[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
RE: Review of draft-ietf-v6ops-nap-02.txt
- To: "'Walt Lazear'" <email@example.com>, "'Fred Baker'" <firstname.lastname@example.org>, "'Thomas Narten'" <email@example.com>
- Subject: RE: Review of draft-ietf-v6ops-nap-02.txt
- From: "Tony Hain" <firstname.lastname@example.org>
- Date: Thu, 1 Jun 2006 08:37:37 -0700
- Cc: "'Jim Bound'" <Jim.Bound@hp.com>, "'Brian E Carpenter'" <email@example.com>, "'EricLKlein'" <firstname.lastname@example.org>, <email@example.com>, "'Ralph Droms'" <firstname.lastname@example.org>, <email@example.com>, "'Lindqvist Erik Kurt'" <firstname.lastname@example.org>, "'Margaret Wasserman'" <email@example.com>
- In-reply-to: <009501c684e2$488021f0$5100a8c0@WaltXP>
Walt Lazear wrote:
> I agree! I'm a firm believer in the need for IPv6, but
> I get a deja-vu knot in my stomach when I hear some IPv6
> evangelists (such as at an IPv6 Summit) talking about the infinite
> address space and great applications it enables. There is
> breast beating about Jon Postel wantonly handing out
> Class A's to buds in IPv4 land and yet I see the same kind
> of large chunks of IPv6 land being distributed freely.
> Perhaps we're letting the RIR's head down the same road?
> /64's for serial links! 15 quintillion addresses where 2-4
> will suffice?!?!?!
> What's scary are the applications that give addresses to every
> retail item so it can be tracked during it's lifetime. No
> reuse of addresses is contemplated (because they're infinite).
> Imagine the ultimate usage, where nano-things in aerosol sprays each
> have an address that is literally "cast to the wind". Is there a
> /70 or /80 in every can of this stuff? Why is there no built-in
> mandatory recycling of address space for these applications?
There appears to be some confusion here. Even if every device that attaches
to the network has a lifetime unique tag, its uniqueness is bounded by the
subnet that it is attached to. The same tag might be in use in some other
part of the network with a different routing prefix.
> We're undoing CIDR-like usage.
That is simply not true. Allocations for subnets are done in exactly the
same hierarchal manner. The only difference is that the subnet size is fixed
and large so an arbitrary number of uses within that may occur. One of the
major new security efforts that is simply not possible with the small subnet
space in IPv4 is SEND (RFC 3971). The subnet is still contained within a
CIDR block, but new capabilities are taking advantage of the larger space.
> Hierarchical addressing is
> going to be wasteful by nature. I don't know how the IETF
> can influence the RIR allocation policy, but it's scary
> how much history seems to be repeating itself. We're
> at the early end of the cycle, so it may not seem to some
> that we're handing out addresses with both hands, but we are.
You are focused on the number of addresses, where the IPv6 allocations are
done on number of subnets. Even so there are larger blocks being handed out
than with IPv4, because fragmentation of the space causes more routing table
bloat than the savings. In IPv4 the resource is scarce. In IPv6 it is not
infinite, but with the HD ratio change being adopted by the RIRs the
lifetime under the current generous allocation practice is projected at over
500 years. This protocol is not the be-all-end-all of protocols. It will be
replaced by some thing better before 500 years pass.