RE: [css3-backgrounds] Should a non-zero border-radius create a new stacking context ?

They already have this same side-effect with opacity and transforms. Hence my asking for evidence that it *is* an issue, especially with opacity which is very popular. I'm not that interested in worrying about side-effects if no one runs into them outside test suites.

It's easy to explain because you're not really explaining anything :) It may be obvious to you what 'not changing z-order' implies for overflow, clipping etc. I'm not sure that's true for everyone.

And if opacity is special in this respect - i.e. one can make an obvious case to create a new stacking context for this particular visual effect - then I'm not sure we can use it to argue that it shouldn't happen in other cases. We can just say that it clearly made sense there. We can ask, however, whether its new stacking context creation has side-effects that cause z-order issues for authors; or we should argue why opacity is unlikely to be used in cases where those would cause a problem.

From: [] on behalf of Robert O'Callahan []
Sent: Monday, November 22, 2010 6:43 PM
To: Sylvain Galineau
Cc: Brad Kemper;
Subject: Re: [css3-backgrounds] Should a non-zero border-radius create a new stacking context ?

On Tue, Nov 23, 2010 at 3:03 PM, Sylvain Galineau <<>> wrote:

What complexity would have been added for authors by creating new stacking contexts for overflow != visible ?

The complexity of "setting this property value over here has this apparently unrelated side effect over there".

I'd argue the latter is what is actually expected by most, especially so when overflow == hidden. Having outside elements layer underneath the children of an element with overflow:hidden - never mind clipped children - is neither easy to explain nor is it common practice.

I think it's easy to explain. The explanation is "'overflow' does not affect z-order".

Now, the rest of the CSS z-ordering rules are not easy to explain, but that's a different problem :-).

 Not sure what we mean by 'reasonable behavior'. Reasonable for authors or implementors ?

Both, actually.

Group opacity means "composite all the elements of the group together into a single image, then composite that image into the destination". There's no rational way to extend that to handle situations where the elements of the group are interleaved in z-order with other elements that are not part of the group.

"Now the Bereans were of more noble character than the Thessalonians, for they received the message with great eagerness and examined the Scriptures every day to see if what Paul said was true." [Acts 17:11]

Received on Tuesday, 23 November 2010 14:59:01 UTC