Re: Fwd: [ipv6mib] So, where were we?

[Michael MacFaden <mrm@riverstonenet.com>]

On Wed, Oct 09, 2002 at 10:11:59PM +0200, Juergen Schoenwaelder wrote:
>We obviously disagree on this since probably our estimation how 
>many boxes support simple FIBs and
>how many boxes support more complex FIBs diverge. 

Here is my estimate of how many boxes are running simple unicast IPv4
FIBs:

Total = number of windows 95/98/NT/2k/XP + all bridges + IOS IOS routers 
*not* running with vrf.

Hosts that have both IPv4 and IPv6 are a minority but growing.
So there will have two FIBs and they should remain two different tables.
I see no reason to make life complex by merging these two tables.

If one has multiple IPv6 or IPv4 stacks, the entity mib approach 
is appropriate. Implementations are doing this for multiple instances 
of STP now.

The last, complex case is of multiple FIBs. They  exist only on routers 
running vrf.  In such cases, I believe this could best be handled 
by using a different rfc/mib module where the tuple is:
  (interface, route-map-name, prefix, who/age, nextop)
 
>Operators want to have interface where they can ask questions such as
>"to which hop is a packet with desctination X forwarded"? 

Exactly. And my example answers that and tells you where that
route came from and how long it has existed in the device. 

If the box supports some fancy routing, and I estimate at 
< 1% of the time, then we should use capacity bits in the base
mib indicate to the mgmt app to look elsewhere.

>See, we do not seem to agree on the 99% case.

Quite.

Regards,
Mike