3 ms·

I've had less than 0.5% of customers ask for IPv6 from my fibre ISP. It's not worth supporting as a result. The main reason is that any service that is not wi

by bcrl 1y ago

I've had less than 0.5% of customers ask for IPv6 from my fibre ISP. It's not worth supporting as a result. The main reason is that any service that is not widely used will have gremlins that result in poor customer experience, and if it's always the same handful of customers hitting problems or finding quirks, there is a real risk of poor word of mouth incident reporting that can harm the business. At least if something goes wrong with IPv4, it's going to be noticed very quickly. Some people will say monitoring is all that you need, but I do not agree. There are a million different little issues that can and do occur on physical networks in the real world, and there's no way monitoring will have a 99% chance of detecting all of them. When incidents like the partial Microsoft network outage that hit certain peering points occurred, I had to route around the damage by tweaking route filtering on the core routers to prefer a transit connection that worked over the lower cost peering point. It's that kind of oddball issue that active users catch and report which does not happen for barely used services like IPv6.

Sanzig 1y ago
If you already have to do CGNAT, why not IPv6 as your core network with NAT64 at the border and 464XLAT on the CPE? It gives you best of both worlds.
- bcrl 1y ago
  I'm not doing CGNAT. We were able to get enough IPv4 addresses directly from ARIN a few years ago after being on the waiting list for a couple of years. It's a pity that widespread fraud depleted that pool faster than it should have been. CPE support for IPv6 has generally been garbage with it taking 15-20 years before the bare minimum was supported by mainstream router vendors. Even today there are still vendors that assume only IPv4 support. In my opinion the IETF really screwed up when they made IPv6 more complicated than just IPv4 with more address bits. The incumbent in my area generally uses PPPoE in their access network, but routers that supported PPPoE and prefix delegation basically didn't exist in 2010, and only started being available circa 2015 (in part due to the required bits not existing in OpenWRT and the hardware vendors' software development kits for their chipsets). Sure, we're 10 years further on now, but there remain a number of vendors that only support IPv4 for management of devices (cough Ubiquiti cough) in parts of their product line. That said, there are features of IPv6 that are absolutely awesome for carriers. The next header feature that pretty much eliminates the need for MPLS in an IPv6 transport network is one such item that makes building transport networks so much cleaner when using IPv6 than IPv4. No more header insertion or rewriting, just update one field and fix up the delta on the checksum and CRC. They just aren't really applicable for smaller networks.
  - Sanzig 1y ago
    Ah, okay, well if you already have the IPv4 address space available for all your customers that's a different story. I can understand why you want to wait to dual-stack in that case. I do think NAT64/464XLAT is a pretty good architecture for new ISPs that can't get their hands on IPv4 space, though. Or even MAP-T, but CPE support isn't really there yet.
    - bcrl 1y ago
      CPE support is the issue and has always been the issue! People assume that IPv6 is done, but it really isn't in the consumer networking gear product space.
- supertrope 1y ago
  Android and iOS have pressure on app developers to support IPv6 or at the very least function on 464XLAT. On home broadband people could connect anything including programs that use literal IPv4 addresses and break on 464XLAT. Things like corporate VPNs, Skype, online gaming. Ironically these services would benefit the most from supporting IPv6 but they've evolved to deal with IPv4 NAT to the extent that they've become dependent on IPv4 connectivity.
  - bcrl 1y ago
    Great: your client device supports IPv6 while your consumer networking gear does not. How's that going to work exactly?
    - supertrope 1y ago
      Most people just use what their ISP issues. Comcast supports IPv6 and so does their CPE.
mrweasel 1y ago
> I've had less than 0.5% of customers ask for IPv6 from my fibre ISP How many ask for IPv4? I understand your situation, it's a lot of work, for something that many won't notice. It's just that saying there's no demand because your average consumer, who also doesn't know what IPv4 is, isn't asking for it, is the mentality that keeps IPv6 from being implemented. On the funnier side of things, we've also sometimes run into the opposite problem that we can't reproduce an issue, because it's only on IPv4 and 95% of the time everything we do is IPv6. But we're also not serving home users.
- bcrl 1y ago
  Static IPv4 addresses are closer to around 5% of customers. Nobody asks for IPv4, but some customers bring their existing or own wireless routers along and occasionally choose devices that are not IPv6 capable. Maybe in another 10 years those devices will finally be fully removed from service. The worst stragglers right now are the old combo DSL modems that effectively have no modern replacements -- it's just not worth spending money to replace them when customers are going to migrate to fibre soon enough.
  - nulbyte 1y ago
    I don't think GP meant static addressing but literally, how many ask for IPv4 service? None, because you just provide it; it's an unstated expectation. Now apply that to IPv6 and you can see the point that (I think) GP is making. Side note: The claim it is not widely used doesn't track. How many people use Google or Facebook? More than half of that traffic is over IPv6. https://circleid.com/posts/ipv6-usage-in-the-u.s-surpasses-50-but-lags-behind-global-leaders https://circleid.com/posts/ipv6-usage-in-the-u.s-surpasses-5...
    - bcrl 1y ago
      I know that the GP wasn't asking about static IP addresses, however I was merely providing some statistics about what customer asks regarding IP addresses are actually seen on the provider side of things. It's not widely used in my customer base as well as among a number of other small independent ISPs I know of. A significant number of the CPE we have deployed today do not support IPv6, therefore a significant number of our end users will not be using IPv6 resulting in only partial usage. That is what I am referring to: an incomplete deployment of IPv6 across a customer base without their knowledge of change and the end user's ability to identify issues will cause support issues which costs money and time. Because of how IPv6 is provisioned, going from a no IPv6-deployed state to fully-deployed state will encounter transition issues. It's not provisioned the same way as IPv4, and that's the whole bloody problem!!! Sure, I can turn on handing out IPv6 addresses from our edge routers via PPPoE today, but it's not going to "Just Work" for all customers without CPE config changes (assuming the CPE even supports IPv6 in the first place for which a bunch of them do not!). It's a complete pain that has less than zero benefit for a small provider today since everything is available over IPv4 and nothing is IPv6 only. Heck, another ISP I know had a customer ask for IPv6 support, and then the customer didn't even bother to provision it on their own router! Another example: take Ubiquiti's AirCube product line. It doesn't support IPv6. To deploy IPv6 it would be necessary to reconfigure things to make the ONU for the customer act as the router for the customer's home network and place the AirCube in bridged mode. That's assuming that the ONU even supports IPv6 yet as it did not when I looked into it on the order of 6 years ago. And yes, there are a few dozen customers on AirCubes that fall under this case. This means moving into a territory where multiple different CPE deployment configs are now required for less than zero benefit. That there are gaps in IPv6 support has been a problem, is a problem, and will continue to be a problem. The industry isn't 100% IPv6, and I don't see how that is going to be the case in the next 5 years or even the next 10 years. Consumer gear isn't there yet. That was the problem 5 years ago, that was the problem 10 years ago, that was the problem 15 years, and that was the problem 20 years ago... What's different today that's any different than yesterday? It's easy to deploy IPv6 in data centers where you control all the infrastructure on servers where operating system IPv6 support was mature decades ago. It's not so easy as a small ISP when dealing with pile of random consumer gear that an ISP's customers bring to the table where IPv6 support remains hit and miss.
massysett 1y ago
> I've had less than 0.5% of customers ask for IPv6 from my fibre ISP. It's not worth supporting as a result. Big, evil, hated Comcast has full ipv6, and I doubt any of its customers asked for it either. Instead people complain they’re only getting a /60.
- bcrl 1y ago
  Comcast was forced to go to IPv6 because they ran out of IPv4 addresses in the private address space to use for management of their network (think of how each and every cable modem needs a an address for management in addition to all the routers and CMTSes). I was a fly on the way inside one of the router vendors when this took place more than 15 years ago.

