[AFS3-std] Re: Encoding IPvN addresses

[Andrew Deason]

On Thu, 10 Feb 2011 14:25:35 -0500
Jeffrey Hutzelman <jhutz@cmu.edu> wrote:

> > But if someone has an endpoint with a different 'port' width, or some
> > entirely different equivalent of a 'port', we have to rev the whole
> > protocol?
> 
> No; we probably need to do something for the transport layer along the
> same lines as what we're talking about for the session layer.  But
> let's not conflate them.

My concern is the overhead. If we do the same thing for all of the
layers, we need effectively the same structure for the IP address, port,
and service number. So, for an RPC that just wants an endpoint to talk
RX to (i.e. all of them, right now), you'd have something effectively
like:

struct {
  int addrtype;
  opaque addr_data<>;
  int transport;
  opaque transport_data<>;
  int app;
  opaque app_data<>;
};

Which is about 16 extra bytes per encoded "address". But is that not
worth fussing about?

-- 
Andrew Deason
adeason@sinenomine.net