On 06/10/10 12:31 PM, Jonathan Hui wrote:
So a packet sent by R1 that will be forwarded outside of the ROLL
network will have a outer IPv6 header whose destination is the BR?
That is where we started. Draft-01 does have a line or two about the
possibility of exempting the last entry in a RH4 from the strict source
route rule, thus allowing R1 to send a packet without an outer IP header
but this idea still needs to be fleshed out. It would require the BR to
strip the RH4 and update the IPv6 Payload Length accordingly. Are there
any implications with this idea?
The first problem is that the draft isn't clear what it is saying on
this point. Hence my clarifying question and the need to clarify the draft.
For the technical point see below.
It wasn't clear to me whether you are proposing IPv6-in-IPv6 tunneling
for packets that come in via a border router.
Not sure I'm clear on your use case, but the restrictions in the draft
are intended to prohibit the case where an RH4 cross a RPL domain
boundary. A RH4 may be added and/or removed at a border router, but it
shouldn't process/maintain an existing RH4 across a RPL domain boundary.
Understood. My question is when the BR adds RH4, does it also
encapsulate in an outer header?
Yes, that is the case with draft-01.
Thus is the only case when there is RH4 without IPv6-in-IPv6 when both
the source *and* the destination is a router in the ROLL network?
Certainly the source must be within the ROLL network. Whether or not the
destination has to be depends on whether it is acceptable for the border
router to strip RH4 headers that have have a Segments Left = 0 after
processing.
Thoughts?
It seems quite asymmetrical to have the packets from R1 via BR to some
non-ROLL destination not use IPv6-in-IPv6, while packets in the reverse
direction must use IPv6-in-IPv6.
That begs the question whether the same asymmetry is there for packets
from R1 to a host that is directly attached to a the ROLL router R2.
Thus would R2 strip RH4 and update the payload length for packets from
R1 via R2 destined to the host, while the packets from the host via R2
to R1 would be IPv6-in-IPv6 encapsulated by R2?
Erik
--------------------------------------------------------------------
IETF IPv6 working group mailing list
ipv6@ietf.org
Administrative Requests: https://www.ietf.org/mailman/listinfo/ipv6
--------------------------------------------------------------------
