty much useless without anti-aliasing, you might as well add an
> anti-alias property.
> 
> 


-- 
http://www.ben-ward.co.uk

Received on Sunday, 10 April 2005 23:42:14 UTC