draft-ietf-6man-prefixlen-p2p-01 clarification

[Miya Kohno]

Hi Michael,

> But if subnet-router anycast is turned of, should that matter?  

As for the actual behavior, that should not be a matter.
It's for the sake of double check. 

> RFC 4291 only states that the zeroth host bits
> *of a given prefixlen* not be used (i.e. the last 128-prefixlen bits
> cannot be zero).

Yes, but in the section 2.5.4, it says:
 All Global Unicast addresses other than those that start with binary
 000 have a 64-bit interface ID field

So some stereotyped implementations could automatically think that the
prefixlen should be 64 bit and so all zeros in the rightmost 64 bits
should mean subnet-router anycast. I think it's not likely, but just to
be safe.

Hope this clarifies.

Miya
> -----Original Message-----
> From: ipv6-ops-bounces+mkohno=juniper.net at lists.cluenet.de
[mailto:ipv6-
> ops-bounces+mkohno=juniper.net at lists.cluenet.de] On Behalf Of Michael
> Sinatra
> Sent: Wednesday, March 23, 2011 9:16 AM
> To: ipv6-ops at lists.cluenet.de
> Subject: draft-ietf-6man-prefixlen-p2p-01 clarification
> 
> Hi,
> 
> I am trying to properly interpret draft-ietf-6man-prefixlen-p2p-01
> (http://tools.ietf.org/html/draft-ietf-6man-prefixlen-p2p-01), which I
> believe has been advanced to proposed standard.  In section 6 of the
> draft, it says:
> 
>     Routers MUST support the assignment of /127 prefixes on point-to-
>     point inter-router links.  Routers MUST disable subnet-router
anycast
>     for the prefix when /127 prefixes are used.
> 
>     When assigning and using any /127 prefixes, the following
>     considerations apply.  Some addresses have special meanings, in
>     particular addresses corresponding to reserved anycast addresses.
>     When assigning prefixes (and addresses) to links, care should be
>     taken to ensure that addresses reserved for such purposes aren't
>     inadvertently assigned and used as unicast addresses.  Otherwise,
>     nodes may receive packets that they are not intended to receive.
>     Specifically, assuming that a number of point-to-point links will
be
>     numbered out of a single /64 prefix:
> 
>        a) Addresses with all zeros in the rightmost 64 bits SHOULD NOT
be
>        assigned as unicast addresses, to avoid colliding with the
Subnet-
>        Router anycast address [RFC4291].
> 
> 
> Okay, so I understand the point about the routers must turn off
> subnet-router anycast; otherwise /127s could never be used.  However,
it
> appears that the second paragraph is trying to recommend what prefix
the
> operator assigns as a /127 point-to-point link.  If that's the case,
> then it further seems that "a)" requires that I not assign the zeroth
> /127 prefix carved out of a /64.  But if subnet-router anycast is
turned
> of, should that matter?  RFC 4291 only states that the zeroth host
bits
> *of a given prefixlen* not be used (i.e. the last 128-prefixlen bits
> cannot be zero).  Of course, if that's the case, then no /127 would
> work, which is why we need section 6 of the draft as quoted above.  So
> is it really the case that we can't assign the zeroth /127 out of a
> given /64?
> 
> This certainly doesn't matter from an addressing point of view, as
there
> are way too many /127s in a /64 for this to be a remote worry.
However,
> I have heard people say that you can't use the zeroth /127 and I have
> experienced other people say "Why the heck are you skipping the first
> /127 and instead using 2001:db8:1:2::2/127 (or whatever)?"  So it's
more
> of a confusion thing for me.  Any interpretations of what the draft is
> trying to say?
> 
> thanks,
> michael