Comment 15 for bug 197781

> We shipped 1.5 with document.loadOverlay in very questionable state, so it's
> probably not such a big deal.

Sad to say, but Toolkit's prefwindow definitely isn't quite ready for realworld usage outside of simple cases like Firefox' prefpanes. See also bug 410562 and others.

> More to the point, this code is rather messy and doesn't care about *so* many
> cases, that making sure it works correctly in all code paths is quite hard.
> It would be easy to cook up a patch to make successive loadOverlay calls work
> properly in the common cases, but then you're replacing a clear bug (only one
> overlay can be loaded at a time) with lots of subtle bugs and the risk to
> introduce security risks.

Shouldn't it suffice to just queue the overlays and then walk back notifying the observers? You'd definitely don't want to load an overlay twice to avoid looping, but (simple) ested overlaying should work.

Which security risks do you expect? (While of course always possible, this pretty much sounds like a scarecrow here.)