Re: [BEHAVE] Amount of fragmentation resulting from translation

[Brian E Carpenter <brian.e.carpenter at gmail.com>]

On 2009-10-31 02:20, Mark Andrews wrote:
> In message <4AEAA15C.4060900 at it.uc3m.es>, marcelo bagnulo braun writes:
>> Mark Andrews escribi�:
>>>
>>> I think turning on PMTUD in the middle of the network is just plain
>>> wrong.  There are OS which incorrectly turn on PMTUD for IPv4 for
>>> all packets.  This causes problems for applications like DNS servers.
>>>
>>> There is only one choice and that is to fragment at network MTU.  Anything
>>> else will cause operational problems.
>>>
>>>   
>> Just to be clear, you support the option that the translator always 
>> sends IPv6 packet of 1280 bytes rahter than the translator doing pmtud 
>> to figure out which is the mtu availbale for that path, correct?
>>
>> Regards, marcelo
> 
> Yes.

BTW, the MTU data observed in draft-defeche-ipv6-traffic-in-p2p-networks
strongly support this for today's IPv6 world. Average observed IPv4
MTU was 1498 bytes and average observed IPv6 MTU was 1289, obviously
telling us that 1280 is the dominant value in practice (at least
for hosts running BitTorrent). So it's not only the safe value to use,
but it's also not very sub-optimal.

   Brian