4 ms·

Why can't we just prefix IPv4

2 points by neilellis 11y ago

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.

detaro 11y ago
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.
- neilellis 11y ago
  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).
  - Piskvorrr 11y ago
    So, IPv4 over IPv4.1 over IPv6? What good is the middle step?
justinsb 11y ago
"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"
- neilellis 11y ago
  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.
  - Piskvorrr 11y ago
    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.)
Piskvorrr 11y ago
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_alternatives_djbs_article_13_years_later.html https://blog.sesse.net/blog/tech/2016-01-06-20-54_ipv6_non_a...

2 points by neilellis 11y ago

detaro 11y ago

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.

neilellis 11y ago
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).
- Piskvorrr 11y ago
  So, IPv4 over IPv4.1 over IPv6? What good is the middle step?

justinsb 11y ago

"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"

neilellis 11y ago
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.
- Piskvorrr 11y ago
  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.)

Piskvorrr 11y ago

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_alternatives_djbs_article_13_years_later.html https://blog.sesse.net/blog/tech/2016-01-06-20-54_ipv6_non_a...