FYI. I've made some progress with my implementation. I held a presentation on 
this at RIPE this week:

http://thingsonip.blogspot.com/2012/04/home-networks-by-magic.html

There are a couple of relevant points to raise -- implementation and usage 
experience always brings up things that one did not think of when writing 
specifications. Here are some high-level ones:

It is surprising how many connections an autoconfiguration software (running 
inside OSPFv3 in this case) has to the rest of the system. Sources of prefixes. 
Users of prefixes. Etc.

There may be a need for system APIs that allow various system components to 
work together with the autoconfiguration software.

The implementation effort made me think about DNS in a completely different way 
than what we had been discussing in the WG before. The WG had focused on name 
resolution Bonjour-style within the entire home vs. a single link. While that 
is useful, name resolution for the Internet seemed an even more important 
topic, and getting that autoconfigured is not trivial (unless you assume IPv4 
name server use in a dual stack setting). I still don't know what exactly the 
right approach is, but I think we need to pay more attention to this than we 
have so far. Or at least I have.

One of aspect of this is to ensure DNS discovery works.

IPv6-only and NAT64-networks require special care. If you have a NAT64 device, 
behind that device you want to use its DNS64 but elsewhere you do not.

Timing of various events needs thought. If you come up, how long do you wait 
until you generate ULA space if you don't hear from others or if the others 
have no global space either?

Jari

_______________________________________________
homenet mailing list
[email protected]
https://www.ietf.org/mailman/listinfo/homenet

Reply via email to