Re: Question on anycast IID range(s)

[Kerry Lynn <[e-mail]>]

On Thu, Jan 3, 2019 at 10:55 AM Ole Troan <[e-mail]> wrote:

> Kerry,
>
> > For practical purposes, particularly in light of RFC 7136, should one
> > consider an anycast address to be any that ends in dfff:ffff:ffff:ff80-
> > dfff:ffff:ffff:ffff OR ffff:ffff:ffff:ff80-ffff:ffff:ffff:ffff?
> >
> > The phrase in RFC 2526 that's causing confusion for me is
> > "Specifically, for IPv6 address types required to have to have
> > [sic] 64-bit interface identifiers in EUI-64 format ..."  To my
> > knowledge, there are address types that require a 64-bit IID,
> > but it seems we've been systematically trying to deprecate
> > *EUI-64 format* IIDs.  In any case, there's nothing to prevent a
> > mix of EUI-64 or *other* format IIDs in the same subnet as far
> > as I'm aware.
>
> The main point of 2526 is:
>
>    "Within each subnet, the highest 128 interface identifier values are
>    reserved for assignment as subnet anycast addresses.”
>
> Is that definition not unambigiuous?
>
> It's ambiguous if 254 IID values meet the definitions further down in
RFC2526
or for prefix > /120

I have to credit my colleague Mike Zhukovskiy for bringing this problem to
our
attention.  When he read RFC2526, he didn't get why the quote above applied
to the range fdff:ffff:ffff:ff80-fdff:ffff:ffff:ffff.  I then told him
about the "u" bit,
RFC7136, and eventually ended up confused myself as I re-read RFC2526.

Thinking about this overnight, I'm not sure that adding an errata to RFC5453
to simply include the range ffff:ffff:ffff:ff80-ffff:ffff:ffff:ffff is the
correct solution;
we'd probably need to widen its scope to include all 64-bit IIDs.

I think it's clear from re-reading RFC2526 that the range
ffff:ffff:ffff:ff80-ffff:ffff:ffff:ffff
COULD be the reserved range, even for /64 prefixes.  Mike points out at
least
one widely-deployed implementation that assumes this:
https://community.cisco.com/t5/ipv6/reserved-anycast-address/td-p/2765408

OTOH, let's say you actually want to deploy Mobile IPv6 home-agents.  Should
you advertise this service at <prefix>:fdff:ffff:ffff:fffe,
<prefix>:ffff:ffff:ffff:fffe, or both?
I agree with Suresh that we may need a more considered solution.

Regards, Kerry

Cheers,
> Ole
>
>
>