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

Tatuya,

This suggestion by you:

From:
   In addition to the Prefix List, individual addresses are on-link if
   they are the target of a Redirect Message indicating on-link, or the
   source of a valid Neighbor Solicitation or Neighbor Advertisement
   message.  Note that Redirect Messages can also indicate an address is
   off-link.  Individual address entries can be expired by the Neighbor
   Unreachability Detection mechanism.

To:
   Section 2 of [RFC4861] defines the complete list of cases where an
   address is considered on-link.  Note, in particular, that Redirect
   Messages can also indicate an address is off-link.  Individual
   address entries can be expired by the Neighbor Unreachability
   Detection mechanism.

makes sense to us - we will take text from you - thanks. 

Further, it's better continuing discussion related to the 4th bullet of
on-link definition from the Terminology section of RFC4861. We can
totally not touch it in our draft - that is what we will do. We will
also remove any mention of the source-address rule in bullet 2 of
section 2 of our draft.  That's closure on our draft for the last call.

However, we can start that 4th bullet discussion right now where I and
Wes will justify why we wanted the 4th bullet deprecated. Thomas has
been saying that as well. We will explain why we do.

You see, the data forwarding table on a host takes precedence over
ND-cache. The data forwarding table consults the Destination Cache. If
an IPv6 node is a router then the data forwarding table is called the
routing table. We router folks do not like the 4th bullet because we do
not want our routing table cache to be touched merely based on source
address of packets received by the router interface. You see, the 4th
bullet seems to only apply to a router-less network but if one starts
applying the rule to routers, that is wrong! There are other reasons too
that we do not like the 4th bullet - the reasons have been given in
emails to this mailer.

Hemant & Wes 

-----Original Message-----
From: Jinmei_Tatuya@isc.org [mailto:Jinmei_Tatuya@isc.org] 
Sent: Tuesday, July 08, 2008 10:32 PM
To: Hemant Singh (shemant)
Cc: Thomas Narten; ipv6@ietf.org; Brian Haberman; Bob Hinden
Subject: Re: 6MAN WG Last Call:draft-ietf-6man-ipv6-subnet-model-00.txt

At Thu, 3 Jul 2008 17:35:56 -0400,
"Hemant Singh (shemant)" <shemant@cisco.com> wrote:

> Will something like this work for you - we have replaced "change" with

> "clarification".
> 
> [The source of an ND message is no longer used for on-link 
> determination, which is a clarification of bullet four of on-link 
> definition in Terminology section of [RFC4861].]
> 
> or
> 
> [The source of an ND message is no longer used for on-link 
> determination, which is a clarification of on-link definition in 
> Terminology section of [RFC4861].]

I object to this type of change (I have an alternate proposal.  See the
end of this message).

> I agree with you, no implementations have actually interpreted this 
> source address rule, but when folks like David notice the bug, folks 
> quote the definition of on-link in Terminology section of RFC4861. 
> That is why we thought of adding this line to our doc.

First off, I believe we should be more careful before "deprecating"
this rule:

                    - any Neighbor Discovery message is received from
                      the address.

As an implementor, I've been aware of the non-trivial flavor of this
on-link condition, and have noticed it creates tricky corner cases.
But I've always thought it's intentional, rather than a "bug".
Although I was not 100% sure in which case this rule is expected to be
used, I've imagined it might be useful in a router-less network (as we
briefly discussed in a different thread) or some NBMA network.

BSD variants have in fact supported this behavior for years: I've just
tested this with a FreeBSD 7.0 box and confirmed it (that is, if that
host receives an NS from an address that is not covered by "on-link"
prefixes advertised by RAs, it will create a specific neighbor cache
entry for that address).  I've also quickly checked that Linux
(seemingly) supports this behavior, too.

I also suspect there may even be a TAHI test item that requires this
behavior (since it's tricky, and TAHI tests check tricky behaviors "by
definition":-).  If my guess is correct, there will be other
implementations that support this behavior to qualify for an "IPv6 ready
logo" or something.

Now that I've seen the "inventors" of the neighbor discovery protocol
say it's a mistake and that I've also been wondering for what purpose
this rule is expected to be useful, I'd not necessarily be objecting to
the idea of deprecating this rule.  But since this will affect
(potentially many) existing implementations, this should be more
carefully and explicitly discussed with the implementors who have
already supported the behavior.  Perhaps we could have this discussion
when we want to revise RFC4861 to the full STD, but it should not be
done in part of a branch thread on a derivative "clarification"
document.

(BTW, I've seen a "security" concern on this behavior in a different
thread.  I was not sure about whether it's a mere FUD or it has a really
serious new security implication, due to the lack of details, but in any
event I'd not buy that argument in this context.  Since ND is a
link-local protocol, any attack including this one must be performed
within the same link as the victim.  And once we consider such a hostile
environment, we already know the neighbor discovery as a whole is just
vulnerable to many serious attacks.  If we want to counter this
situation, we must deploy SEND; simply deprecating one minor corner case
doesn't help).

Now, going back to the wording of the "subnet model" document, I suggest
the following change to reflect this discussion:

From:
   In addition to the Prefix List, individual addresses are on-link if
   they are the target of a Redirect Message indicating on-link, or the
   source of a valid Neighbor Solicitation or Neighbor Advertisement
   message.  Note that Redirect Messages can also indicate an address is
   off-link.  Individual address entries can be expired by the Neighbor
   Unreachability Detection mechanism.

To:
   Section 2 of [RFC4861] defines the complete list of cases where an
   address is considered on-link.  Note, in particular, that Redirect
   Messages can also indicate an address is off-link.  Individual
   address entries can be expired by the Neighbor Unreachability
   Detection mechanism.

The key intent of the proposed change is to avoid the explicit duplicate
of the conditions.  So, even if we eventually make a consensus on
whether to deprecate any "on-link" conditions described in RFC4861,
probably in a context of revising the RFC, it will be less likely that
some implementors naively implement the would-be-obsolete rule by simply
referring to the "subnet model" document.  I also believe this approach
is generally desirable because this document just tries to clarify some
subtle points of RFC4861, rather than making a substantial change to it.

---
JINMEI, Tatuya
Internet Systems Consortium, Inc.