9 ms·

I guess this is related to the recent kerfuffle about to making (most of) 127/8 publicly routable to stave off the IPv4 shortage a little while longer. We’re j

by mikl 4y ago

I guess this is related to the recent kerfuffle about to making (most of) 127/8 publicly routable to stave off the IPv4 shortage a little while longer. We’re just delaying the inevitable. Sooner or later, the band-aid has got to come off. ~4 billion IP addresses are simply not enough for ~8 billion people. So maybe we should just refuse to pile on more hacks and force the ISPs to get off their ass and upgrade their networks.

Arubis 4y ago
I hadn't heard about that proposal; do you have a link to discussion/source? It sounds like a Bad Idea and I'm curious for the arguments in favor.
- placatedmayhem 4y ago
  https://www.ietf.org/archive/id/draft-schoen-intarea-unicast-127-00.html https://www.ietf.org/archive/id/draft-schoen-intarea-unicast...
  - Arubis 4y ago
    Thank you. This is wild.
  - blibble 4y ago
    that seems completely insane that logic has been hard-coded inside every OS and vast amounts of software for 30+ years and to gain less than 0.8%? they'd do better targeting the reserved multicast range
    - justsomehnguy 4y ago
      > inside every OS and vast amounts of software for 30+ years Ha. Ha. Ha. > that logic has been hard-coded inside The thing is what some things are literally hardcoded. I have a whole bunch of network gear circa 2010 in production. It does support IPv6, but for the most part it is extremely rudimentary. What do you want from a box what implements SSH by running a telnet session over a no-auth SSH session? And while I can run the latest Windows/Linux/Whatever on my server gear, the network gear is not that simple to upgrade.
      - vlan0 4y ago
        >I have a whole bunch of network gear circa 2010 in production I ran into this 10 years ago. I decided to roll out IPv6 to clients. Router docs said it was supported. Then I watched CPU usage peg because every packet hit the CPU. Turns out it was implemented in software; no ASIC support. Was quite the buzz kill.
        p_l 4y ago
        Now consider that there are network components in the wild that have literal latch on first octet equaling 127 which triggers packet drop, and this proposal could drop hw acceleration for v4 too
    - greyface- 4y ago
      > they'd do better targeting the reserved multicast range https://www.ietf.org/archive/id/draft-schoen-intarea-unicast-240-02.html https://www.ietf.org/archive/id/draft-schoen-intarea-unicast...
  - kortex 4y ago
    What in the actual eff. Anything except actually switching fully to ipv6, amirite? Trying to see the most charitable position here - is there an aspect in which redefining 127/8 to 127/16 across billions of devices is actually feasible?
    - accountofme 4y ago
      I used to work tech support for an ISP years ago.... Trying to get people to find their modem and see if it was plugged into the wall was effort. Imagine trying to get the same people to upgrade router firmware...... no chance
      - jSherz 4y ago
        I assume that most ISPs push out updates automatically unless the hardware is user provided (and thus user supported)? I've been through several rounds of "we're sending you a new router" with UK ISPs where you're aggressively reminded to switch.
      - judge2020 4y ago
        This is why the ISP routers and modems have all introduced remote updates from the ISP, sometimes going further with extra telemetry[0]. Not a great situation for those with the technical knowledge, but I don't think we can expect people to upgrade their router FW on their own. 0: https://old.reddit.com/r/ATT/comments/mqfy9a/just_got_our_new_bgw320_installed_however_the/guhaanr/ https://old.reddit.com/r/ATT/comments/mqfy9a/just_got_our_ne...
        wyager 4y ago
        I was pissed to learn that arris DOCSIS modems do not allow you, the owner, to install firmware updates or adjust most settings. Only the ISP gets to do that.
        X-Istence 4y ago
        That's because bandwidth caps are enforced by the cable modem and not further downstream. When you pay for 10 Mbps down, 1 Mbps up, that is enforced by the cable modem. Earlier versions of cable modems had flaws that allowed consumers to push configs on the ethernet port and uncap their cable modems and gain higher speeds than they had paid for.
  - bombcar 4y ago
    A slightly more insane but maybe a bit more practical would have been to try to move 10.x and friends into 127.x.
  - Tsiklon 4y ago
    Ohh wow - I didn't think there were any serious proposals for this, this sounds like madness.
  - twic 4y ago
    Expires: 12 May 2022 I couldn't find a newer edition, so hopefully this has gone to the great standards track in the sky.
    - greyface- 4y ago
      Here's a newer edition! :) https://www.ietf.org/archive/id/draft-schoen-intarea-unicast-127-01.html https://www.ietf.org/archive/id/draft-schoen-intarea-unicast... Expires: 8 September 2022
1ncorrect 4y ago
Any effort that may be applied to reclaiming v4 address space will always be better directed to expanding global v6 connectivity.
- yardstick 4y ago
  Wouldn’t it involve mostly different groups of people? IPv6 is now supported in the vast majority of operating systems and routers and firewalls. It’s now down to the carriers/ISPs to configure their networks to use it. Ie the IPv4 issue is a programming issue, while the IPv6 issue is an administration/configuration issue. Ie Ciscos software engineers aren’t the ones holding back your ISP from turning on IPv6. Cisco engineers aren’t toiling away adding much needed missing IPv6 features. Unless I’ve missed something?
  - nousermane 4y ago
    It's not that somebody is holding your ISP, it's that your ISP doesn't see IPv6 as something that they should bother doing (very low demand from customers). And if that ISP happens to be one of those that decided that their core competency is marketing, and outsourced everything else, including network engineering... Well, than they view IPv6 as something expensive/risky (in project costs) in addition to useless.
