On 8/15/19 2:06 PM, tom petch wrote:
----- Original Message -----
From: "Keith Moore" <moore@xxxxxxxxxxxxxxxxxxxx>
To: <ietf@xxxxxxxx>
Sent: Thursday, August 15, 2019 4:29 PM
On 8/15/19 3:33 AM, shyam bandyopadhyay wrote:
To:
The Entire IETF community
Sub: Why do we need to go with 128 bits address space if
whatever is been trying to achieve with the existing
approach of IPv6, can be achieved by 64 bits address
space as well?
Dear Folks,
I raised this issue couple of times earlier. My intention was to
collect
all the points in support of 128 bits address space and try to
figure out
whether they can be solved with 64 bits address space as well.
The answer is no, they cannot, particularly not if either address
assignment or routing are hierarchical.
It remains to be seen whether 128 bits is sufficient. I'd place even
odds on an extension to IPv6 to permit addresses longer than 128 bits,
gaining favor while I'm still alive.
(Though it's also possible that global climate change will eventually
make even 32 bits more than enough.. But it doesn't seem prudent to
presume that.)
More generally, any decision made long in the past can always be
second-guessed. But the people who insist that a different decision
should have been made, have little reason to have confidence in their
belief. After all, they didn't have the burden of making their
preferred choices work in the real world.
It's also hard to believe that with nearly 25 years of painful
investment in IPv6, dual-stack transition, /and/ NAT, with plenty of
accumulated evidence that IPv6 actually does (mostly) work better than
the alternatives, that people are going to somehow abandon all of that
investment for a completely new packet format that has the worst
features of both IPv4 and IPv6 - insufficient address space AND
transition burden. To put it differently, IPv6 is going to have to
fail a lot worse than it has, before there's sufficient interest to
face
another transition. And while anything is possible, IPv6 really was
designed to last for several decades (and with the assumption that
human
population AND internet appliances will continue to grow
exponentially). 64bit addresses were considered for IPng, and that
would have likely been the choice had they appeared to be sufficient.
Keith
When I started teaching IPng in the late 1990s, ATM was the
up-and-coming link protocol and what immediately struck me was that the
link identifier of ATM was (potentially) more than 64 bits and so could
never fit into the bottom half of a 128-bit address. So yes, I always
expected the address structure of IPng to be too small.
If you want to look at a different approach, the following was recently
developed:
https://datatracker.ietf.org/doc/draft-moskowitz-flexip-addressing/
There is more, but I never posted it, as there was little I could do
with FlexIP that I could not do with IPv6.
ATM addresses would have been one possible advantage that I did not
investigate.
Bob
