In message
On Wednesday, Feb 12, 2003, at 21:00 Canada/Eastern, Frank March wrote:
[APEC workshop will cover] Issues surrounding the current standard of Internet Protocol version 4 (IPv4) and how it limits the number of Internet Protocol (IP) addresses available globally.
Someone should tell these people that we are not running out of IPv4 addresses any time soon
They actually only talk about IPv4 limiting the number of addresses available -- which is true. There are only 2**32 of them :-) But it's not quite the point. However people outside the network-core could be forgiven for believing that there's a dire shortage of IP addresses. When only a few years ago a business connection would often have a /24 or greater address space, often one which was "portable" (in theory, but perhaps not in practice), these days plenty of business connections are given a single (gateway) IP address, and it is strongly suggested they use NAT -- even when hosting multiple servers requiring incoming connections. (In at least one instance I've seen multiple connections installed primarily "to get another IP address". When they started in on adding a few more hosting servers, and ordering their third connection "to get another IP address" I put my foot down, and they have finally applied for a /28, which will already be > 50% filled simply properly numbering the existing servers.)
and that the real pressing issue is how you make the routing system scale (an issue which IPv6 does not attempt to address).
This is the flip side to the "one address, or maybe a few more if you beg lots". You end up lots of very long prefixes. To be sure many of those couldn't care less about routing, and CIDR works well for them. But you end up with a certain percentage that have a legitimate need to do their own routing, for redundancy or otherwise, and still have fairly long prefixes. Even if they go through the (non-obvious to those outside the core networking arena) process required to get a /24, that still gives you a lot of /24s floating around to deal with routing. I tend to agree with you that by and large IPv6 will just shuffle this problem around some, and still leave basically the same issue with "how do we deal with all these routes". (The address space will be larger, and in theory will be more hierachically allocated, but there'll still be the same people, and the same portion of them will want/need to do their own routing, etc.) I'd be happy to see IPv6 deployed to get rid of some of the worst excesses of NAT I see (particularly the "traffic in to one IP address splits off to N internal servers in difficult to keep track of fashions") but other than that I don't think it's going to solve all that much by itself. Ewen