IPv10 Draft Specification Published

[nanonyme]

The thing is, as long as ISPs have to support IPv4, IPv6 is nothing but added cost to them. Which I believe is the problem IPv10 is trying to fix: let the ISP configure one protocol and one protocol only and be able to talk to the world.

Dual stack running costs aren't really that big. The problem has been ISP's needed to upgrade core switches with average life span of decades

[bug77]

Dual stack running costs aren't really that big.

They are when you factor in customer support

[M@yeulC]

It's a process called port forwarding. I think most modern routers though call it something else and do it automatically.

Port forwarding, yes, and modern routers handle this trough uPnP, which is a security hole, as it gives the false impression of being connected behind the router's firewall while you're not.

As a matter of fact, ipv6 and firewalls are not exclusive. But the big problem with NAT/ipv4 is that if I want to host multiple ssh servers at home, I can't set them up on the standard port 22 (the alternative would be to expose only one computer and hop from there, but that's far from perfect).
In theory, port multiplexing is possible, but not practical for every protocol. And, let's face it, NAT's raison d'être is ipv4 pool exhaustion.

As for NAT, there are multiple ways to establish a connection trough one (or even one at both ends), most of which require the assistance of a third party server.

This ipv10 is a joke. Or I hope it is, as it just looks like a joke, and doesn't even solve any practical problem. ipv6 needs to be deployed as soon as possible, starting with the mobile networks, and the IoT. One of the nice side effects of the much bigger address space is that it's currently highly impractical to scan the whole address space, so it might actually reduce exposure (even if we get rid of NAT).

Of course, NAT still has a few practical uses, but will be relegated to professional use, as it should be.

This is a timely reminder that I need to harass both my ISPs (mobile and DSL) to provide the ipv6 that they always promised for 2017
Seriously, it is time to go ipv6 only, and provide legacy ipv4 service with DNS64 or similar. Then, extend the ipv6-only regions until DNS64 isn't needed anymore.
Once the critical mass is reached (ipv6 only websites start to appear), the trend will be inverted and those remaining on ipv4 will need some special translating mechanisms to access them.

As a side not, ipv6 is perfectly suitable for local networks. Those are generally determined by the broadcast domain, anyway.

[rastersoft]

When I read the document I scratched my head, because that proposal looked insane. The idea is to create a new IP protocol (with its own IP version number) with, basically, the same header than IPv6, but with an specific syntax to allow to embed an IPv4 address. That means that all routers in the internet will need to receive a firmware update to be able to work with that (both the ones in the backbone and the users routers). But also the operating systems running in both the servers and the clients would need to be updated to be able to use the new packets, and, in case they use IPv4, update also the software to be able to use long addresses... But if you are going to do that, you can just migrate to IPv6 and forget IPv10...

But after reading two comments, now I see that this must be just an elaborated joke, like the RFC that explains how to implement IP over pigeons.

[DMJC]

The IPv6 standard is dumb. They should have just added extra octets to v4. 192.168.xxx.xxx.xxx.xxx.xxx.xxx.0 makes a hell of a lot more sense than retraining everyone on yet another dumb standard.

[torsionbar28]

I would not say 'nobody' uses it:

https://www.google.com/intl/en/ipv6/statistics.html

Not only that, but many large organizations have switched to using ipv6 exclusively. If you have a Verizon cell phone, you are using ipv6 exclusively. Verizon has disabled ipv4 on their cellular network.

[pal666]

In a home LAN, you do nothing with ipv6.

ipv6 is mandated by lte

[pal666]

LAN ipv4 addresses dont cost a thing

they cost nat

[pal666]

It's a process called port forwarding. I think most modern routers though call it something else and do it automatically.

port forwarding works only with static ports and with one client behind router
and for many protocols it will require protocol knowledge and mangling from router

[clockley1]

We do not maintain web servers. To make your kernel cleaner and faster, remove ipv6 support.

What evidence do you have to support your assertion that removing IPv6 will result in increased performance?