IPv4 is now a minority share

Some time in the last few months, Google's "ipv6 share" from US traffic went above 50%. Ipv4 is now a minority of traffic to Google from the US. Moreso even in France (72%) and Germany (66%).

OpenWrt handles Ipv6 flawlessly out of the box for most providers.

I know I'm celebrating!

10 Likes

U.S.A.! U.S.A.! :us:

IPv6 only make senses in BIG data centers.
IPv4 is best for home users.

4 Likes

Tell it to those users whose ISPs only offer IPv6.

3 Likes

ISPs that offer only IPv6 without NAT64 (via RFC6146) are broken. Thankfully, OpenWRT supports RFC6146 via the 464xlat package, and, to the connected clients, it then looks like the ISP supports both IPv4 and IPv6, with some strange encapsulation hop just after the router.

Don't get too excited yet. The world isn't Google colored. APNIC uses Google ads to measure ipv6 deployment. Only about 30% of users worldwide are ipv6 capable. The industry has been working on the ipv6 transition for like 25 years to get to 30% deployment, and it wouldn't surprise me if it takes another 25 years or more to get to 100%. There are some parts of the world that rely heavily on carrier grade nat to get by without ipv6.

https://stats.labs.apnic.net/ipv6/

1 Like

The problem is the looooong tail of services and servers (and end users) only available via IPv4, so not that much to celebrate yet.

1 Like

while the percentage of people with ipv6 is small, there is little incentive to change. When more than 50% of your users have ipv6 there's now less of an incentive to keep ipv4 at all. Of course globally we're still below 40%, but globally less than something like 20% of all people even have internet access. Africa, south america, and such could change that a lot.

This is just not true at all. I'd love to be ipv6 only on my LAN. In fact I ran that way for a full year (DNS64 and NAT64 at the router) and no-one in my household cared. Even minecraft works fine on ipv6 now, and my kids occasionally share their "LAN" games via ipv6 with friends.

Ipv6 is literally easier to configure, easier to administer, and dramatically more capable than ipv4. the only reason to have ipv4 is the stupid stuff that doesn't yet handle ipv6, like my TP-link omada controller (they're working on it though).

Even my IP cameras support ipv6.

The reason to celebrate is that soon people are going to figure out that ipv4 only is a minority of people in major western countries.

100% is irrelevant 95% would be fantastic, and the difference between 95% and 100% could easily be 100 years. The question is, how long to 95%? Native IPv6 hit 1% at google in 2012, now almost 2023 it's about 40%. It hit 20% in 2018, if you extrapolate at purely linear rate you'd expect 90% in about 10 more years. But the more people have ipv6, the more incentive there will be to add ipv6 as an ISP and server farm etc... I expect a burst of transition about 3 years out which suddenly adds a dramatic amount of ipv6 capacity. Finally, ISPs etc will realize that they need to do it, and once they enable ipv6 big swaths of people will have it. Incentives matter, as long as it's 10% of people or less you have little incentive, once it's 50% or more, there's big incentive. Things should be nonlinear.

4 Likes

If you drank the coolaid ;).... that is IPv6 with all its warts fixes IPv4's single biggest impediment, the lack of enough addresses, but that about sums it up.
The other changes are IMHO more of a mixed bag, including 64bit interface identifiers, but I do not want to rehash this discussion here and now.

Easier to configure? Yes and no, or only if you buy in with the 'new way' of doing things, trying to keep well established practises in use can result in loads of unhappiness (cf DHCPv4 and DHCPv6). I am not trying to restart that discussion but 'easier to configure' is not unconditional.

3 Likes

In the context of for a home user, I think easier to configure is just definitely true. Plug in and everything gets an ipv6 move on. For someone who doesn't even know what ipv4 DHCP is, that's a good thing.

2 Likes

This is IMHO a matter of taste, on OpenWrt routers both IPv4 and IPv6 just work out of the box for endhosts. Endpoints generally just need to be plugged in. However making an internal host reachable from the outside seems simpler with IPv4, in IPv6 one needs to poke the same hole through the firewall as with IPv4 (and rightly so, no complaints on this one) but then one needs to wrestle with the endhosts own IPv6 address stability on the to be exposed endhost*.
Trying to be fair here, I can not see a clear winner either way, both have their warts...

