4 ms·

I have to say I’m super disappointed in the ignorance and negativity in the comments on this thread. Ignorance of both the difficulties inherent in upgrading a

I have to say I’m super disappointed in the ignorance and negativity in the comments on this thread. Ignorance of both the difficulties inherent in upgrading a fixed size wire protocol designed for a research network fifty years ago, and the widespread adoption of ipv6 for real customer deployments. Heck most of you are probably using ipv6 through your mobile carrier and don’t even know it!

gnfargbl 4y ago
I think you're being unfair to the audience. The main complaints here aren't about the consequences of those legacy problems, they're about the fact that IPv6 is just such a mediocre designed-by-committee 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.
- zamadatix 4y ago
  IPv6 is very similar to IPv4 it's just IPv4 wasn't a consumer grade protocol either. 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.
  - gnfargbl 4y ago
    Again, this is kind of making my point. IPv4 had an excuse for being a bit crap, because it was designed in the early 1980s. IPv6 should have been much better than it is. 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.
    - keiyakins 4y ago
      IPv6 was designed in the late 80s/early 90s, before home users had internet access. At the time there was serious debate over switching to the OSI stack instead. You're showing your lack of history knowledge more than anything.
    - NegativeK 4y ago
      IPv6's introduction is closer to IPv4's introduction that to now. By nearly twice as much.
    - ipython 4y ago
      It should be mentioned that 2^32 was widely considered to be unfathomongly large when ipv4 was developed as well. 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.
    - zamadatix 4y ago
      I could say IPv6 addresses should have been 8 bits because it would have been more usable but that's not reasoning I'm right and IPv6 is a mediocre design-by-committee protocol it's just me lambasting the designers based on some number I thought sounded nice. 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).
- ipython 4y ago
  What parts of ipv6 used in consumer grade equipment are not understandable or “consumer grade”? Ipv6 in some areas is actually simpler than ipv4. 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.
  - mianos 4y ago
    It is worse than that. They implement 'carrier grade NAT' on top of double NAT. 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.
- mplabelspace 4y ago
  >I think you're being unfair to the audience. The main complaints here aren't >about the consequences of those legacy problems, they're about the fact that IPv6 >is just such a mediocre designed-by-committee protocol. 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.
usr1106 4y ago
> most of you are probably using ipv6 through your mobile carrier and don’t even know 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?
- wmf 4y ago
  T-Mobile has been IPv6-only for a while. I guess you can't use NAT64 at Hetzner because Hetzner doesn't provide it.
  - usr1106 4y ago
    Why would Hetzner need to provide it? If you have several IPv6-only nodes you can rune one NAT64 node yourself. Of course that would work only if you save enough on IPv4 to pay for the extra node. And not if you are heavily network-bound. 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.
- ipython 4y ago
  Yes, T-Mobile for example. I just dumped the interface information from my iPhone and the only address available is an ipv6 address. Everything else is put through an ipv6 to ipv4 gateway. Works fine. Apple has required app developers to support ipv6 only deployments since 2015: https://developer.apple.com/support/ipv6/ https://developer.apple.com/support/ipv6/
- hinata08 4y ago
  I had IPv6 only on the default APN configuration pushed by my carrier on Android, last time I checked it And IPv4 only for roaming Both v6 only, v4 only and v4+v6 were supported 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 ?
AtlasBarfed 4y ago
You're essentially blaming the victim. 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.
- Symbiote 4y ago
  45% of users are on IPv6. Clearly almost half of ISPs have adopted it, probably more than half of consumer ISPs.
  - AtlasBarfed 4y ago
    How many of those are Mobile? 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.

gnfargbl 4y ago

I think you're being unfair to the audience. The main complaints here aren't about the consequences of those legacy problems, they're about the fact that IPv6 is just such a mediocre designed-by-committee 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.

zamadatix 4y ago
IPv6 is very similar to IPv4 it's just IPv4 wasn't a consumer grade protocol either. 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.
- gnfargbl 4y ago
  Again, this is kind of making my point. IPv4 had an excuse for being a bit crap, because it was designed in the early 1980s. IPv6 should have been much better than it is. 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.
  - keiyakins 4y ago
    IPv6 was designed in the late 80s/early 90s, before home users had internet access. At the time there was serious debate over switching to the OSI stack instead. You're showing your lack of history knowledge more than anything.
  - NegativeK 4y ago
    IPv6's introduction is closer to IPv4's introduction that to now. By nearly twice as much.
  - ipython 4y ago
    It should be mentioned that 2^32 was widely considered to be unfathomongly large when ipv4 was developed as well. 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.
  - zamadatix 4y ago
    I could say IPv6 addresses should have been 8 bits because it would have been more usable but that's not reasoning I'm right and IPv6 is a mediocre design-by-committee protocol it's just me lambasting the designers based on some number I thought sounded nice. 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).
ipython 4y ago
What parts of ipv6 used in consumer grade equipment are not understandable or “consumer grade”? Ipv6 in some areas is actually simpler than ipv4. 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.
- mianos 4y ago
  It is worse than that. They implement 'carrier grade NAT' on top of double NAT. 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.
mplabelspace 4y ago
>I think you're being unfair to the audience. The main complaints here aren't >about the consequences of those legacy problems, they're about the fact that IPv6 >is just such a mediocre designed-by-committee protocol. 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.

usr1106 4y ago

> most of you are probably using ipv6 through your mobile carrier and don’t even know 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?

wmf 4y ago
T-Mobile has been IPv6-only for a while. I guess you can't use NAT64 at Hetzner because Hetzner doesn't provide it.
- usr1106 4y ago
  Why would Hetzner need to provide it? If you have several IPv6-only nodes you can rune one NAT64 node yourself. Of course that would work only if you save enough on IPv4 to pay for the extra node. And not if you are heavily network-bound. 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.
ipython 4y ago
Yes, T-Mobile for example. I just dumped the interface information from my iPhone and the only address available is an ipv6 address. Everything else is put through an ipv6 to ipv4 gateway. Works fine. Apple has required app developers to support ipv6 only deployments since 2015: https://developer.apple.com/support/ipv6/ https://developer.apple.com/support/ipv6/
hinata08 4y ago
I had IPv6 only on the default APN configuration pushed by my carrier on Android, last time I checked it And IPv4 only for roaming Both v6 only, v4 only and v4+v6 were supported 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 ?

AtlasBarfed 4y ago

You're essentially blaming the victim. 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.

Symbiote 4y ago
45% of users are on IPv6. Clearly almost half of ISPs have adopted it, probably more than half of consumer ISPs.
- AtlasBarfed 4y ago
  How many of those are Mobile? 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.