cvs commit: src/sys/dev/bce if_bce.c src/sys/dev/bge if_bge.c
src/sys/dev/em if_em.c src/sys/dev/ixgb if_ixgb.c src/sys/dev/nfe
if_nfe.c src/sys/dev/nge if_nge.c src/sys/dev/re if_re.c
src/sys/dev/stge if_stge.c src/sys/dev/ti if_ti.c src/sys/dev/txp ...
Bruce M. Simpson
bms at FreeBSD.org
Sun Sep 17 13:08:23 PDT 2006
Julian Elischer wrote:
> As I mentioned before, I am slightly uncomfortable with the
> implementation of this
> change as it puts protocol specific items into the protocol
> independent mbuf header.
> The fact that 99.99% of network traffic coming in and out of a
> machine uses this protocol
> at the the moment makes it understandable but if in 2 years a new
> transport mechanism sweeps
> the world for which this is irrelevent, or worse, has a different
> requirement for similar fields,
> are we going to add fields for that too? should this be defined as a
> link layer specific union for
> which we can add future variants?
This argument seems a case of putting the cart before the horse -- it's
a bit up in the air. If anything Ethernet is more likely to increase in
popularity, and it is the most common use case. Particularly so, given
that VLAN encapsulation is specified for 802.1p priority tagging, and
there is more and more interest in L2 QoS because of VoIP.
Kudos to Andre for sorting it out.
More information about the cvs-src