InNoRoutes / OutNoRoutes
MIB-II has ipOutNoRoutes for IPv4.
The old IPv6 MIB doesn't have any equivalent counter,
but adds ipv6InNoRoutes with a different definition,
of questionable usefulness (and no rationale in the RFC).

The current draft deprecates ipOutNoRoutes, and
leaves InNoRoutes as current.  

In our implementation, OutNoRoutes is still very useful,
and I see no reason for InNoRoutes.

I'd like to propose that we keep the standard MIB-II
definition of OutNoRoutes, and deprecate the ipv6InNoRoutes 
object.

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

Further investigation reveals that ipForwDatagrams in MIB-II
and ipv6IfStatsOutForwDatagrams in RFC 2465 actually have
quite different semantics, and so the suffix collision is
somewhat confusing.  

The ipv4 counter counts the number of ATTEMPTS to forward,
whereas the ipv6 counter counts the number of SUCCESSES.

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

Erik Nordmark:

I don't have the MIB or the Solaris code in front of me but I recall
looking at this ...
Assuming per interface statistics, when forwarding datagrams and there
is no route found, using OutNoRoutes is a bit undefined - for which
interface
do bump the counter?
In this case InNoRoutes makes more sense.

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

Dave:

That's true.  MIB-II uses global counters, and the old IPv6
MIB uses per-interface statistics, and now I understand
why the difference, so thanks for the explanation.
(I think this motivation should be added to the draft.)

Perhaps there should be separate counters for InForwNoRoutes,
and OutNoRoutes (not including packets being forwarded).