by bcrl 1y ago

Sanzig 1y ago

If you already have to do CGNAT, why not IPv6 as your core network with NAT64 at the border and 464XLAT on the CPE? It gives you best of both worlds.

bcrl 1y ago
I'm not doing CGNAT. We were able to get enough IPv4 addresses directly from ARIN a few years ago after being on the waiting list for a couple of years. It's a pity that widespread fraud depleted that pool faster than it should have been. CPE support for IPv6 has generally been garbage with it taking 15-20 years before the bare minimum was supported by mainstream router vendors. Even today there are still vendors that assume only IPv4 support. In my opinion the IETF really screwed up when they made IPv6 more complicated than just IPv4 with more address bits. The incumbent in my area generally uses PPPoE in their access network, but routers that supported PPPoE and prefix delegation basically didn't exist in 2010, and only started being available circa 2015 (in part due to the required bits not existing in OpenWRT and the hardware vendors' software development kits for their chipsets). Sure, we're 10 years further on now, but there remain a number of vendors that only support IPv4 for management of devices (cough Ubiquiti cough) in parts of their product line. That said, there are features of IPv6 that are absolutely awesome for carriers. The next header feature that pretty much eliminates the need for MPLS in an IPv6 transport network is one such item that makes building transport networks so much cleaner when using IPv6 than IPv4. No more header insertion or rewriting, just update one field and fix up the delta on the checksum and CRC. They just aren't really applicable for smaller networks.
- Sanzig 1y ago
  Ah, okay, well if you already have the IPv4 address space available for all your customers that's a different story. I can understand why you want to wait to dual-stack in that case. I do think NAT64/464XLAT is a pretty good architecture for new ISPs that can't get their hands on IPv4 space, though. Or even MAP-T, but CPE support isn't really there yet.
  - bcrl 1y ago
    CPE support is the issue and has always been the issue! People assume that IPv6 is done, but it really isn't in the consumer networking gear product space.
supertrope 1y ago
Android and iOS have pressure on app developers to support IPv6 or at the very least function on 464XLAT. On home broadband people could connect anything including programs that use literal IPv4 addresses and break on 464XLAT. Things like corporate VPNs, Skype, online gaming. Ironically these services would benefit the most from supporting IPv6 but they've evolved to deal with IPv4 NAT to the extent that they've become dependent on IPv4 connectivity.
- bcrl 1y ago
  Great: your client device supports IPv6 while your consumer networking gear does not. How's that going to work exactly?
  - supertrope 1y ago
    Most people just use what their ISP issues. Comcast supports IPv6 and so does their CPE.

