Your point about mobile networks kind of illustrates this. There are all sorts of weird and wonderful protocols used in the mobile world, and it doesn't matter because 99.999% of us never have to deal with it, and the 0.001% that do have time and space to become experts. IPv6 isn't like that -- it needs to be a consumer grade protocol that people can understand and work with. It either fails that test, or perhaps at best scrapes a borderline pass. We deserved better.
That's every standard protocol.
>Your point about mobile networks kind of illustrates this. There are all sorts of >weird and wonderful protocols used in the mobile world, and it doesn't matter >because 99.999% of us never have to deal with it, and the 0.001% that do have >time and space to become experts. IPv6 isn't like that -- it needs to be a >consumer grade protocol that people can understand and work with. It either fails >that test, or perhaps at best scrapes a borderline pass. We deserved better.
Wireline networks aren't any different, gobs of weird and wonderful protocols in MEF, carrier TE, DOCSIS, etc. Implementation issues drive 90% of the problems described. And don't think for a second that those kinds of issues are unique - all of this happened the same way in IPv4. It just happened so long ago that most either were not around for it, or forgot the pain. that shit box was classful in its original state. The difference now is that a bunch of people can see the sausage getting made and they don't like the ingredients, even though they've likely been eating the sausage for years.
I gave the example of mobile networks because it’s one of the few places consumer facing networks are growing rapidly and in some cases build entirely greenfield networks. And they have overwhelmingly chosen to build using ipv6 to do so.
Maybe mobile devices don't need incoming connectivity but doing a hack on a hack to save IPv4 addresses is a reflection on the genuine expertise at some of the national carriers.
When I was younger, telecom providers had all the professional expert engineering talent.
(I'm on superloop down under here, they actually have engineers and seem to have deployed a, what seems to be to me, a faultless IPV6 setup. The only issues I have found have been with toy things like pihole not behaving properly. Once I went to the Technitium thing my ipv6 experience over the last few years has been faultless.
Adoption was always going to be slow, but it never needed to be this slow. If the designers had thought a but more about usability and not done stupid stuff like making the address space contain twice as many bits as it actually needs, then we would have transitioned by now.
However now we are pumping out something like 4 ARM cores per person on the planet, per year. Clearly we will need to plan for those devices to communicate.
Also the effective address space in ipv6 is much smaller than 2^128 as the default subnet prefix is /64. So in a sense, it has already “addressed” your concern? Pun intended.
If the addresses were 64 bit instead of 128 bit the situation wouldn't be any different, beyond IPv6 would be harder to use. For humans unused bits end up hidden when written out and for machines most of the bits are used and are used in a way that make it scale and be simplified over using fewer bits. E.g. the assumption that every user subnet is a /64 instead of based on the number of devices in it is a massive simplification that hides a ton of complexity people ran into with IPv4 networks. /48 being the minimum aggregate advertisable on the internet makes the number of participants in the internet scale much better. Giving RIRs blocks of /23 leaves room to expand as assignments grow in the future - avoiding the problem we hit with IPv4. Just saying "half the bits is simpler" doesn't actually make your assumptions correct or the deployment happen any faster.
To re-iterate on this part: The problem isn't the protocol it's that there is little incentive for the average user to switch to anything else on their own volition when most didn't even really set up what they are switching from (and already working on today) anyways.
That's not to say the end user needs a more "usable" protocol so it becomes attractive to switch - that's not where the lack of attractiveness sources. The cost to use the protocol could be 0 (and IPv6 is actually extremely close to that to be honest) and there is still no incentive for consumers or businesses to make the switch yet. Even if everything goes perfect and is 100% done without any human intervention it still provides them no benefit until very recently when IPv4 prices started to rise so why would anyone have rushed to make the switch 20 years ago?
