David Conrad wrote:
Eliot,
On Oct 15, 2005, at 2:34 AM, Eliot Lear wrote:
The IETF cannot legislate prefix lengths, but the argument behind
conservation beyond /48 would be utterly silly and demonstrates a
"revenue opportunity", plain and simple.
When multiple /19s and /20s have been allocated and there are rumors of
much shorter prefixes which can be justified under the current rules,
I'm not so sure discussions about conservation can be classified as
"utterly silly".
This is a matter of timescale. Prefixes should be expected to
change. In fact, SHIM6 should be able to provide equivalent of SCTP's
ADD-IP to *devalue* prefix stability.
Unfortunately, since applications are aware of IP address structures
(both v4 and, sadly, v6), you'll need to rewrite all applications,
libraries, and kernels to tolerate changes in those addresses at
arbitrary times or face broken connections or misdirected packets. An
apparent base assumption in IPv4 was that addresses were stable over a
"session" (be it connection-oriented or connectionless). This
assumption permeates every IP stack. Given DNS root server IP
addresses retired over 10 years ago still get 30 queries per second, I
doubt prefixes can be expected to change by all applications in our
lifetimes.
That is the main driver behind the way shim6 preserves the ULID (i.e.,
the address as seen by the upper layer).
The other main driver is that after 13 years of knowing that a scaleable
PI solution would make this problem set go away, we still don't have one.
(Defining PI address space that can be used by a few tens of thousands
of large companies doesn't have anything to do with IDR scaling to tens of
millions of smaller sites.) The multi6 WG did thrash out all these arguments
before converging on a direction, by the way. The only new argument I've
seen recently is the need for some support of TE functionality in shim6, which
is a very valid point but doesn't belong on this list.
Instead of predicating the future of the network on a non-existent solution,
we're designing one that has a chance of working on the scale of a billion
node Internet.
Brian
--------------------------------------------------------------------
IETF IPv6 working group mailing list
[email protected]
Administrative Requests: https://www1.ietf.org/mailman/listinfo/ipv6
--------------------------------------------------------------------
