RE: 6MAN WG Last Call:draft-ietf-6man-ipv6-subnet-model-00.txt

David,

What is UHL?

Then please see in line below between <hs> and </hs>

-----Original Message-----
From: MILES DAVID [mailto:David.Miles@alcatel-lucent.com.au] 
Sent: Sunday, July 13, 2008 11:15 PM
To: ipv6@ietf.org; Hemant Singh (shemant); erik.nordmark@sun.com; Wes
Beebee (wbeebee)
Subject: RE: 6MAN WG Last Call:draft-ietf-6man-ipv6-subnet-model-00.txt

Wes,

There are implementations that update their routing table on the receipt
of a Neighbour Advertisement which we should consider. One example is
the code developed in the KAME project, which can be found in many
BSD-based distros. On receipt of a valid NS, the neighbour cache is
updated with a stale entry and if a route does not exist for the
destination a host route is created (with flags UHL). 
This behaviour makes sense if we consider the old on-link assumption -
but it opens the security concern I expressed in v6ops. It is quite
possible for an on-link node to create route entries in that may affect
other links (say the target of this were an ISP router). I do not think
this behaviour is desirable in a router.

I would prefer to avoid behaviours that create bogus entries. It seems
that we defiantly need clarification around correct node behaviour, and
if we are clarifying (to be explicit that the receipt of a ND has no
affect on forwarding) then should we go so far as to avoid the bogus
entry?

For interest; in the KAME example will drop a received NA when the
target-address is not in its Neighbour Cache (the NA is discarded).

<hs>
This behavior by KAME is correct. See text snipped from section 7.2.5
that mentions this behavior reported by you for KAME.
[When a valid Neighbor Advertisement is received (either solicited or
   unsolicited), the Neighbor Cache is searched for the target's entry.
   If no entry exists, the advertisement SHOULD be silently discarded.]
</hs>

Hemant

-David

>>>
Sorry to reopen this, but do you think that the following clarification
could be added to the IPv6 Subnet Models draft to address bullets three
and four of the on-link definition in the Terminology section of RFC
4861:

"Since only the Neighbor Cache is updated with the source address of a
received ND packet or the target of an NA packet, and the Destination
Cache and Prefix List are not updated, an ND packet cannot indicate that
a destination is on-link in the absence of corresponding on-link prefix
information."

What does the WG think?

- Wes

*****
--------------------------------------------------------------------
IETF IPv6 working group mailing list
ipv6@ietf.org
Administrative Requests: https://www.ietf.org/mailman/listinfo/ipv6
--------------------------------------------------------------------