See comments in bug 296639. Basically, I think that we should never null out mScriptGlobalObject on documents; the inner window should just be owned by the document. Conceivably they should even share a refcount, though that's hard to do with the cases where we reuse inner windows...

I agree, this should make it possible for documents in the bfcache to be much more 'alive' (although fiddeling with them too much will get them evicted from the cache). What mScriptGlobalObject should documents loaded from xmlhttprequest and the likes get? The same as the 'owner document' or a new one?

That's a good question. What script global and container, if any, should those get? Right now, loading via XMLHttpRequest sets no container or script global, iirc. Creating via DOMImplementation sets a container but no script global. I'd say documents not attached to a window should have no container (which is basically the docshell/outer_window) and documents that don't plan to run scripts or dispatch events should have no script global. So do we ever run scripts or dispatch mutation events or whatnot in such documents? Should we?

(In reply to comment #2) > So do we ever run scripts or dispatch mutation events or whatnot in such > documents? Should we? Bug 201236 for mutation events; I plan on dispatching events if they have a mutation event listener, even if they lack a window.

*** Bug 308902 has been marked as a duplicate of this bug. ***

https://bugzilla.mozilla.org/show_bug.cgi?id=1472046 Move all DOM bugs that haven’t been updated in more than 3 years and has no one currently assigned to P5. If you have questions, please contact :mdaly.