UDP/TCP stats

The scalars at the top of each MIB are currently
IPv4/6 independent.  However, we find it useful to
track them separately.  So we actually keep separate
global counters for IPv4 and IPv6 (and indeed it's
infeasible to keep combined counters since our
current TCP/UDP modules are different modules
in IPv4 than in IPv6).

Now I can have the mib agent add together the two
counters, to get a "total" running counter, but
I guess I'm asking for the scalars to be duplicated
in the MIB between IPv4 counts and IPv6 counts.  
This has the additional benefit, when errors get 
counted, one knows whether to look at IPv4 or IPv6 traffic.

------------------------------------------------------------------------

Juergen's reply:

So you are saying that every implementation must be capable to keep
IPv4/IPv6/... specific counters in the transport protocol stack? 

I have not checked what the Linux kernel internally does (the /proc
interface is a little bit confusing).

From an architectural point of view, I think the proposal is heading
back into the direction of having IPv4 and IPv6 MIBs...

------------------------------------------------------------------------

Bill:

On the other hand, FreeBSD has a unified TCP and UDP implementation;
TCP is aware of whether it's an IPv4 or IPv6 connection but doesn't
count anything seperately.