Brian E Carpenter <brian.e.carpen...@gmail.com> writes: > Jared,
> On 2010-08-16 13:06, Jared Mauch wrote: > ... > > Is there a legitimate operational reason a host should not know > > the subnet length it sits on? A host should not be *required* to know the subnet length. Very simple devices may have "simple" stacks. And indeed, for a simply host, all you need to know is a first hop router address. Redirects can do the rest. This may not be the ideal situation, but it is one the architecure allows for. It is also fine for more sophisticated implementations to know what the subnet length is. > It needs to know, which is why this is part of the basic information > in RAs. Um, trying to run without RAs won't get you very far in IPv6. > The question seems to be whether the information should be repeated > in DHCP. My conclusion is now yes. The obvious scenario is one that is in use today in IPv4. Hosts learn router addresses via DHCP. They also learn on-link prefix info from DHCP. This is essentially static information and does not change. The network infrastructure (i.e., the first hop routers) all run VRRP and handle faults and failovers without hosts having to do anything. Works fine in IPv4 and is an appropriate choice for some (emphasis on *some*) environments. IPv6 should allow for the same functionality, to ease transision to IPv6 for such networks. I plan to resurrect an old DHC ID on this topic, as I view this is a gap of the form "why can't I do what I do in IPv4 today in IPv6". Thomas -------------------------------------------------------------------- IETF IPv6 working group mailing list ipv6@ietf.org Administrative Requests: https://www.ietf.org/mailman/listinfo/ipv6 --------------------------------------------------------------------