> On Tue, Sep 21, 2004 at 01:44:22AM +0200, Pavel Cahyna wrote:
> > [kqueue and events via knote and kcont]
> 
> I think I like this approach, a lot, though care and thought will be
> required around the security potential for arbitrary kernel callbacks
> if the framework exposes too much to users.

I don't think that it exposes anything more to users, because the only way
to manipulate knotes from userland is via kqueue(2) API, but this sets the
callback uncontitionally to a callback which queues the event.

(BTW probably I won't be able to reply to emails until Monday.)

Bye	Pavel