[c-nsp] IPv6 subnets for point-to-point links

[Mohacsi Janos]

On Mon, 9 May 2005, Crist Clark wrote:

> Gert Doering wrote:
>> Hi,
>>
>> On Mon, May 09, 2005 at 01:19:29PM +0200, Frotzler, Florian wrote:
>>
>>> To my knowledge there is mostly a consensus within the IETF (and Ripe
>>> :-) IPv6 community to not use /127 for various reasons. Personally I
>>> agree to that, whether or not this has a technical background, I would
>>> never use a /31 in the IPv4 world so why go for a /127.
>>
>>
>> In the IPv4 world, /31 is "the way to go" :-)
>>
>> (Just as a side note...)
>
> In the IP world, or any world for that matter, why does one need to
> assign a netmask to a point-to-point link at all? A point-to-point
> link is, by definition... uh, point-to-point. Netmasks are for broadcast
> media.
>
> The need for netmasks on a point-to-point link only happens when you
> pretend a broadcast link is point-to-point.
>
> Please enlighten me if I'm missing something here.
> --

Yes I agree with you. The point-to-point link does not need IP addresses 
for run them. It is useful only for debugging: pinging the other end.

With IPv6 you don't need to assign global addresses to point-to-point link 
, even if it is implemented with Ethernet type of media. With IPv6 you can 
rely on the automatically assigned link-local addresses for pinging you 
end point. Link-local is enough for running OSPFv3. You can rely on 
loopback or some service interface address for traceroute. This is working 
for us for more than 3 years now.

Regards,


Janos Mohacsi
Network Engineer, Research Associate
NIIF/HUNGARNET, HUNGARY
Key 00F9AF98: 8645 1312 D249 471B DBAE  21A2 9F52 0D1F 00F9 AF98