RE: [nesg] RE: status update for ESnet

> Mike Collins wrote:
> Michel:
> Ok, I see your point now.  You are correct we do not
> aggregate the loopback addresses.

This configuration is clever and has some increased operational
robustness that IMHO more than balances the increase in routing table
size.

The pros and cons are:
Pros: even if a bozo messes up routing, you are likely to still be able
to access the router without jumping hop-per-hop.
Cons: one /128 route in the routing table per router.

Michel.


>Mike / Ole
>
>> Mike Collins wrote:
>> The simple answer is that 128-bit prefixes are required for the
>> loopback address in a Juniper router. Host address are required
>> for the loopback for both IPv4 and IPv6.
>
>Got it, thanks.
>
>
>> Ole Troan wrote:
>> both /64 and /128 can be aggregated. what is your point?
>
>With regard to what Bob Fink posted quoted below:
>
>> Interfaces addresses are globally unique and are /64s.
>> Loop back addresses are globally unique /128s from a common /64.
>
>The way I understand this is:
>[I just made up these numbers, good enough for example]
>Let's say you have been assigned 2001:xxxx::/32
>Out of this, you reserve 2001:xxxx:FFFF:FFFE::/64 for loopbacks,
>And you assign
>Router 1's loopback: 2001:xxxx:FFFF:FFFE::1/128
>Router 2's loopback: 2001:xxxx:FFFF:FFFE::2/128
>Router 3's loopback: 2001:xxxx:FFFF:FFFE::3/128
>
>If what is above is not correct, I find the "from a common /64" part
>confusing.
>
>If what is above is indeed correct, you have to explain me how you
>aggregate these loopback addresses.
>
>Back to your 2001:xxxx::/32, you might want to do something like:
>
>2001:xxxx::/40 Illinois
>  2001:xxxx:0001::/48 Chicago
>2001:xxxx:100::/40 California
>  2001:xxxx:101::/48 Berkeley
>  2001:xxxx:102::/48 Livermore
>
>Then there is no way router 1's loopback can be part of the Chicago
>block, and router 2's loopback address part of the Berkeley block.
>
>Michel.
>