*) Exposing MAC addresses in interface identifiers was/is just way too naive, while privacy extensions push the needle more into way too paranoid territory, I would say if the original policy would not have been as bad as it was, the 'fix' might not have swung the pendulum so far into the other direction; but see @bmork's remark on a better middle-ground by RFC7217 (published 2014, after 2007's RFC 4941 and 2001's RFC3041, both describing "Privacy Extensions for Stateless Address Autoconfiguration in IPv6")

Disclaimer, I mostly agree with you and think that all in all IPv6 is am improvement, I a, just not as exited as some as I see considerable squandered possibilities as well.

Yeah, if you want to share a device to the outside world the easiest way is with tokenized addresses... basically just give your device a token like

::abcd:1234:0101:1

and then it adds that to the end of all the prefixes... done.

Which is a step that for IPv4/DHCPv4 (as used by OpenWrt) is simply not needed at all, hardly 'easier to configure'. The sad thing is that a desire for equally expressive/capable DHCPv6 was totally predictable or should have been, but since IPv6 apologets did not considered a valid use case we ended up with the status quo where DHCPv6 works for many IPv6 endhosts while others (looking at you Android) don't play ball without being in violation of the specs.
Again 'easier to configure' only if one buys into a whole new 'philosophy' to go with >32bit addresses.
I am again fine with IPv6, but let's not overpromise, as what it does deliver (addresses enough for the foreseeable and unforseeable future) is totally sufficient to justify/motivate its deployment (in spite of the warts).

well, you either need a reservation on the router, or you need a long enough lease that your device just never loses its lease. Both of which are options for ipv6 on any device you'd actually want to use as a server endpoint (DHCPv6 does work fine on Windows, Mac and Linux). I mean, unless you're nuts enough to want to use an android cell phone as a server...

Doesn't uptake of IPv6 undermine further uptake because it frees up addresses for IPv4 again, undermining the need to switch over to IPv6?

2 Likes

If team IPv6 had not screwed up, this would be as a consequence of decent standards documents and not a pick and choose approach by endsystems If at all the pretty comprehensive list of DHCPv6 supporting endsystems implies that it would have behooved team IPv6 to make DHCPv6 a first class supported option in which case we would not have to come up with excuses for android.

I work for a hosting company and we are never going to support IPv6 on internal networks, it is far too "magical" and we don't want all that stupid quorum/management traffic. We like static routing and static addresses without an RA and without lame ass broadcast packets (besides ARP). Where IPv6 has been intentionally routed through, we do it fully static without any router-advertisement and related other stupid PnP magic fully disabled.

Same as how we don't run DHCPv4 to assign addresses to internal equipment like routers: those things get assigned a static IP forever (and can't be lied to by a rogue DHCP server, same as a rogue RA server could hijack IPv6 realms). It's just insecure to have magical stuff going on like automatic addressing or automatic routes or etc. And managing IPv6 fully statically is sort of a pain in the butt.

7 Likes

Android made a very politically motivated choice, this has zero to do with the IPv6 designers. The point of the Android choice is to force it to be the case that address hoarders can't keep addresses away from end users. Unfortunately we live in a world where there's a group of people who would LIKE for it to be the case that you must beg them and/or pay money to use an address. I think Android saying F-that we're going to make the biggest most important operating system on the internet with multiple billions of users push back hard against that.

In the absence of rent seekers, where for example a /56 is guaranteed to every person on the planet and a /48 to anyone who asks, such a choice would not be required.

Never say never. I predict within 10 years you'll be LOVING ipv6, it'll happen when the people you hire grew up natively speaking ipv6.

Maybe or maybe not. Sweden as a country has pretty much ignored the whole IPv6 concept up until now and there isn’t showing up any will from the ISP to implement it either.

Yes cos much like that idiot pottering with systemd, google only supporting SLAAC is down to one idiot engineer who refuses to follow international standards. There is a bug filed on android's tracker that is marked as won't fix so don't ever expect to have dhcpv6 on android.

Apple understood this and made their ecosystem use dhcpv6 and allowed proper management of their devices. Like it or not that's why apple devices are much more easily used in corporate environments and android devices are hated.

1 Like