Re: router behaviour with prefixes longer than /64

[Peter Dordal <[e-mail]>]

On 03/17/2014 02:17 PM, David Farmer wrote:

On 3/17/14, 12:45 , Manfredi, Albert E wrote:

From: Lorenzo Colitti [mailto:[e-mail]]

I didn't miss that. People seem to be reading this clause backwards.

I'm saying that "those that start with binary value 000" are the vast
majority of potential IPv6 global unicast addresses.

No, the way to read it is exactly the way it's written: for *all*
unicast addresses, except those that start with 0 or 1, IIDs are 64
bits long, period.

Are you disagreeing that unicast addresses starting with 000 are the vast majority of potential future unicast addresses?

The only currently defined IPv6 addresses that start with the binary value 000, are special-purpose addresses, all other addresses that meet this definition are reserved, and not currently available for allocation, though they are reserved for future unicast allocation.

My understanding of the 0::/3 block has always been that it represents a fallback
in case 64-bit prefixes don't work out for some reason. This address block is "held in reserve".
That is, if we run out of /64 IPv6 address prefixes, we can always divide up the 0::/3 block using, say, /96 prefixes.
(And then at that point figure out how to manage with 32-bit interface IDs.)

...
There has never been any general-purpose unicast addresses that start with the binary value 000, only the current and historic special-purpose addresses.

But this could change in the future, though that would be likely only if there were a problem with the 64/64 division.
Which is to say not very likely at all.

Peter Dordal
Loyola Univ Chicago CS Dept