An example of this comes back to the cell carriers - they didn't go to IPv6 because they just like and have the time to do complicated things it was the simplest way to enable mobility across their networks with a large scale of devices. They hit this need much sooner than the price of a /24 went up enough for people to start caring so they pushed for their networks and phones to support this faster (on top of needing to build out a lot of new greenfield stuff anyways so why bother building greenfield with the old).
The ISPs still haven't adopted it after 25 years.
ISPs are a massive monopoly in the USA. IPV6 can't be adopted by software until a sufficient number of the backbone works. I've heard that "the backbone is all working for ipv6". Um, is comcast? Wasn't last time I had comcast.
If the ISP monopolies aren't 100% ipv6 (and you see lots of comments here that ipv6 support in ISPs is still "substandard") and convenient, then you can't blame the software people.
Look at your success story in mobile (which is behind a huge NAT to translate things to the "real" internet by the way). How did that work? Oh, you probably wrangled the three or four mobile companies into a room and got them to agree on protocols. Wow, success.
This needs to happen for the rest of ISPs. The fact it hasn't isn't a software issue, it is a governance issue. The failure is in the governance, the outreach with the real policy hammers like the FCC and ISP monopolies.
The governance has failed. It's been failing about 15 years longer than it should. FIFTEEN YEARS OF FAILURE!
Stop blaming reticent programmers, because IPV4 networking is still much much much easier than ipv6 in software, and IPV4 networking SUCKS between NATs and bridges and internal/external IPs and port mappings and what's-my-ip-on-the-other-end and dynamic DNS. Ohmygod it sucks. And ipv6 is worse than that!
Stop blaming the software people. IPV6 governance and outreach failed. Failed failed failed failed.
I don't want ipv6 to fail. I WANT STATIC IPs! EVERY PROGRAMMER WANTS STATIC IPS!
Get the ISPs and FCC in a room. Get google and microsoft and whoever else you need to lean on them. Get Amazon and google and microsoft IAAS into a room (hm, look, the same companies basically) and lean on them to support ipv6.
Oh, and you have the government to thank for the useless posix subsystem for windows back in the 90s (remember that?) as a way to force software “interoperability” between Unix and windows. Companies vying for federal contracts will look at the poorly defined requirements and find a way to just barely meet the letter but not the spirit.
"THOU SHALT USE IPV6 BY XXXX date or you don't get subsidies and/or lose spectrum and/or lose IPV4 ranges".
This is not designing a new complicated api compatibility. ALLEGEDLY the ipv6 geniuses have everything figured out technically. There is ALLEGEDLY no protocol development, RFC, standards body formation. ALLEGEDLY it's all been laid out. There are some success stories, there are "reference implementations" in production.
Again, it's not like this is hard to disseminate to the necessary powers that be. America is a bunch of monopolies and cartels. That means the regulatory agencies can call like 10 people and tell them the lay of the land. This is not herding a thousand cats with a fork, this is using shock treatment on elephants.
And how many of those mobile ipv6 addresses are communicating with the "real" internet through a carrier grade NAT to translate/intermediate with IPV4 servers/addresses?
And about the most offensive, disgusting thing to ipv6 people is the NAT. It's the thorn in their side that 1) IPV6 "trivially" solves (allegedly) but even worse 2) it's what keeps ipv4 on life support with the ISPs.
Can some ipv6 person tell me with a straight face that mobiles being ipv6 behind a huge cgnat is "success"?
So the fact that the big success story of ipv6 is all enabled by a huge massive NAT is ... morbidly hilarious.
So that actually means that most "regular internet" HAVEN'T adopted it, a statistic meant to show good uptake actually shows the real problem. The ISPs for cable/internet, Despite huge huge monopoly-protected revenues, aren't doing anything at all.
Again, as a programmer, I would like to work with static IPs. NATs suck, ipv4 sucks, private addresses vs public addresses suck. If not for the condescending, ivory tower, shove-down-throat, bungled incompetent planning, utter lack of outreach and transition planning, an addres transition plan (per the MX + A guy's thesis), I would love ipv6. I WANT A STATIC IP. EVERY PROGRAMMER WANTS A STATIC IP. We don't want to think about all the ipv4 bullshit.
