RE: what problem is solved by proscribing non-64 bit prefixes?

[Pekka Savola <pekkas at netcore.fi>]

On Thu, 2 Oct 2008, Dunn, Jeffrey H. wrote:
> While I agree with your assertions concerning flexibility and
> robustness, I do not agree that 2^64 is the minimum number of nodes
> that should be supported on a link.  Consider current IPv4 deployments.
> I doubt anyone have configured a single router interface with any
> prefix shorter than /8 (24 bits for hosts).  As a result, it would seem
> to me to be a reasonable and conservative to assume that one might want
> to use an IPv6 /96 (32 bits for hosts).  Regardless of that, I see no
> reason to constraint folks from using other numbers of bits, both more
> and less than 64.  The bottom line is that I do not understand why
> folks feel the need to constrain network architects and administrators
> to use only 64 bit prefixes.

Back then, folks felt that mapping an Ethernet MAC address 
(one-to-one) was a good idea.  That requires 48 bits.  Initially, the 
interface identifier part was 48 bits (RFC1884), but that was thought 
inadequate and it was expanded to 64, probably because other media 
needed more than 48 and bits were there.  I don't think this design 
decision is worth revisiting.

But no one will force you to use this numbering in your own network if 
you don't want to.  It just might be that some protocols don't work 
quite as well.

> > You're not reserving a block of 2^12 IPv4 prefixes (or addresses in
> > degenerate case) for each VLAN-capable router interface today, so I
> > fail to see how this argument would apply to IPv6.
>
> That is because you did not read the initial paragraph, which points
> out that SLAAC requires a one-to-one mapping of prefixes to links,
> e.g., Ethernet broadcast domains.  In addition, OSPFv3 also forbids
> prefixes from spanning interfaces.  As a result, 2^13 prefixes could be
> required if one wanted to be sure that all 2^13 VLAN identifiers were
> available to a given physical interface.  This fits in with your
> statement about the ability to design new protocols and mechanisms to
> leverage address bits.

I read it, and read it again.  Again, I ask, how do you solve this 
problem with IPv4?  I'm not seeing it.  IPv4 has (AFAIK) the same 
constraints except that IPv6 has link-local addresses and you actually 
don't need any prefix at all unless you attach hosts to the link.

--
Pekka Savola                 "You each name yourselves king, yet the
Netcore Oy                    kingdom bleeds."
Systems. Networks. Security. -- George R.R. Martin: A Clash of Kings
--------------------------------------------------------------------
IETF IPv6 working group mailing list
ipv6 at ietf.org
Administrative Requests: https://www.ietf.org/mailman/listinfo/ipv6
--------------------------------------------------------------------