Subject: Re: NetBSD master CVS tree commits
To: None <perry@piermont.com>
From: Gordon W. Ross <gwr@mc.com>
List: current-users
Date: 02/01/1996 13:11:28
> Date: Wed, 31 Jan 1996 18:28:49 -0500
> From: "Perry E. Metzger" <perry@piermont.com>

> I keep wondering whether the bpf stuff should be replaced with an
> (easy to produce) RARP protocol socket, MAC protocol raw socket, or
> similar kernel hook. The code for it is trivial and needing to do BPF
> all the time and take the performance hit on each packet on servers
> doesn't feel right. It could also help other things if, for
> instance, it was a general MAC protocol hook.

It would be nice if one could open some sort of "MAC protocol" socket
(or device? as in DLPI?) and bind that to a MAC protocol, thereafter
receiving all packtes with that protocol (i.e. RARP, Appletalk,...)
Such a thing probably needs to be per-interface to be useful, because
most things one would do with it need to know which interface each
packet came in on, and send needs to go to a specific interface...

Maybe we could have a "Raw MAC protocol" socket which needs to be
bound to both an interface and a MAC service access port (SAP).

Gordon