border-radius seems to be a hot topic these days, what with CSS3 including it and some broswer vendors already implementing it. I just think it looks nice.

I originally created this table to see how the various browsers acted, but it’s also a very useful table of examples of how to use border-radius in your designs. Even after many readings, some of the wording in the proposed specification can be confusing. I’ve done my best to get things right here.

The top half of this table is looking pretty good these days, if you ignore the elephant in the room whose name will not be mentioned (until later). Eventually this page will be a syntax reference instead of a compliance table. That will be a Good Thing.

Safari and Chrome
Big news, everyone! Safari 5 is out and has caught up with Google Chrome. And with Safari 5.1 we finally have support for expressing the radii as percentages.

If you do not need to support older versions of Chrome or Safari (and Apple’s Software Update will make sure Safari 4 goes away very quickly), then the -webkit- prefix is a thing of the past!

Firefox 3
As of Firefox 3.6.3, it is still necessary to use the -moz- prefix. More importantly, the syntax for specifying individual corners is different for Mozilla. I tested to see if the latest might also accept the syntax from the spec, but it did not. For the time being I believe it’s better to use the shorthand to specify the different corners. That way your CSS markup will be the same as the standard except for the -moz- prefix.

Firefox 4 and 5 and… 6?
Firefox 6? Really? Are we suddenly in a rush to bigger version numbers? Anyway, no big change since Firefox 4. I’m going to add some new tests, but I won’t go back and fill in old browser versions. At first glance, at least, FF6 = FF5 = FF4.

The biggest news with Firefox 4 is the move to the standard CSS for specifying individual corners. Hooray! It will still accept the old -moz-wrong-order prefixes, but going forward Firefox will be playing well with others. Firefox worked out a couple of other kinks as well, and is looking pretty good overall.

Opera
Opera 10.5 has border-radius and it’s pretty complete. There are a couple of odd bugs that make me think they just need to look at this page and their programmer will slap his head and say “d’oh!” and all will be well in the next release. [Note: Opera 10.6 still has those bugs.]

Internet Explorer
Wow! Internet Explorer 9 beta has, as of this writing, the best border-radius support of any browser. It will be a good day when we don’t have to support Internet Explorer 8 anymore.

Other notes
I used SVG graphics to create most of the images for the specification column. That way I was able to type in the exact drawing commands based on my interpretation of the spec. The graphics are my own work and any errors are all my fault.

So here’s the table. I’ll try to remember to update this as new browser releases come out, but you can help too! If there are interesting cases that should be in the table, or if I’m slacking and browser support has changed, please drop me a note!

Border is drawn in a way that does not match the proposed specification.

Border is drawn correctly but property name is different. Specifically, Mozilla specifies individual corners differently. Web designers should use the different syntax with the understanding that it will be obsolete later. Later rows in the table assume you understand that Mozilla is different.

Technically wrong, but not by much. Considering the things that the draft doesn't specify, the angle of the line between the colors seems pretty minor. The spec also "recommends" a gradient. NOTE: I am reviewing this right now; I suspect that Webkit is also subtly wrong. The spec uses English to describe a complex geometric relationship, rather than using math. There are several places they could have removed ambiguity and made compliance easier just by using numbers rather than words.

These two rows are exactly equivalent in terms of the CSS3 specification. I color-coded the values to indicate how the browser should expand the shorthand into the longhand.

The CSS3 draft merely says the transition between borders of different thicknesses be smooth, so although the Mozilla implementation does not match the illustration in the spec (it uses an elliptical curve on the interior where the illustration in the draft uses a circular curve), it is probably compliant. Web developers note that they cannot rely on inner borders looking exactly the same between fully-compliant browsers.

Note:

I recently updated my “Who the hell knows?” page. I added a section on how I think the standard should be for defining the transition between two styles at a curved corner. While I was at it, I corrected a couple of errors in the formulas as they were printed on the page. My interactive demo actually outperforms all current browsers for solid borders. The new section is really pretty slick!

Firefox 3 note
I should have tested with the latest Firefox 3 (3.6.12, I think) before chucking it off my computer. I’ll probably go back and test it at some point, or if anyone out there wants to check and give me an update, that would be great!

More on Webkit (Safari and Chrome)
The overhaul I said was necessary has happened. I don’t know if my test thingie helped them or not, but some very smart people went in and made things a lot better. Still waiting for dots, though…