The failure is with the ipv6 mafia. They failed. They keep failing. And based on everything they say and their representatives, they will continue to fail.
Yes, mobiles on v6-only with NAT64 to reach legacy v4 hosts is a massive success story. How can you say with a straight face that it's not?
Most big landline ISPs in the US are doing v6. Comcast, the poster child for awful ISPs, winner of multiple worst company in America awards, has been running v6 over their entire network for years now.
It IS a success, but it's a success that is singularly enabled by the #1 thing that ipv6 people hate with a passion: a NAT. The evil NAT that has kept IPV4 address space alive, that was the crutch that kept ipv6 from being adopted earlier. The evil NAT isn't a firewall, the evil NAT can be replaced by string-of-acronyms.
And, it's a success which paves the way for practically any protocol to replace ipv6 as the real successor, like the casually thrown out ipv4.4++v2, which is getting dismissed right and left by the ipv6 people here. Or god forbid a better protocol. Please give me ipv8.
Not that I was in the room, but it's apparent that the mobile success is due to the fact the cartel of phone makers and spectrum owners/mobile carriers got in a room and adopted ipv6 head to toe.
I would like to know if the FCC had a role here, or it's just that the mobile carrier industry is used to adopting new standards so this wasn't a big deal, or it was the OEMs that enforced/enabled it.
For the rest of ipv4-land, where is this cooperation/regulation/coordination for migrating?
My comcast modems do not support ipv6 last I tried, but that was three years ago. Maybe the magic wand has been waved finally. But I look at this thread and think: Eh, nope.
NAT is a necessary evil to deal with address space exhaustion in v4. NAT64 is just another application of that. In v6, when not dealing with backwards compatibility to v4, NAT is an unnecessary evil. Do you see the difference?
> And, it's a success which paves the way for practically any protocol to replace ipv6 as the real successor, like the casually thrown out ipv4.4++v2, which is getting dismissed right and left by the ipv6 people here.
It's being dismissed because it brings nothing new to the table. The people casually throwing out alternatives aren't thinking through them enough to realize that they've either come up with something that doesn't work, or they've come up with something that's basically v6 and has the same limitations v6 does. There's zero point in replacing v6 -- which 40% of the Internet's clients are already using -- with another protocol that's just as hard to deploy.
Comcast has had ipv6 support rolled out for some time now. I don’t know why you don’t see it.
In the UK, the two largest providers have been running IPv6 for years (BT and Sky), that's about 55% of the market.
Virgin and TalkTalk are ignoring it, that's a further 25%.
Vodafone are working on it ("next year due to Covid"), 5%.
The remaining 20% is small ISPs, and many of these support it.
My mobile carrier provides dual stack (with NAT in the IPv4 side). So no problem with any kind of site. Are there operators who provide IPv6 only? Haven't really read up on that topic, but I guess that requires NAT64 because the average user needs to reach IPv4-only sites. Does that generally work well?
If yes, why can the poster of the submission not use that on his Hetzner instances?
Apple has required app developers to support ipv6 only deployments since 2015: https://developer.apple.com/support/ipv6/
For landlines, I'm surprised because some ISPs who were not having difficulties with ipv4 are being the fastest to switch to ipv6. While some who had to deal with CG-NAT and other horrible stuff are fine with ipv4, and haven't yet planned to move to ipv6
Is this due to corporate culture or something ?
I guess you can't use NAT64 at Hetzner because Hetzner doesn't provide it.
There are also free NAT64 solutions, but of course you can't base anything but a random hobby project on something free. Maybe there are also commercial solutions, but IPv4 might still be too cheap to make that really interesting.
Actually the AskHN talks mostly about client cases. Those still seem "easy". Offering more than one service on let's say port 443 would require a full application level gateway / forward-proxy.