Slashdot Mirror
ARIN Letter Says Two More Years of IPv4
dew4au writes "A reader over at SANS Internet Storm Center pointed out a certified letter his organization received from ARIN. The letter notes that all IPv4 space will be depleted within two years and outlines new requirements for address applications. New submissions will require an attestation of accuracy from an
organizational officer. It also advises organizations to start addressing publicly accessible assets with IPv6. Is ARIN hoping to scare companies into action with the specter of scarce resources? This may be what's needed to spur adoption since there appears to be no business case for IPv6 deployment."
I just got back from the ARIN meeting this week and the letters are, indeed, a "scare tactic". Network providers keep reporting that PHBs won't spend any money on IPv6 even though engineers are begging for it. Most corporate officers probably think IP is only Intellectual Property and this is an attempt to draw their attention to the fact that the network world as they know it is going to end soon and that the only way to avoid serious problems is to either stop growing or to start IPv6 deployment. PHBs sometimes get the idea when they realize that not spending some money will lead to big problems in a few years. Others figure that if it's over a year away, it really does not matter because it won't impact their bonus this year, so it may not work, but we can hope.
Kevin Oberman, Network Engineer, Retired
How about they take back the Class A address space owned by companies who probably aren't even utilizing it. Here's a list of a few companies who have class A licenses and you wonder how much of it they are even using:
General Electric 3.0.0.0 - 3.255.255.255
IBM 9.0.0.0 - 9.255.255.255
Xerox Palo Alto Research Center 13.0.0.0 - 13.255.255.255
Hewlett-Packard 15.0.0.0 - 15.255.255.255
Hewlett-Packard (originally DEC, then Compaq) 16.0.0.0 - 16.255.255.255
Apple Inc. 17.0.0.0 - 17.255.255.255
Massachusetts Institute of Technology 18.0.0.0 - 18.255.255.255
Ford Motor Company 19.0.0.0 - 19.255.255.255
Royal Signals and Radar Establishment 25.0.0.0 - 25.255.255.255
Halliburton Company 34.0.0.0 - 34.255.255.255
Why the hell do some of these companies even need 16+ million addresses? I can't see them utilizing the space available, but maybe someone here can enlighten me on how that is done (aside from trying to justify a public IP address for every workstation).
Awesome idea. We'll give Google 1/40, The government can 2/40, IBM will get 3/40, etc etc etc
Same problem. The ipv6 is not a "bad" idea, it's just sort of like...imagine in 1950s if the phone company decided "we could go with area codes to subdivide numbers to prevent running out, or we could use letters AND numbers".
Can you imagine the upheaval?
In a lot of ways, that would have been even easier to deal with, because everyone's phone was owned by AT&T. New phones could have been issued without too much problem.
No, imagine it instead in the mid 1980s. Ma Bell doesn't own the phones any more, in fact there are tons of cheap phones available, cell phones are starting to come out, and there are still rotary AND push button phones.
That's more like what the IPv6 switch is like. Do you give the new people 2 numbers, so that grandma can still call them? How long is it before you stop accepting legacy phones that only have 10 dialing options? How the hell do you get DTMF to work with 36 numbers? Do we need area codes? It would be weird without them, but we don't really need them.
The equivalent of these questions are still being asked. Just a couple of months ago, there was a huge to-do about NAT and IPv6. "IPv6 is a world without NAT". The hell it is. My internal routers don't get publicly routable IP addresses, even if I have to NAT back to IPv4.
When the wrinkles get ironed out, we're going to wonder how we ever did without it. During the transition, it's going to be hell for everyone (with the possible exception of the clueless end user, who might have to buy a new router at most).
Check out my sysadmin blog!
I have your killer app right here.
upon the advice of my lawyer, i have no sig at this time
Just a couple of months ago, there was a huge to-do about NAT and IPv6. "IPv6 is a world without NAT". The hell it is. My internal routers don't get publicly routable IP addresses, even if I have to NAT back to IPv4.
I agree with the sentiment - however, it's one of policy, not mechanism. NAT is a pretty poor substitute for a router that implements policy (known as a firewall). NAT has literally an all-or-nothing granularity. For instance, I might want to specify that an internal host can enable BitTorrent via UPnP, but under no circumstances can CIFS be allowed through - in either direction. An internal host sending a CIFS solicitation out does not mean a pinhole should be opened and some set of hosts (depending on cone of restriction) free to respond. NAT is just not a practical policy tool. It's an address space recovery tool. Reverse NAT, however, has some redeeming qualities for load balancing and failover - I'm not versed well enough in IPv6 to understand how they'd be implemented without NAT. (Anycast addressing, I suppose.)
But you can implement NAT in IPv6 just as much as in IPv4 if you wish. A router could appear to have a single interface ID and translate to/from that. It's largely unnecessary though since instead of a handful of IPv4 addresses you have an entire 64-bit space to yourself (and maybe even the SLN prefix, not sure about that).
IPv6 really is a major cleanup and simplification from IPv4. I'm slightly disconcerted by the increased dependency on DNS however.
> I thought IPv6 split the network and local address segments right down the middle (i.e. each is 64-bit).
From what I remember, that was more or less the plan circa 2002-2004. The main problem with the original address allocation scheme was that it left big gaps in places that made it nice to route, but a bitch to memorize and rendered the proposed shortcut notation all but useless. Originally, they planned to use the upper 3 bits as a grand macro-level version indicator, then leave the next byte zero for now, then hop and skip over the next few bytes using the lower bit or two of each byte until they got to the "meat" of the address somewhere around bytes 5-8. That would have resulted in lovely addresses like 100:103:401:3f7a:xxxx:xxxx:xxxx:xxxx which, if you actually used your MAC address to set the lower 48 bits, would have been all but impossible to meaningfully encode with the "::" zero-packing shortcut. At best, you might have ended up with 2 pairs of sequential zero bytes to compress, and had to pick one or the other.