Hi Mark,
Adding a address range as non-public to existing equipment is a lot
easier than adding IPv6 so that you can use DS-Lite. Much CPE
equipment doesn't have the flash capacity to do the later. The former
is trival provide the company that supplied the fireware is still in
business.
I'm mixing this conversation with the discussion on Class E, because I think
that your responses are perhaps less true if the address space is sourced from
that block.
I would contend that attempts to use non-traditional (Class A/B/C) addresses
will not be trivial for two reasons:
Firstly there are inherent assumptions in the uses of particular address
classes which can be littered through code.
This is a similar issue to the signed-integer time_t issue (though perhaps
simpler, still not trivial).
Additionally, host communicating with address classes outside the predefined
unicast set may not be able to reliably connect to older IPv4 devices, or
devices on specific networks (e.g. with existing Bogon filters).
Class E relies upon the endpoint understanding the format and purpose of the
address in a similar fashion to IPv6.
From my point of view, CGNs and Class E usage are part of a continuum of IPv4
address quality degradation. I'm not against these addresses being used, but
would seek to ensure that any uses of addresses in this fashion reflects the
current internet environment, and doesn't require engineering changes on third
party, legacy devices.
Sincerely,
Greg Daley
_______________________________________________
Ietf mailing list
Ietf(_at_)ietf(_dot_)org
https://www.ietf.org/mailman/listinfo/ietf