Hi Shoishi:

I agree with you that the expanded packet is UDP. Whatever is inside
that UDP packet is not IETF concern long as the UDP guidelines
(http://tools.ietf.org/html/draft-ietf-tsvwg-udp-guidelines) from
transport area are followed by the upper layer.

So again (and again):

- The draft is not talking about a zero checksum: there appeared to be
some confusion with the discussions around UDP tunnels in CAPWAP, AMT or
LISP. Here, once the packet is expanded it is a valid UDP packet with a
correct checksum. Routers on the way can verify that checksum if they
care. The receiver receives a UDP packet with a proper UDP header and
has to verify that checksum. What happens next is beyond UDP. 

- The draft not talking about hop-by-hop Message Integrity Check within
the LoWPAN: Geoff introduced some confusion yesterday. 

- The draft is not loosing end to end integrity: the checksum can be
elided ONLY IF there is an end-to-end integrity mechanism above UDP that
is better than UDP checksum, covers everything that the UDP checksum
covers (including the UDP pseudo header) etc... Yes we are talking about
an improved UDP, but it inherits all the UDP properties so it is still
UDP.

But it is critical that the HC2 defined in the draft expands into a NH
of UDP (17) even when the checksum is elided:

- UDP is ULP agnostic: 6LoWPAN constrains the range of ports (F0Bx) so
it implies that a port number will be overloaded depending on the
deployments. Considering this, it makes a lot of sense that some magic
happens at the ULP to make sure that the message being received is
expected, IOW it makes a lot of sense to have an additional MIC that not
only verifies the integraty but also authenticate the data.

- UDP is well understood by now: with UDP, come the UDP guidelines and
all the litterature.

- UDP is a good fit for smart objects: It is lightweight and enables
applications to save battery in constrained devices by piggy-backing
acks and payload.

- what would it expand to anyway? Defining an adhoc (6LoWPAN?) transport
protocol, we'd loose UDP handling in the Internet. NATs (yes they are
still there). Firewalls. Not defining it, we'd force other parties that
would implement 6LowPAN with checksum compression to invent their own
stuff. That's exactly the opposite of our mission, not only as 6LowPAN
but as IETF.

My take is to stick to what the draft does. We can improve the wording
but changing the bottom-line would hurt terribly the applicability of
this work.

Pascal

>-----Original Message-----
>From: [EMAIL PROTECTED] [mailto:[EMAIL PROTECTED] On
Behalf Of Shoichi Sakane
>Sent: mardi 18 novembre 2008 19:47
>To: 6lowpan
>Subject: [6lowpan] UDP checksum elide
>
>Hi,
>
>In today's meeting, we didn't make a consensus to have the UDP checksum
>elide bit in the HC.  I think that it is something like UDP-NAT.
>The 6lowpan adaptation layer is responsible to realize it.
>In my understanding, the UDP stack of both end devices should calculate
>the checksum, and the 6lowpan layer elides or recalculate the checksum.
>I believe it is UDP still.
>_______________________________________________
>6lowpan mailing list
>[email protected]
>https://www.ietf.org/mailman/listinfo/6lowpan
_______________________________________________
6lowpan mailing list
[email protected]
https://www.ietf.org/mailman/listinfo/6lowpan

Reply via email to