Re: IPv6 prefix lengths - how long?

[Philip Homburg <[e-mail]>]

>The RFC 4291 IPv6 addressing and the original stance on 64bit boundary was bas
>ed on the EUI64 format for the Mac generated FFFE bit stiff between the 3rd an
>d 4th octet of the station id or random station id.
>
>I agree that a 64bit station id was overkill and was made 20 years ago thinkin
>g futuristically that every physical device you can think of would have an IP 
>but now in reality large in my opinion that would never be used and really was
>ted valuable bits on a station Id that could have been used more wisely to exp
>and the prefix length.

With SLAAC, EUI64 has to some extent been replaced with pseudo-random 
numbers.

For that to work, the number of bits needs to be sufficient to avoid
collisions. The exact number of bits you need depends on the number of
addresses on a subnet and how much you care about the damage done by
collisions.

The problem is that we know that a 64 bit pseudo-random number is safe,
but there is no sensible way to set a lower bound. There certainly is
no consensus on a lower bound.

At the same time, we are not running out /64 prefixes. We have prefix
allocation problem. We have enough /64s to give every virtual subnet its
own /64.

So the obvious thing to do is to leave the 64 bit boundary in place for SLAAC.

If you really need longer prefixes, then just use static configuration or
DHCPv6.