Why can't we just prefix IPv4 | Hacker News Reader

Why can't we just prefix IPv4

2 pointsneilellis7 comments

Instead of the complexity of switching to IPv6 couldn't we have a an IPV4.1 that supports variable length addresses with 0 indicating legacy.

So:

0.192.168.0.1 -> 192.168.0.1

0.0.0.192.168.0.1 -> 192.168.0.1

25.192.168.0.1 -> new address space 25 with IP of 25.192.168.0.1

25.100.1.2.3.4 -> new address space 25.100... with IP of 25.100.1.2.3.4

then the CIDR notation would just become:

25.192.168.0.1/32 for 25.192.168.0.0 ->25.192.168.0.255

Router's could use the x.y.z.0.0.0.1 address as gateways by default, making routing easy.

Under the covers it would be possible to use IPv6 as the protocol like 6to4 (https://en.wikipedia.org/wiki/6to4) but with variable sized addressing from a consumer's perspective.

Just a thought.

View on HN

There was a longer article on this, linked a few weeks back (actually written in 2002).

TL;DR: this would work, except for a completely minor issue: a large-scale HW+SW overhaul of everything up and down the net stack would be necessary before this extension would work (just as with IPv6). Therefore, no significant benefit over IPv6 (plus a bunch of brand new, IPv4.1 specific problems).

https://blog.sesse.net/blog/tech/2016-01-06-20-54_ipv6_non_a...

What would that make easier? You still need new networking stacks, have to redefine all protocols carrying IPs to allow for those, legacy hosts still can't use it, ...

+ Networking really doesn't like headers of varying length if avoidable, because it makes fast parsing (esp in hardware) harder.

That's a very good point. So here's an idea. Use the scheme I mentioned from a user point of view but translate to IPv6 as a transport protocol (since we're tunneling 6 over 4 at the moment this is not impossible).

So, IPv4 over IPv4.1 over IPv6? What good is the middle step?

"Legacy" IPs would have no way to access "new" IPs; they have no way to differentiate "0.192.168.0.1" from "1.192.168.0.1"

You'd need translation on router's of course because without upgrades they wouldn't understand the variable sized address space - but since I'm seeing people route IPv4 over IPv6 it's not completely insane.

You'd need translations everywhere - you couldn't just "let the router decide", the end nodes would need to support this as well.

( Let's say we had IPvIII, consisting of just three bytes per address: 10.1.2 - and now we went from that to 10.1.2/24, i.e. 10.1.2.0-255. How does the router to know that an IPvIII client is requesting 10.1.2.15 ? It doesn't. The extension from IPv4 to IPv4+ is equivalent.)

Reply on news.ycombinator.com