mrweasel 1y ago

> I've had less than 0.5% of customers ask for IPv6 from my fibre ISP How many ask for IPv4? I understand your situation, it's a lot of work, for something that many won't notice. It's just that saying there's no demand because your average consumer, who also doesn't know what IPv4 is, isn't asking for it, is the mentality that keeps IPv6 from being implemented. On the funnier side of things, we've also sometimes run into the opposite problem that we can't reproduce an issue, because it's only on IPv4 and 95% of the time everything we do is IPv6. But we're also not serving home users.

bcrl 1y ago
Static IPv4 addresses are closer to around 5% of customers. Nobody asks for IPv4, but some customers bring their existing or own wireless routers along and occasionally choose devices that are not IPv6 capable. Maybe in another 10 years those devices will finally be fully removed from service. The worst stragglers right now are the old combo DSL modems that effectively have no modern replacements -- it's just not worth spending money to replace them when customers are going to migrate to fibre soon enough.
- nulbyte 1y ago
  I don't think GP meant static addressing but literally, how many ask for IPv4 service? None, because you just provide it; it's an unstated expectation. Now apply that to IPv6 and you can see the point that (I think) GP is making. Side note: The claim it is not widely used doesn't track. How many people use Google or Facebook? More than half of that traffic is over IPv6. https://circleid.com/posts/ipv6-usage-in-the-u.s-surpasses-50-but-lags-behind-global-leaders https://circleid.com/posts/ipv6-usage-in-the-u.s-surpasses-5...
  - bcrl 1y ago
    I know that the GP wasn't asking about static IP addresses, however I was merely providing some statistics about what customer asks regarding IP addresses are actually seen on the provider side of things. It's not widely used in my customer base as well as among a number of other small independent ISPs I know of. A significant number of the CPE we have deployed today do not support IPv6, therefore a significant number of our end users will not be using IPv6 resulting in only partial usage. That is what I am referring to: an incomplete deployment of IPv6 across a customer base without their knowledge of change and the end user's ability to identify issues will cause support issues which costs money and time. Because of how IPv6 is provisioned, going from a no IPv6-deployed state to fully-deployed state will encounter transition issues. It's not provisioned the same way as IPv4, and that's the whole bloody problem!!! Sure, I can turn on handing out IPv6 addresses from our edge routers via PPPoE today, but it's not going to "Just Work" for all customers without CPE config changes (assuming the CPE even supports IPv6 in the first place for which a bunch of them do not!). It's a complete pain that has less than zero benefit for a small provider today since everything is available over IPv4 and nothing is IPv6 only. Heck, another ISP I know had a customer ask for IPv6 support, and then the customer didn't even bother to provision it on their own router! Another example: take Ubiquiti's AirCube product line. It doesn't support IPv6. To deploy IPv6 it would be necessary to reconfigure things to make the ONU for the customer act as the router for the customer's home network and place the AirCube in bridged mode. That's assuming that the ONU even supports IPv6 yet as it did not when I looked into it on the order of 6 years ago. And yes, there are a few dozen customers on AirCubes that fall under this case. This means moving into a territory where multiple different CPE deployment configs are now required for less than zero benefit. That there are gaps in IPv6 support has been a problem, is a problem, and will continue to be a problem. The industry isn't 100% IPv6, and I don't see how that is going to be the case in the next 5 years or even the next 10 years. Consumer gear isn't there yet. That was the problem 5 years ago, that was the problem 10 years ago, that was the problem 15 years, and that was the problem 20 years ago... What's different today that's any different than yesterday? It's easy to deploy IPv6 in data centers where you control all the infrastructure on servers where operating system IPv6 support was mature decades ago. It's not so easy as a small ISP when dealing with pile of random consumer gear that an ISP's customers bring to the table where IPv6 support remains hit and miss.

massysett 1y ago

> I've had less than 0.5% of customers ask for IPv6 from my fibre ISP. It's not worth supporting as a result. Big, evil, hated Comcast has full ipv6, and I doubt any of its customers asked for it either. Instead people complain they’re only getting a /60.

bcrl 1y ago
Comcast was forced to go to IPv6 because they ran out of IPv4 addresses in the private address space to use for management of their network (think of how each and every cable modem needs a an address for management in addition to all the routers and CMTSes). I was a fly on the way inside one of the router vendors when this took place more than 15 years ago.