Re: auth handshake and rendevouz objects

[Marcus Brinkmann]

On Wed, Oct 30, 2002 at 12:01:01AM +0100, Bram Meersman wrote:
> Marcus Brinkmann wrote:
> 
> >
> >We talked about auth and rendevouz objects before, and I think we concluded
> >that the rendevouz object should be managed by auth (at least that's how I
> >recall it).
> >
> >I just thought about it, and I think it should stay with the user.
> >
> Hi,
> I have read the discusion, and I don't think I like either of the protocols.
> 
> In both cases proxies are required to rpc to {Auth | Client} in order to 
> pass on the rendezvous handle ,
> so proxies will have to trust {Auth | Client}( if a thread performing an 
> rpc sets a 0 receive timeout, the rpc will fail ).
> 
> In Hurd/mach proxies just pass on port rights, only Client and Server 
> have to communicate with Auth.
> Client and Server might even be the only tasks who trust Auth.

I don't understand what you are saying.  Maybe this is because I don't know
what you mean by "proxy".  Passing a handle in HUrd on L4 is a complex
operation, it requires the recipient and the sender of the handle to contact
the server that provides the object which we pass the handle for.  But that
is ok, we make all the right guarantees so that it will have similar
semantics to passing a send right in Mach.  Except that in Mach we have one
central object server shared by all object providing servers, while in Hurd
on L4 we have one object server per object providing server process.

I don't see any other process involved in the protocol we discussed except
the user, the server, and auth.

Beside explaining what you don't like about this, I would also like to know
if you have any other suggestions?

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/