It seems to be rounding to the nearest integer.
Why? Maybe performance.

hhp

Wouldn't surprise me. Safari has a pretty strict performance over features strategy.

Safari is doing it right, there is no such thing as “half a pixel”.

In the CSS spec there are half pixels.

http://www.w3.org/TR/CSS21/syndata.html#length-units
http://webkit.org/blog/57/css-units/

Thanks for the links, Jens. Very informative.

In my opinion the definition of “pixel” in the CSS spec is fundamentally wrong. But what do I know.

> Safari is doing it right, there is no such thing as “half a pixel”.

Hmmm, closet Apple-berserker talk? There most certainly are fractional pixels, on many levels. Especially on an OS that worships wysiwyg!

hhp

Looks like the genius of Håkon Lie is behind this one too – see his reply to the question about “Definition of pixel” here.

Seriously, if he knew what he was doing we wouldn’t be in such a mess right now.

WYSIWYG is the root of all font rendering evil.
I propose WYSIATI instead — What You See Is All There Is.

> Safari has a pretty strict performance over features strategy.

Well then, for max performance just snap to powers of 2!
Who really needs 9, 10, 11, 12, 13, 14 and 15 point anyway?
I'm sure the right kind of press conference can spin those sizes into oblivion.

> WYSIWYG is the root of all font rendering evil.

That's a primo cushy bandwagon now.

hhp

"Safari is doing it right, there is no such thing as “half a pixel”."

Non-sequitur, and I just showed you Firefox doing it.

and I just showed you Firefox doing it.
How dare you!

How dare you!

Brass balls.

Probably Safari rounds values from the start to avoid inconsistencies caused by rounding later, like the so-called Firefox leading bug?