Dan Matejka <[EMAIL PROTECTED]> also wrote... > > In the end, and you might be interested in this, apparently the reason > for implementing this in JS in this case was that doing it in C++ > required writing a callback for use from JS, and that for some reason > necessitated a compile-time dependency on LDAP. But it seems bug 89137 > is about moving the offending code to the LDAP directory anyway, so > that's not an issue and it looks like Mitesh will be doing it in C++ > after all. > > Short summary: I can't say that doing this stuff in JS is the root > of all evil, but I'd really rather see it done in C. I was thinking > of making nsIEventQueueManager unscriptable. Would that screw you > over?
I'm guessing that you really meant nsIEventQueueService. Yeah, making that unscriptable would make it impossible to use nsISupports proxies from JS, which seems like a bad thing, because then it's impossible to write pure JS components which do work on other threads but need to ultimately call back to the UI thread since lots of the UI stuff isn't threadsafe. This would break at least the LDAP datasource code; maybe other stuff too. Dan
