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

["Dunn, Jeffrey H." <jdunn at mitre.org>]

Brian,

Your point about raising the size of the address space to the fourth
power is not quite correct.  If we constrain the network prefix to be
64-bits, then, assuming that the longest routable IPv4 prefix is 24, we
have only raised the prefix size by the power 2.67.  Now assume that
we start handing out /28 prefixes to ISPs for their residential
customers (who are now all behind a NAT), we have now only raised the
number of subnets to the 1.17. My argument is that it is not the number
of unique end system addresses that is the issue, rather the number of
subnets.

As to your observation concerning this generation's engineers, I whole
heartedly second your suggestion.

Best Regards, 
  
Jeffrey Dunn 
Info Systems Eng., Lead 
MITRE Corporation.
(301) 448-6965 (mobile)


-----Original Message-----
From: Brian E Carpenter [mailto:brian.e.carpenter at gmail.com] 
Sent: Wednesday, October 01, 2008 4:43 PM
To: Dunn, Jeffrey H.
Cc: Alexandru Petrescu; michael.dillon at bt.com; Sherman, Kurt T.;
ipv6 at ietf.org; rbonica at juniper.net; Steve_Eiserman at ao.uscourts.gov;
Pasi.Eronen at nokia.com; draft-ietf-v6ops-addcon at tools.ietf.org;
ralph.liguori at disa.mil; night at nist.gov; dougm at nist.gov;
v6ops-chairs at tools.ietf.org; Martin, Cynthia E.
Subject: Re: what problem is solved by proscribing non-64 bit prefixes?

On 2008-10-02 02:04, Dunn, Jeffrey H. wrote:
> More to the point, what would a individual household do with
Avogadro's
> number worth of IPv6 addresses (2^80 = 1.2x10^24)?  This seems
> extremely wasteful.  Further, a reasonable sized ISP with a couple of
> million customers would require a /28 or more just for their
> residential customer base.  This sounds like a prescription for
address
> exhaustion.

Not in the least. Please remember that we have raised the size of the
address space to the *fourth* power; we squared it and squared it
again.
Even if we'd stuck to the original plan of assigning /48s everywhere,
that means we've *squared* the size of the subnet prefix space (/24 to
/48).
Unless really stupid allocation mechanisms are allowed, that is enough
for any imaginable future. And all the evidence is that the RIRs are
being extremely conservative in their practices for allocation. So
I don't see even a remote cause for concern.

On the other hand I see enormous value in allowing any size of network
on any customer's premises. It's not for this generation of engineers
and ISPs to constrain what our great-grandchildren might invent.

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