Re: [BEHAVE] Handling fragments in the stateful NAT64

I agree that 10 seconds in Internet today is a long time for fragments to
arrive. Should we optimize for the common case or the worst case?

Although there will always be exceptions as mention by Iljitschv, providers
should be allowed to optimize for the their common case. So, I would suggest
to include text to say that the the interval MAY be configurable.

Thanks,

Reinaldo

On 12/1/09 9:50 AM, "Iljitsch van Beijnum" <iljitsch@muada.com> wrote:

> On 1 dec 2009, at 10:41, marcelo bagnulo braun wrote:
> 
>>  Assuming it otherwise has sufficient resources, a NAT64 MUST allow
>>  the fragments to arrive over a time interval of at least 10 seconds.
> 
> The first clause makes the requirement meaningless. Someone could always claim
> lack of resources.
> 
> Also, is this 10 seconds between all fragments (so 30 seconds in the case of 4
> fragments) or 10 seconds from the first fragment to arrive until the last?
> 
> I'm not very happy about the 10 seconds, in practice all fragments will arrive
> within a few hundred milliseconds. But I guess if we assume a worst case
> network speed of 8 kbps (really slow GPRS link) and packet size (64k) then it
> can take more than a minute for all fragments to arrive...
> _______________________________________________
> Behave mailing list
> Behave@ietf.org
> https://www.ietf.org/mailman/listinfo/behave