Rounded performance issues

[bill]

From the mailing list we had a discussion about performance issues with the current rounded corners implementation. Tom wrote:

The current solution is a "thousand div" solution that uses block elements to emulate pixels; I personally never liked it but the source was ported by permission of the author. There are a few other similar solutions out there that are a lot more streamlined, but almost all of them were released under LGPL.

If the group wants to come to a consensus conclusion about what we are willing to support, I'd be *more* than happy to trash the current solution.

The current implementation supports:

1. Anti-aliasing
2. different radii (w/h) per corner
3. full background image support
4. full transparency support
5. automated padding, if asked for
6. liquid box sizing

Current implementation does not support:

1. nested widgets (my fault, need to go back and change the way the padding is implemented; the original used innerHTML and it should have used appendChild).

---

Possible alternatives include: CSS3 radius (my personal preference, of course no IE support) gfx with some fallbacks, if needed image-based solutions (i.e. if it becomes a reality, tapping into that Google service)

[Adam Peller]

is this code going to dojox, or should we close this?

[bill]

I remember Tom saying we wouldn't port Rounded to Dojox, so I'm gonna close this bug. Tom can reopen if I am misquoting him.