Hello all,
Proposal:
IPv6 unicast routing is based on prefixes of any valid length up to
128 bits [BCP198]. However, as explained in [RFC7421], the Interface ID
of unicast addresses is generally required to be 64 bits in length, with
exceptions only provided in special cases where expressly recognised
in IETF standards track documents.
The only place in IPv6 standards where the 64 boundary is mandated is on the
2000::/3 rule.
There is not a single implementation or deployment of that rule. There is no
instance where
trying to configure a prefix of length different than 64 will work when the
prefix is within 2000::/3, and fail when it is not.
That is a sufficient reason to not include this rule as part of the standard
track.
Nevertheless, SLAAC over Ethernet links indeed requires prefix length of 64.
As documented in the RFC7421, there are a few examples of protocols which
require 64 bits long interface identifiers.
SLAAC is not the only way to configure IPv6 hosts though.
Manual configuration, automated configuration, and DHCP are all IETF approved
ways of configuring IPv6 hosts.
They all work with prefixes of various lengths.
Does 'special cases' in the proposed text covers manual/automated configuration
or DHCP ? If so, it is far from being clear.
You obviously should use /64s if you want SLAAC to work, as well as other
protocols which are documented in RFC7421,
but /64 is not the rule, it is an architectural consideration that you have to
take into account when you design your network depending on the protocols that
you want to run there.
Therefore, as far as this discussion is concerned, there is no reason to
mandate 64 bits boundaries as part of the IPv6 specification.
- Pierre