On Jul 26, 2011, at 6:09 PM, Alexandru Petrescu wrote:
> Le 27/07/2011 00:05, Joel Jaeggli a écrit :
>> If you're commenting on v6nd draft I don't think we have any intention of
>> connecting ~1.8 * 10^19 hosts to a subnet...
>
> Yes, commenting on the v6nd draft.
>
> Listening to the disxcussion it so sounde to me that these were ND
> optimizations such as to support 2^64 nodes in a subnet.
goal is:
to retain entries for things that we know under duress
to be able to learn new addresses when that router is resource
constrained by prioritizing activities.
the implicit assumption is that memory is a limited commodity and there are
higher priority things for the control plane processor to be doing instead of
discovering hosts that don't exist due to incoming unicast packets...
> Alex
>
>>
>>
>> On Jul 26, 2011, at 5:10 PM, Alexandru Petrescu wrote:
>>
>>> 2^64 is 10s of trillions (not just trillions) and only for anglo-saxons.
>>>
>>> 2^64 is exabytes of memory (not giga, nor tera, nor peta).
>>>
>>> I am not sure you look at adapting nd for these figures.
>>>
>>> Alex
>>> --------------------------------------------------------------------
>>> IETF IPv6 working group mailing list
>>> [email protected]
>>> Administrative Requests: https://www.ietf.org/mailman/listinfo/ipv6
>>> --------------------------------------------------------------------
>>>
>>
>
--------------------------------------------------------------------
IETF IPv6 working group mailing list
[email protected]
Administrative Requests: https://www.ietf.org/mailman/listinfo/ipv6
--------------------------------------------------------------------
