This would solve lots of usecases: - Add a custom search bar - Add a complicated toolbar widget - ... Allowing extensions to add sandboxed iframes as toolbar items seems like a safe and sustainable way to satisfy the usecases above.

We talked about this during the last advisory board meeting, and the consensus was that we almost certainly do not want to support this.

I think the next step on this bug is to ask Kev for the PRD for if we want to do this. If we did, it would probably be lower priority than a sidebar.

I really feel strongly about this. It's much more flexible that allowing an iframe as toolbar because you can drag around individual elements from toolbar X to toolbar Y, and that's a really important thing that is part of the Firefox identity. People would be able to inject (in a sandboxed iframe that has no access to the chrome): - a custom search bar - a weather widget - a 3-button media player widget - lots of other use cases, I don't think for any of these cases we want to inject a complete toolbar. The best part is they can put it anywhere they want it in the UI, like the menu panel, the nav-bar, the tab-bar, ... This really doesn't hold back the development of Firefox and isn't susceptible of breaking anything, so I really insist in supporting this.

There was a strong consensus that we don't want to implement this, for a variety of performance, maintainability, and UX reasons, particularly those we've already experienced in the SDK implementations. As follow-ups: - Mike Kaply is going to look into, and file bugs for, UI improvements we can make to the existing toolbar button support. - Shane Caraveo is investigating custom toolbar support, which will likely include some sandboxed iframe functionality.

@Kris Maglione I'm not versed in the terminology, so I'd ask directly: is it a problem for pentadactyl? (I'm asking that because you said to be one of its devs, and because I'm worried about an API for it to work).

(In reply to Hi-Angel from comment #5) > @Kris Maglione I'm not versed in the terminology, so I'd ask directly: is it > a problem for pentadactyl? (I'm asking that because you said to be one of > its devs, and because I'm worried about an API for it to work). Pentadactyl only needs bug 1215064 I think.