Re: [homenet] HNCP

[Mikael Abrahamsson <[e-mail]>]

On Thu, 13 Feb 2014, Mark Townsley wrote:

>> Drafts:
>>   http://tools.ietf.org/html/draft-stenberg-homenet-hncp-00
>>   http://tools.ietf.org/html/draft-stenberg-homenet-dnssd-hybrid-proxy-zeroconf-00
>>   http://tools.ietf.org/html/draft-pfister-homenet-prefix-assignment-00

I have read all of the drafts. I still need to wrap my head around all the 
details, but I think this is excellent work. It solves most if not all the 
areas of concern within the home when it comes to service discovery, 
prefix assignment etc.

I am a bit confused about the disposition of the drafts, but I guess it 
can be carved a lot of different ways, so I don't see this as a problem 
really.

The choice to invent a new transport protocol is a concern, however as 
things should be fairly modular, let's focus on the mechanisms and 
architecture, and as far as I can tell, this looks fine. If I understood 
correctly, there is running code for most of this so it has actually been 
tested? I think this is extremely good if it has, because getting all of 
these mechanisms right is the important thing, then we can discuss 
transport as a separate issue.

As I am a router guy, redistribution between protocols is done all the 
time. It's hairy when you do redistribution between OSPF->BGP->OSPF and 
BGP->OSPF->BGP etc. There you have some identifier that helps you to stop 
redistribution loops. Do we need something like this here? If we decide 
that HNCP is not what we want long term and we want to migrate to 
something else, there might be the need for redistribution between old and 
now. How can we prepare for something like this so it's as least 
distruptive as possible?

I was thinking of failover. What happens with a LAN with 1 DR and 3 other 
routers, and I turn off the DR? There is a prefix, there might be some 
IA_NA handed out by the DR, and there is a lifetime of the prefixes etc. 
I can't find the relevant section in the prefix assignment draft that 
handles this (or it's implicit somewhere and I just missed it). Anyhow, I 
would like to see a whole new section on this topic that describes what 
happens in different scenarios with pointers to desired behavior, as 
described in other sections of the text.

-- 
Mikael Abrahamsson    email: [e-mail]