Re: Necko e10s specs

Robert O'Callahan wrote:
> For those not following the WHATWG thread --- there's interest in adding 
> an asynchronous transactional API for localStorage, but less interest in 
> breaking compatibility with the existing API. There's quite a bit of 
> discussion about what properties the transactional API should have.
> 
> Also it seems to me that the global storage mutex can be implemented 
> with a per-domain mutex; we should be able to avoid all situations where 
> script in one domain synchronously triggers script in another domain 
> while holding the storage mutex.
> 

And if someone can think of a situation *not* involving plugins (the 
current spec handles that case) where script in one domain could 
synchronously invoke script in another, he/she should file a bug *now*. 
  From what I understand this is a potentially major security hole, in 
addition to a bother for storage mutex implementors.

I wonder if we shouldn't try just killing off plugins if they attempt 
the sequence of events "page A call-into plugin, plugin call-into page 
B."  What good could come of that?  (Of course it might massively break 
plugins and we'd have to turn it off.)  Disallowing that would make the 
storage-mutex impl easier for us and better for web pages because we 
wouldn't need to drop the mutex when calling into the plugin.

Cheers,
Chris