stig.venaas at uninett.no
Mon Aug 28 09:58:26 CEST 2006
Iljitsch van Beijnum wrote:
> On 25-aug-2006, at 14:52, Nick Hilliard wrote:
>>> PMTUD problems aren't caused by any particular MTU, but by the source
>>> host not seeing "packet too big" messages. If you remove this router
>>> and use one that allows for a bigger MTU the problem will simply
>>> resurface once packets hit another link with a smaller MTU.
>> This is true, but it's also true that having links with MTU < 1500 tends
>> to cause trouble. Of course, in a perfect world, you could run at any
>> MTU you cared for, but we don't live in that world.
> I would say that's true for IPv4, where having links with an MTU
> smaller than 1500 bytes causes so much trouble that you can't
> realistically do anything other than ignore the DF bit or have a 1500
> byte or larger MTU. But in IPv6, you can't clear the DF bit and there
> is still a lot tunneling going on, so both of those aren't options, and
> we have a chance to get it right so we should really kick people in the
> sensitive parts until they get their PMTUD working. It's not like
> getting PMTUD to work is so incredibly hard...
A common problem with tunneling is also that the two ends of the tunnel
are configured with different MTUs, which cause problems even if PMTUD
might work. I've seen this happen many times.
> There is one other option that may work to some degree in IPv6:
> rewriting the TCP MSS option. But I'm not aware of any gear that
> supports this for IPv6.
And it is ugly, so I don't really miss it :)
More information about the ipv6-ops