Date: Tue, 29 Nov 2011 21:09:22 -0700
From: Sumanth Channabasappa <sumanth@xxxxxxxxxxxxx>
Message-ID: <76AC5FEF83F1E64491446437EA81A61F81D7CBBA11@srvxchg>
This whole question is weird, when someone needs an address to use,
and given that the pool of free (or close to it), that is, easily
available, addresses no longer exists, I'm going to take whatever
address will work for me for its purpose.
What the various addresses are marked as in some RFC or IANA list, is
immaterial, all that matters is what works.
If that means "borrowing" (or squatting if you prefer) on the addresses
some ISP in some remote part of the world uses for its customers, that my
customers will never need to communicate with (in my opinion), then that's
what I will do.
More likely, if I can, will be to take some address that I know can't
be needed, and use that - 1918 addresses are in that category. So is the
"documentation prefix" (which only has as its problem that it is too
little address space to actually use for most purposes). The only real
criteria is whether the address I'm considering is in the routing tables or
not - if it isn't, regardless of reason, then I can use it without harm.
For most purposes, the 1918 address space is going to be the right choice
whenever I don't need a routable address, and it is here too.
| ISPs have already indicated (a few times) that RFC1918 space is not
| practical behind the CGN due to the (real) possibility of overlap
| with customer addressing.
Frankly, that's nonsense. Not that there's the possibility of overlap,
of course there is that possibility. And not that things would break
if there was a duplicate allocation on both sides, they might, probably
even would. But that that actually matters if done sensibly.
We know that normal consumer CPE equipment doesn't use network 10 (you
just have to look at some to see that, they almost all use 192.168
addresses). Further, it is hard to imagine how any network not managed
by network professionals would, or could, ever use network 10, with the
possible exception of 10.0.0/24 or 10.255.255/24 which are numbers that
one might imagine some odd CPE manufacturer might just pluck from the
air.
If the ISP were simply to use 10.64.0.0/10 as the /10 they claim to need
for their CGN's, the chances of that conflicting with any customer who
doesn't have a network staff smart enough to deal with the issue is
basically zero.
For the one in a thousand (probably less) customers who do end up having a
problem, there are plenty of other 1918 addresses that the ISP could pick from,
for that individual customer, which would not cause problems. Certainly,
no-one is going to want to have to deal with every individual customer,
but dealing with one or two odd cases should be no burden.
Further, I can't imagine that the ISPs aren't aware of this, they know
what CPE equipment is being used, and how it is typically configured.
They also know they could cope with the one in a thousand end user who
actually has configured 10.64/8 in the inside of their CPE, and isn't
willing to change that.
The paranoid in me suspects that the IESG should not be making a decision
to approve a /10 for CGN usage without making sure the anti-trust policy
that's being discussed in another thread is in place, and that everything
in this request has been in accordance with that policy. That's because,
to me, this smells like a cartel of major ISPs with plenty of allocated
IPv4 address space (legitimately allocated, and used properly right now)
looking for a blessing from the IETF (that using private addresses and CGNs is
OK) to be able to reclaim much of their currently allocated v4 address space.
And I cannot imagine a single one of them (or almost any) just returning
that space to the RIR's for allocations - in many countries, doing so (by
any of the major ISPs that are listed public companies) would probably be
a breach of the director's duty to properly manage the resources of the
company - since addresses (address blocks) sell for such high returns these
days, simply giving away whatever the company has had allocated could easily
lead to prison terms for the directors of the ISP... Other than small
privately owned ISPs, they really cannot return the addresses, so their
only option is to be selling the things for the profit they'd gain.
Does the IETF really want to be blessing a doc written by a group who all
appear to come from the very types of ISPs who would be subject to & likely
to benefit from such an action? (Note: I am not intending to suggest that
any of the authors of the doc was actually planning such an action, just
looking at how it all appears from the outside.)
Lastly, while I am here, I have absolutely no sympathy at all for
manufacturers of consumer equipment (or any equipment) that is currently
being sold, or has sold in the past 5 years (at least) that is not
able to use IPv6 rather than IPv4 for network connectivity. Which it
should actually use in practice is a different issue, but any IPv4 only
equipment on the market is not our problem. We do not need to keep making
that trash functional. If people buy it, and it fails to work, let them
just return it for a refund. The manufacturers will soon learn then.
At the minute, this policy of "we must keep IPv4 working because there
are still people with no other choice, and there keep being more new
users with no other choice" is just penalising the equipment manufacturers
who have invested the time & effort to be ready for when IPv6 is needed,
in order to favour those who have only ever been concerned about today's
bottom line. It is time to draw the line, and start telling people that,
as predicted, IPv4 addresses are now exhausted, and to connect, they need
to use IPv6. If, short term (one hopes) we need some IPv6->IPv4 type NAT
(NAT-PT) then at least that's a step towards a long term solution, rather
than yet another attempt to avoid change, which is all CGN's are.
If the IESG eventually decide to go ahead and publish the draft in
question, the number they should fill in in the IANA considerations
section should be 10.64.0.0/10 or 10.192.0.0/10 - reducing the pool of
available routable addresses, even further, even by just a /10 (0.1% of
the available address space) would be insane. Reducing the pool of 1918
addresses by a fraction would be harmless. If ARIN have spare /10s
lying around that they don't need to allocate, I'm sure APNIC, or RIPE,
or LANIC (or anyone else) would be happy to take them.
kre
_______________________________________________
Ietf mailing list
Ietf@xxxxxxxx
https://www.ietf.org/mailman/listinfo/ietf