Re: auth handshake and rendevouz objects

[Marcus Brinkmann]

On Tue, Nov 05, 2002 at 08:25:25AM +0100, Niels Möller wrote:
> The entity that keeps track of the number of references is S, right?
> Perhaps one could give S the responsibility for doing the right thing
> with references and decisions on killing the underlying object.

Maybe.

> I guess there *has* be some timeout anyway, so that the "notes" or
> "unacknowledged references" that S keeps track of will expire if they
> aren't acknowledgemed in a timely manner. Who should determines this
> timout, A or S?

Well, one question is surely resource allocation.  I think A should provide
the resource to S that contains the note that B should get a reference to
the handle.  And if you do that, I don't think that it is a problem if A
decides about the timeout, because it also provides all the resources
related to it.
 
> You may still want an acknowledgement so that A can know if the
> transfer was successful. This can be sent by either B or S, it depends
> on who you think are most qualified to know about success. B seems
> more natural to be, but I'm not sure how much it matters. But this is
> a success/failure indication for A, independent from reference
> counting and object destruction.

S should not send messages to A.  Any server to client message is only
adding hair (well, you can make it safe with zero timeouts etc, but still). 
In the usual case, if A sends a message to B (containing a handle or not) it
will block without a timeout until it receives the reply message, so it is
usually safe to just make the acknowledgement of the handle implied by the
reply message, and acknowledge it from B to S before sending the reply.

Exact details like this probably depends on what we really need and use in
the Hurd.

Thanks,
Marcus

-- 
`Rhubarb is no Egyptian god.' GNU      http://www.gnu.org    address@hidden
Marcus Brinkmann              The Hurd http://www.gnu.org/software/hurd/
address@hidden
http://www.marcus-brinkmann.de/