pTLA 3FFE:FFF0::/28 for test & example use per draft-blanchet-ipngwg-testadd-00.txt, closes 20Jun01

[Robert Elz]

    Date:        Wed, 06 Jun 2001 18:43:12 -0700
    From:        Bob Fink <fink@es.net>
    Message-ID:  <5.1.0.14.0.20010606182835.00af9bd8@imap2.es.net>

  | that Marc Blanchet has proposed a 6bone pTLA prefix be assigned for private
  | testing or examples in RFCs.

Yes.

  | I propose assigning the 6bone pTLA 3FFE:FFF0::/28 for this purpose.

No, please don't do that - that's just as bad as the example that Marc
selected and used in the draft.

What's needed is an address block that is patently invalid, not one
that comes from a valid range, but just happens to be unassigned.

That is, it should have an appearance more like net 127 in the v4
space (which of course, is only "patently invalid" with hindsight,
but in v4 there was no other real choice), that is, rather than the
rfc1918 set of nets, which you can only detect are "different" by
very close attention to the RFCs.

Make the address block for this purpose be one that can trivially, and
safely, be filtered by everyone, forever, and easily.  That is, one
that never is, or will be, valid for use for IPv6 forwarding.

The whole point of the kind of "dummy" address that Marc is proposing
(and it is a good idea, we should do it) is that it must never be
thought by anyone to be an actual address that can be used.

I'd suspect that something in the FE::/8 range would be the right choice
(the 0000:: range is the other possibility, but those addresses don't
look enough like "normal" IPv6 addresses to really meet the purpose).

kre