Re: draft-ooms-v6ops-bgp-tunnel-06.txt and v6ops review

[Eric Levy-Abegnoli <elevyabe@cisco.com>]

Hi,
MTU is not discussed in the draft, and I think MPLS carries  some specific 
issues that are worth describing.  Basically, if the MTU in  the MPLS core is 
smaller than 1280, IPv6 packets greater than 1280 bytes are likely to be 
dropped silently.
Any chance (if this not too late) to add a small section/text on MTU handling 
in the MPLS core ?  (at lean the first sentence)

Here is a proposed text:

MTU consideration
--------------------------
[IPv6] requires that every link in the internet have an MTU of 1280 octets or 
greater. On MPLS links where no link-specific fragmentation and reassembly is 
provided, MTU must be configured to have an MTU of at least 1280 octets plus 
the encapsulation overhead in order for 6PE traffic to be forwarded.

(Add optionnaly the following text)

Some IPv6 hosts might be sending packets greater than the MTU available in the 
IPv4 MPLS core and rely on Path MTU discovery to learn about those links. 
Core routers are typically not IPv6 nodes and cannot build an IPv6 ICMP 
"packet Too Big" message. Even if they can,  they may not have knowledge of 
the destination of the ICMP message and would be forwarding this message to 
the egress PE, using mechaninsm described in [RFC3032].
According to [RFC2463], the ICMP "Packet too Big" message can be filled with 
the invoking packet up to 1280 bytes. If any links on the path to the egress 
PE has an MTU smaller than 1280, the ICMP message won't reach its 
destination,  preventing the originator of the traffic to be notified about 
the MTU too small. This may cause significant operational problems; the 
originator of the packets will notice that his data is not getting through, 
without knowing why and where packets are discarded.
To minimize problems, it is advised to engineer the MTU on the ingress 6PE to 
core interface, consistent with the core MTU, so that ICMP 'Packet too big" 
can be sent by this router without these packets or the ICMP messages 
entering the core MPLS.

Thank you

Eric Levy-Abegnoli
On jeudi 9 Mars 2006 01:16, Fred Baker wrote:
> We have been asked to read and comment on
>
>    http://www.ietf.org/internet-drafts/draft-ooms-v6ops-bgp-
> tunnel-06.txt
>    "Connecting IPv6 Islands over IPv4 MPLS using IPv6 Provider Edge
> Routers
>    (6PE)", Jeremy De Clercq, 18-Jan-06, <draft-ooms-v6ops-bgp-
> tunnel-06.txt>
>
> which is about to go into IESG review. As you will recall, this has
> been round the block, and deserves speedy review. May I suggest that
> you get your comments in by 17 March at the latest. Reply to the
> authors copying idr; if you are not on the list, the working group
> chairs will deal with the moderation issues.