True, but until the 281 trillion network addresses are exhausted, nobody will ever need them. With perhaps 28000 network addresses available for every person on Earth during the next 30-70 years, it’ll be a while before anyone needs to borrow host addresses from anyone else.

I wouldn’t be surprised if some individuals only needed a thousand or so networks. There might be a market for selling a few if any are ever seen as spare network addresses. I’m pretty sure I won’t use more than a few hundred, and I could probably get my wife to give one of hers up if I need to add another quadrillion or so hosts.

The numbers are too big for any foreseeable need. Granted, we’re running into the exhaustion phase of the approx 16M (max) networks that IPv4 tried to handle (assuming /24 networks). IPv6 allows adding 10000 entire IPv4 internets each year for the next thousand years. And every one of those can have more hosts inside of them than will likely ever be used.

I suspect that the networks themselves will have intelligence enough to replace addressing before a significant problem of exhaustion shows up. The whole concept of addressing seems likely to be superseded first.

I just can’t see waste being a part of it when it’s simply a portion of a logical structure that no one will need.

Tom

That’s a little difficult to determine. It’s kind of like saying that it’s wasteful to have numbers like -62168943.77344987 since no one is likely ever to use them.

Tom

Its not like IPv4 where you could subnet a /24 to, say, several /30′s. once you hit the IPv6 /64, thats it.

If youre getting the prefix from somewhere, aaske them for a larger block. If you yourself are setting up an IPv6 lab, then start with a /48 and use the remaining 16 bits for your own internal subnets…

I’m not an IPv6 subnet expert by any means, but I can see how the simplest examples can work. As I understand it, your basic addresses will look like this:

gggg:gggg:gggg:ssss:iiii:iiii:iiii:iiii

Those represent the 32 possible hex characters that may make up a full IPv6 address — eight blocks of four hex characters per block.

The first three blocks (gggg) will be the (global) part of your address block. That takes up the first 48 bits (or six bytes) of your addresses. The next block (ssss) will be used for your subnetting. That takes 16 bits (or two bytes). The last four blocks (iiii) will be for your interfaces. That takes the final 64 bits (eight bytes).

All total, the address is 128 bits (16 bytes).

You can’t do anything with the ‘global’ bits. Those will be assigned to you when you receive your address block.

For your subnetting, you’re interested in the 16 subnet bits in the middle of your addresses. A full IPv4 address is only 32 bits. For IPv6, you get to use 16 bits just for the subnets you want to set up.

Say, for example, that you need 1000 subnets for your labs. (Remember that each subnet can have 64 bits worth of interfaces under it, or 18 quintillion hosts per subnet.)

You have 16 bits available for subnetting and you want your labs to have 1000 subnets available. Since 1000 will fit in only 10 bits (2^10=1024), that leaves six bits left over. Those six bits let you set up a hierarchy of 64 (2^6=64) higher level subnets, each of which can have 1000 subnets beneath it, each of which can address 18 quintillion interfaces.

Or maybe your labs only need 30 subnets each.

Since 30 fits in five bits (2^5=32), you have eleven bits left over to set up a hierarchy of subnets. You might use one bit to have one level in your hierarchy indicate labs or “not labs”. That still leave ten bits to set up more subnet levels in your hierarchy; each of those would aloso have the labs/”not labs” bit.

All you really need to know is that you will use 16 bits for your subnetting and that the bits represent whatever hierarchy you want to create.

A 16-bit subnet field is enough for 64K subnets within your address block. If you can’t get by with that, it’s hard to see how you could be getting by with IPv4 now. It’d be hard to see how you could get by with any addressing scheme at all.

Tom