> I wish kill() had a richer interface so one could pass data along with
> a signal and have one generic "something has changed" signal and tell
> what has changed in the passed data.  I believe the kernel can sort of
> do this right now for some signals...

As you say, there are a lot of events that a program might potentially
want to know about.  I think an asynchronous delivery mechanism like
kill() is probably the wrong way to send them.  I think that actually
providing an interface similar to the routing socket might be the way
to go with this.   Because there are a lot of potential things that
might need to be signaled, I'm not sure that actually using the
routing socket is the right thing, although it makes sense for
interface address changes.   Some other events I'd like to see
delivered are ``we just got back from sleeping,'' ``we're about to
shut down - please clean up,'' and ``all network users are shut down -
please shut down the network.''

I don't need this stuff right now, which is why I didn't propose it
earlier, but if you're talking about what sort of interface to
provide, I'd like this suggestion to be taken into account... :')

			       _MelloN_