I encourage my competitors not to pursue IPv6.

I wouldn't be so fast declaring Ipv6 as a failure.

First, ten years of limited progress is not a lot of time compared to a lot of other, now succesful protocols.

Second, having two addresses in parallel really isn't that big of a deal. Typical networked devices already keep track of dozens of addresses at different levels. It's an engineering problem that can be handled quite easily.

Third, reclaiming addresses would require politics, which is much slower than technology. It's primarily an american political issue, since the US has 60% of all v4 addresses for 5% of the world's population - ie almost all unused v4 addresses are american. Twenty years of solving that problem, and lots of people have tried that as well, has yielded little return. One might argue that these "lazy, greedy, short-sighted and/or ignorant" people should've done something, but they haven't. The address growth is mainly in the rest of the world where operators probably feel that the chances of US giving up most of its address space are quite low, hence the interest in v6.

This last issue should not be taken lightly. Telecoms are closely linke to politics all over the world, and I think many countries see a chance to get a more fair and US-independent address allocation with v6.

Rainbow, I agree, it's time to give up on v6 and move on.

As to what to use to switch on other than names, the idea is that the name should map to a 48-bit application address. This is a locally-significant value, meaning that there is no need for global uniqueness.  The 48 bits are the "IP address" (often begins with 10. or 192.168.) and "port".  NAT maps a 48-bit value in to a 48-bit value out, the mapping being done at connection establishment time.

A lot of people in the network vendor and service provider community will say privately that IPv6 has already failed. The core protocols have been around for over 10 years with almost zero global adoption. V6 advocates say that this is because the industry is lazy, greedy, short-sighted and/or ignorant- but the other side of the coin is that a set of protocols that don't offer enough inherent value to make lazy, greedy, short-sighted people want to implement them don't deserve to be implemented.

A lot has changed since V6 was proposed back in the mid-90's. Is it time to write-off v6 and move on?

Can we start having an open and honest conversation about this? Or do we all have to politely pretend to be making the transition?

Okay, so I agree with your first paragraph (espeically if you read my comments to Carol).

I guess I am confused on how application addressing is going to help us with address exhaust.  At some level some switching/routing device must move packets along to the next stop.  Even if there is an application address, that name will not be unique per endpoint.  So, we would still have to resolve that application on that endpoint.  I think the idea of separating the addresses here is that you do not want to have to update the network to be able to introduce a new application.  You want the endpoints to be able to talk applications to each other while the network blithely shuffles packets between them.

Don't get me wrong, I am not thrilled with IPv6 but I think that is a ship that has sailed soon if we don't reclaim addresses.

seven

Reclaiming space would buy a lot of time.  But the point is that client devices never belong on public IP address space.  The v4 address space should be used for gateways and public-facing servers.  Private nets should stay in net 10.

IPv4 addressing architecture is incomplete.  Applications should be addressed by name.  IPv6 does not fix this; it just makes for more wrong numbers, so to speak. As a stopgap (not to v6; as you might remember, I advocate RINA as the real answer), one should think about the "address" as being one 48-bit (IP+port (field, not as if they were separate layers. NAT gets this, but fundamentalists who believe old textbook descriptions of ARPANET protocols don't. And that's who wrote v6.

I think BrooksSeven has a salient point.

With 84% of the current IPv4 address space completely unused, doesn't it make sense to just go through the administrative excersise of reclaiming and reallocating them? Isn't that easier than a compelete vertical and horizontal change to all applications, networking gear and back office systems that IPv6 requires?

I think BrooksSeven has a salient point.

With 84% of the current IPv4 address space completely unused, doesn't it make sense to just go through the administrative excersise of reclaiming and reallocating them? Isn't that easier than a compelete vertical and horizontal change to all applications, networking gear and back office systems that IPv6 requires?

Is your solution then to recover by force many of the IP address blocks that are wasted?  If we did that we could certainly delay IP address exhaust a LONG time.  Or are you suggesting we make an IPv7 which is completely different?

And I know he was only talking about double 4 NAT.  The problem is that this only solves one of the two huge hurdles on top of us.  Not the big one either.  Even if you are not happy with IPv6, then how about something more helpful than IPv6 sucks.

seven

He was talking about 4-4 NAT, not 4-6 NAT.  I see no point to 4-6 NAT, 6-6 or anything else with a 6 in it.  Once you realize that v6 was and is an Epic Fail of massive proportions, it al starts to make sense.

The problems he cited were that some NATs add latency or have limited capacity.  That's a capacity engineering problem; buffering strategy is often a problem, since some folks insert too much.