[prev in list] [next in list] [prev in thread] [next in thread] 

List:       ipng
Subject:    Re: IID length text [was Re: Review of draft-ietf-6man-rfc4291bis-06]
From:       Tore Anderson <tore () fud ! no>
Date:       2017-01-23 14:07:43
Message-ID: 20170123150743.6960611b () envy ! e1 ! y ! home
[Download RAW message or body]

* Lorenzo Colitti

> On Mon, Jan 23, 2017 at 8:49 PM, Tore Anderson <tore@fud.no> wrote:
> 
> > Isn't there a conflict with RFC 6052, though?
> >
> > E.g., if you're using an RFC 6052 NSP such as 2001:db8:6052::/96,
> > the IPv4-translatable representation of an IPv4 subnet such as
> > 192.0.2.0/24 would be 2001:db8:6052::192.0.2.0/120
> > (2001:db8:6052::c000:200/120) and nodes in this subnet/link would
> > necessarily have 8 bit long IIDs. Right?
> 
> I'm not sure it makes sense to draw that conclusion.
> 
> If you follow that reasoning, /96 might make sense, but all the other
> RFC 6052 prefix lengths don't really make sense. For example, if the
> NSP is 64 bits, how long would the IID be? If you say 8 bits, then
> that means there are 2^40 addresses routed to the same host, which
> means that the prefix length (88) plus the IID length (8) does not
> add up to 128. And if you say 48 bits, then that means that a host
> has 2^40 interface IDs for the same interface

So with an NSP of 2001:db8:6052::/64, the IPv4 link subnet of
192.0.2.0/24 would be as I understand it be represented as
2001:db8:6052:0:c0:2::/96.

In the  «vanilla » RFC7915 operational mode, the default gateway on that
link (a.k.a. 192.0.2.1 from the IPv4 point of view) would be configured
with 2001:db8:6052:0:c0:2:100:0/96, the node reachable at 192.0.2.10
would have 2001:db8:6052:0:c0:2:a00:0/96, 192.0.2.254 would be
2001:db8:6052:0:c0:2:fe00:0/96, and so on.

The IPv6 addresses you actually need to configure on the routers and
the nodes on this link will have an IPv6 prefix length of /96, so
wouldn't that mean that the IIDs in question are 32 bits long?

> the interface ID isn't really an ID any more.

I suppose you can look at it that way too...but wouldn't that then mean
that you're essentially saying that you can have whatever amount of
suffix/node/host bits in an IPv6 address you'd like, but you're only
allowed to call it an  «IID » if it just so happens to be exactly 64 of
them? :-)

Tore

--------------------------------------------------------------------
IETF IPv6 working group mailing list
ipv6@ietf.org
Administrative Requests: https://www.ietf.org/mailman/listinfo/ipv6
--------------------------------------------------------------------

[prev in list] [next in list] [prev in thread] [next in thread]