Re: [v6ops] BGP Identifier

[Nick Hilliard <[e-mail]>]

On 15/02/2014 15:26, Sander Steffann wrote:
> Op 15 feb. 2014, om 16:13 heeft Shane Amante <[e-mail]> het volgende geschreven:
>> On Feb 15, 2014, at 12:55 AM, Randy Bush <[e-mail]> wrote:
>>> I use this funny thing called DNS. 
>>
>> And that has what to do with the problem of determining liveness or determining where in the topology is a ROUTER_ID?
> 
> And what does an integer called ROUTER_ID tell you about that?
> 
> And hey, you can always create records like
>   1.2.3.4.router-id.castlepoint.net IN CNAME router1.somewhere.castlepoint.net

guys, we're all having a bit of an ietf moment here: within the space of 48
hours, the conversation has ratholed.  Let's love up a bit.

The authors of the draft have a simple operational desire to make their
lives a little easier and as an operator I sympathise with this,
particularly because they are using ipv6-only networks in anger which is
more than I do.

There are two advantages of having a 128 bit router-id (with the unstated
convention of tying router-id == address of first loopback):

- firstly it's easy to identify the location of prefix announcements on
your network
- secondly the router-id can be autoconfigured.

On the other hand, implementing this will require that the authors define a
transition mechanism to allow 128 bit router-id routers interoperate with
32 bit router-id routers in such a way that router-id collisions don't occur.

If the authors want to progress this, they need to state a stronger use
case and they need to create a transition mechanism.  If they can do this
in a reasonable way which doesn't break backwards compatibility, then it
might be appropriate for idr / v6ops / etc to take another look at the draft.

Also, I wish them well because everyone understands what a router ID is and
everyone will want to paint theirs a different colour.

Nick