Re: Path MTU for draft-ietf-v6ops-mech-v2-00.txt

["Fred L. Templin" <ftemplin@IPRG.nokia.com>]

Hi Vlad,

Vladislav Yasevich wrote:
> Fred
>
> Fred L. Templin wrote:
>
> > Were there any technical objections to setting a larger non-dynamic
> > tunnel MTU of 1380?
>
> If you start mandating a default static MTU of 1380 on a configured
Perhaps I havn't made myself clear on this point, but I'm not
seeking to mandate a static MTU of 1380. What I am meaning to
suggest is something along the lines that the static MTU assignment
"MUST be no less than 1280 and SHOULD be no greater than 1380".

> tunnel, then you may very likely have interoperability problmes
> with implementations of rfc 2893 that only support static
> MTU of 1280.
On this point, it's not clear to me that RFC 2893 specifies a static
MTU of 1280. If you read section 3.2 of that document carefully, you
will see that 1280 is only mentioned within the context of using the
IPv4 path MTU to determine the tunnel MTU. The only text in RFC 2893
that speaks to a static MTU assignment is in the next-to-last paragraph
of section 3.2 where it says that a node may: "...avoid using the IPv4
Path MTU algorithm accross the tunnel and instead use the MTU of the
link layer (under IPv4)...". (But, in many cases, such a choice would
lead to excessive fragmentation).

> Since v6ops-mech-v2-00 draft is updating RFC 2893, lets try to
> keep it backwards compatible.  Neither drafts absolutely require
> PMTU Discovery and thus having different default MTUs may cause
> problems and will require reconfiguration.
As far as I can tell, it is OK for the MTU determination schemes
implemented by the tunnel endpoints to be asymmetric. For example,
if nodes A and B are the tunnel endpoints it could be the case that
A uses a static MTU assignment while B implements the dynamic mechanism.
The fact that A and B might be using different default MTUs does not
seem to me have any implications for interoperability.

Regards,

Fred
ftemplin@iprg.nokia.com

> -vlad
> ++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++
> Vladislav Yasevich        Tru64 UNIX Network Group
> Hewlett Packard         Tel: (603) 884-1079
> Nashua, NH 03062        ZKO3-3/T07