Re: Node Requirements: Elevating DHCPv6 from MAY to SHOULD

[Markus Hanauska <[e-mail]>]

On 2011-05-31, at 11:19 , Philip Homburg wrote:

> No, ND is more clever than that. All traffic between prefixes that are on-link
> goes directly between the hosts. Even when the prefix is off-link it is
> possible for the router the send a redirect ICMP to cause further traffic
> to be directly between the hosts.

I don't think this related to ND, is it? ICMP redirects also exist for IPv4 and IPv4 doesn't know ND. I think only difference is that ICPMv6 optionally allows an link layer address in the redirect message. How good IPv6 implementations really support ICMPv6 redirects as of today is another question.

My main problem with that approach is, that not everyone has a $5000++ Cisco router available and the configuration capabilities of some more inexpensive routers are quite limited; especially regarding IPv6, which is still not mainstream (the majority of routers on the market still has no IPv6 support at all). Also what are you going to do, if your ISP only gives you a single /64 prefix? To subdivide it, you have to use some bits for subnetworking, which means your hosts might only have /48 addresses and that disables SLAAC entirely.

I still think it would have been much easier, to define a second bit in the /64 address space. Just like the 'u' bit, which says that an address is globally unique or not, there would be the 'a' bit, which is set if the address has been assigned "automatically" (SLAAC w/ or w/o Privacy Extension) and not set, if this address has been "manually" assigned (either by manual config on the node or by DHCP or anything comparable to DHCP). This will effectively eliminate collisions, except for MAC address collisions or collisions caused by misconfiguration of manual IPs and/or DHCP.

Regards,
Markus