Re: A proposal for draft-ietf-6man-rfc4291bis-07

[David Farmer <[e-mail]>]

On Thu, Mar 2, 2017 at 8:55 PM, Manfredi, Albert E <
[e-mail]> wrote:

> From: David Farmer [mailto:[e-mail]]
>
> > Because, it doesn't say IIDs are themselves required.
>
> I see this as being a semantic game. As long as an address consists of a
> prefix, and it has to in order for the packet to be forwarded "the usual
> way" across routers, the remaining bits will be an ID pointing to a
> specific interface. I don’t think we are suggesting here that any prefix
> greater than 64 bits MUST result in a flat network. Or at least, I'm not.
>
> > In fact it says "at a minimum, a node may consider that unicast
> > addresses (including its own) have no internal structure".
>
> As a minimum, sure, that's always the case. A node might possibly assume
> this, as a minimum. "As a minimum" the rule can apply to any IPv6 node,
> even /64 nodes. A router, for instance, might be configured to only route
> ten 128-bit addresses, each one to a particular egress interface, and
> blindly send the rest over an 11th egress interface, for all other traffic.
> That works fine, as long as the number of 128-bit addresses doesn't grow
> too much.
>
> > And I'd now suggest it should say, if you use an IID it must be
> > 64 bits unless overridden my an IPv6-over-foo spec.
>
> But this would make "as a minimum" phrase lose all its meaning? Now it
> says that this flat routing scheme must be the only one allowable, for
> prefixes > 64. So I don’t think that's right.
>
> Bert
>

Well I'll agree it's subtle, it needs to be explained better in RFC4291bis,
and I probably haven't hit the mark yet, but it's more than just
semantics.  If I configure an interface as 2001:DB8::1/64, that is saying
that the interface has 128 bit address and a subnet length of 64 bits.
It's not saying the subnet prefix is 2001:DB8::0/64 and the IID is 1 with
an IID length of 64.

Similarly, if I configured my interface with 2001:DB8::1/120, that is
saying that the interface has 128 bit address and a subnet length of 120
bits. It's not saying the subnet prefix is 2001:DB8::0/120 and the IID is 1
with an IID length of 8.

Now I could also configure the address of an interface as 2001:DB8::1 with
no prefix length, it then should look for an RA for that link and if there
is a PIO for prefix 2001:DB8::0, and lets say it has a prefix length of 120
and with an "L" flag and no "A" flag.  In my opinion that should work and
communicate with another host on the link with an interface configured
similarly, with say 2001:DB8::2.  Furthermore, if another host on that link
had an interface configure with 2001:DB8::1001, they shouldn't communicate,
as that address isn't on link by that PIO.

And in all of those cases, 2001:DB8::1/64, 2001:DB8::1/120, or 2001:DB8::1,
2001:DB8::2, 2001:DB8::1001 with no prefix length, there is also a link
local address with an IID of length of 64 and a prefix of FE80::0/64 and
was auto generated with SLAAC.

And stateful DHCPv6 doesn't provide an IID either, it provides a full 128
bit IPv6 address, but doesn't provide a prefix length or a router, that is
learned from the RA and it's PIOs.  I agree with Randy that's another thing
to fix, but that's a fight for a different day, and completely independent
of this.

So saying that all IIDs are required to be 64 bits, doesn't actually imply
all subnet prefixes are /64.  It only says you can't generate an IPv6
address using an IID on a network with a subnet prefix other than /64, and
you probably can't (or at least shouldn't) set the "A" flag on PIO for
prefix other than /64. However, can manually configure or get an address
via DHCPv6 for subnet prefix other than /64.  This why we have to correct
RFC4291bis by recommending /64 subnet prefixes, because it's ambiguous now,
and would be much cleaner if it was clearly recommended.

Thanks.




-- 
===============================================
David Farmer               Email:[e-mail]
Networking & Telecommunication Services
Office of Information Technology
University of Minnesota
2218 University Ave SE        Phone: [phone] <(612)%20626-0815>
Minneapolis, MN 55414-3029   Cell: 612-812-9952 <(612)%20812-9952>
===============================================