Re: [DNA] Prefix information for link identification in DNA

[Erik Nordmark <erik.nordmark@sun.com>]

Greg Daley wrote:

> I'd like to know how you think that relates to DNA for IPv6 though.

Good question.

For IPv6 we've been assuming that
  - we have global IPv6 prefixes (thus no confusion with 192.168.0.0/255 
being used in multiple places)
  - Router Solicitation/Advertisement being reasonably efficient (more 
efficient than sending to a DHCP server and getting a reply)

As a result of this, we can do better than IPv4 in that we can use a 
RS/RA exchange to both verify whether we are on the same link, and also 
get the new configuration information if we have moved to a different link.

So it doesn't seem that useful to follow the IPv4 model of first 
verifying that an IP+MAC address is still present on the link, since if 
we have moved we'd need to send additional packets.


*If* we want DNAv6 to handle the case when we don't have global IPv6 
prefixes, I'd suggest we do that by allowing something other than 
prefixes be used as the link identifier (as Keith Moore suggested).
This would still assume that something is sending out router 
advertisements; but a node can send a router advertisement without 
claiming to be a default router so this doesn't actually assume that 
there is a a router. One just need one or more nodes that will send out 
the RAs.

    Erik