Re: FW: 6PE Path MTU Discovery

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

Hi Roberta,
see comments in line

At 12:35 20/02/2007 +0100, Maglione Roberta wrote:
> Re-sending because I didn't see the original
> mail on the mailing list...
> Sorry for double posting.
>
> Roberta
>
> ------------------------------------------------------------------------
> ----
> Roberta Maglione
> Telecom Italia - Broadband Network Services Innovation
> Via G. Reiss Romoli, 274 - 10148 Torino - Italy
> Phone: +39 011 228 5007
> e-mail: roberta.maglione@telecomitalia.it
> ------------------------------------------------------------------------
> ------
>  
>  
> -----Original Message-----
> From: Invernizzi Fabrizio 
> Sent: Friday, February 16, 2007 10:40 AM
> To: 'v6ops@ops.ietf.org'
> Cc: Maglione Roberta
> Subject: 6PE Path MTU Discovery
>
> Hi all
>
> we are currently investiganting on the deployment impacts of the 
> 6PE
> solution as described in the recently published RFC4798 (Connecting
> IPv6
> Islands over IPv4 MPLS Using IPv6 Provider Edge Routers (6PE)) and
> we
> have some doubts about the IPv6 Path MTU discovery mechanism
> traversal
> of a MPLS cloud described in the RFC.
> In particular the section 3 says: 
> "... , routers in the IPv4 MPLS network have the option to
>    generate an ICMP "Packet Too Big" message using
> mechanisms as
>    described in section 2.3.2 "Tunneling Private Addresses
> through a
>    Public Backbone" of [RFC3032]."
>
> What is not clear to us is:
> 1- how can a P router, IPv4 only, generate an IPv6 icmp message 
> (for
> example, which source addres should it use?)

If the router is IPv4 only, it simply cannot. However, if it has IPv6
capability, but not IPv6 connectivity, then it is not very different from
a P router when it needs to generate an ICMP message to a VPN
destination. Does not know the destination; and does not have any
topologically  valid source address. For reaching the destination,
it tunnel the icmp message in mpls. For the source, it picks one among
the one it has, and "hope" the destination will not filter
it.
For 6PE, it's the same. The source address can be any IPv6 address in the
box, and if they are none, it can be a v4-mapped IPv6 address. None is
going to be topologically correct of course ...

> 2- does this imply enabling IPv6 on all the
> MPLS network? If yes, is
> correct to say that this would reduce the benefits of using the 6PE
> solution? (no changes required on the MPLS network and IPv6 enabled
> only
> on the "edge" routers, the 6PEs)

No, not at all. IPv6 does not need to be enabled. But P-routers need to
be capable of generating IPv6 icmp messages, and tunnel them per
RFC3032,
section 2.3.2 (?Tunnelling Private Addresses through a Public Backbone?)
Eric

> Would be interesting for this community to investigate a mechanism to
> make the 6PE routers able to dynamically discover the PMTU for each LSP
> to be used to transport the IPv6 packets through the MPLS network(may be
> leveraging on the native MPLS PMTU discovery mechanism?)
> If yes, we are available to cohoperate to produce a proposal on this
> topic.




> Thanks in advance
>
> Roberta and Fabrizio
> --------------------------------------------------------------------
>
> CONFIDENTIALITY NOTICE
>
> This message and its attachments are addressed solely to the persons above and may contain confidential information. If you have received the message in error, be informed that any use of the content hereof is prohibited. Please return it immediately to the sender and delete the message. Should you have any questions, please contact us by replying to webmaster@telecomitalia.it.
>
>         Thank you
>
>                                         www.telecomitalia.it
>
> --------------------------------------------------------------------
>