Re: [v6ops] question about draft-jjmb-v6ops-unique-ipv6-prefix-per-host

[Brian E Carpenter <[e-mail]>]

On 05/07/2017 08:35, Ole Troan wrote:
> Brian,
> 
>>> Since we have to solve the problem for multiple links anyway, does it then make a difference that logical topology follows physical topology and each host is on its own link?
>>
>> If the discovery traffic is a very small proportion of total traffic, it probably doesn't matter exactly which solution you adopt. But (as Tim implies for DNSSD), I think you should use link-local multicast when it's available. Are we disagreeing?
> 
> Are you implying there are links where link-local multicast aren't available?

I'm not aware of any, but do we actually have a standard that makes it mandatory?

> If course on the link type I describe there is ever only two nodes connected. 

Sure, I understood that.
 
> If we can't do better than require every link to have a service discovery "helper" then sure link-local multicast is fine ( benefit is well known addresses).

Yes. I think it's unavoidable. Certainly our intention in ANIMA is that discovery must work on any topology without prior configuration, and that requires every node that has more than one interface to automatically deploy a discovery relay between its interfaces. (And yes, that in turn requires loop prevention.)

  Brian