    - yardstick 4y ago
      No disagreement there.
    - dijit 4y ago
      > very low demand from customers This actually pisses me off enormously. It’s not that there’s low (or no) demand, it’s that people are not being recorded as having voiced this concern. I can say for certain that over the last ten years when ISPs call to upsell their packages I always respond with “I’m not going to upgrade unless it gives me ipv6”. When talking to new ISPs I always ask if they support ipv6 and if not: when. The answer is always “it’s not on the roadmap, we don’t have any people requesting this”- the follow up I usually ask is “are you going to write down anywhere that I am requesting this? Because others could say the same” and they always say that they won’t record my interest, even though it costs them an upsell or a sale. So, I figured maybe I’m weird, maybe I am the only one. I have no way of knowing. Except; I met someone else who also does this- we have the same ISP and they told him: “nobody has ever asked for this” which is categorically untrue since I have asked for this. We’re being gaslit by the industry.
b112 4y ago
ipv6 is untenable. We need ipv4.1. Just add another octlet at the start, but stored in the options field. Old stuff won't get confused, and new stuff will know that empty means 1.x.x.x.x. Problem solved.
- kortex 4y ago
  How does that solve anything? Any hardware is gonna try to route packets to the "original" x.x.x.x and completely ignore this option field.
- zokier 4y ago
  Over third of the world is happily using that untenable ipv6. That train left the station long time ago and is chugging along at steady pace now.
- dtech 4y ago
  I fail to see how that is better than the already widely used NAT
- Beltalowda 4y ago
  The problem with this is that it requires hardware upgrades across the world: your computer, router, all the world's switches, a lot of software (e.g. software that stores IP addresses as uint32, but also native types that map to uint32), DNS infrastructure, etc. etc. etc. This is basically the same problem as IPv6, except worse since we'll start from zero. In hindsight, clearly a better plan than "this is IPv6, it's new, it's better, everyone will adopt it!" was needed for IPv6, and it would have helped if the IPv6 people were a bit more pragmatic from the start in their design (religious opposition to NAT springs to mind), but now it's about 20 to 25 years too late.
  - b112 4y ago
    Easy to fix. All we have to do, is explain that ipv6 reduces anonymity online, and thus, makes it easier for predators to track children. And thus, ipv6 endangers children! This will enable a wellspring of governement grants, to roll out firmware updates for switches and routers, etc. See? A non-problem.
deknos 4y ago
i would LOVE it, if there was some project which would test popular appliances and linux at home, if v6(only) presents any problems...
heretogetout 4y ago
What are users of these ISPs missing out on my not having IPv6? They need to rely on NAT for their devices, but that works fine as far as they know. Their IoT devices need to connect to remote servers instead of wait for remote servers to contact them directly, but that won't change how users interact with their devices so they won't care. Maybe this isn't a serious enough crisis and we can just wait til it becomes a real issue for users. When it does, it will probably be incremental rather than catastrophic, so we'll still have time to upgrade networks piecemeal, and with more reason to do so.
- throw0101a 4y ago
  > What are users of these ISPs missing out on my not having IPv6? They need to rely on NAT for their devices […] This assumes that they can even get an IPv4 address. New ISPs and telcos may not be able to get enough IPv4 addresses to assign each individual connection one. So either (a) the CPE only gets an IPv6 address and they have to use DNS64/NAT64 or 464XLAT to reach the IPv4 Internet, or (b) the CPE gets a shared address space (100.64.0.0/10) IPv4 address and CG-NAT ensues: * https://en.wikipedia.org/wiki/IPv4_shared_address_space https://en.wikipedia.org/wiki/IPv4_shared_address_space * https://en.wikipedia.org/wiki/Carrier-grade_NAT https://en.wikipedia.org/wiki/Carrier-grade_NAT In which case the end-user gets to deal with double-NATing (first with their CPE/home router, and second by the ISP). Case in point someone that is surprised that every customer gets a public IP address and CG-NAT is not done: * https://old.reddit.com/r/networking/comments/uw0qwy/ https://old.reddit.com/r/networking/comments/uw0qwy/
vlan0 4y ago
>So maybe we should just refuse to pile on more hacks and force the ISPs to get off their ass and upgrade their networks. I haven't recently looked at who has deployed what, but is this really it? US ISPs like Spectrum has been handing out IPv6 addresses for years now. My Verizon wireless phone can access IPv6 resources w/o issue as well. However, show me the cloud services folks are implementing that are dual stack. Start mandating new service deployment is dual stack, and it'll happen. Similar to how the US Govt is pulling it off.
- WorldMaker 4y ago
  Right, consumer ISPs seem to be driving IPv6 adoption not lagging it. It's Big Enterprise, including all the supposedly "Cloud Native" companies, dragging their feet real hard right now.

by mikl 4y ago

Arubis 4y ago

I hadn't heard about that proposal; do you have a link to discussion/source? It sounds like a Bad Idea and I'm curious for the arguments in favor.

placatedmayhem 4y ago
https://www.ietf.org/archive/id/draft-schoen-intarea-unicast-127-00.html https://www.ietf.org/archive/id/draft-schoen-intarea-unicast...
- Arubis 4y ago
  Thank you. This is wild.
- blibble 4y ago
  that seems completely insane that logic has been hard-coded inside every OS and vast amounts of software for 30+ years and to gain less than 0.8%? they'd do better targeting the reserved multicast range
  - justsomehnguy 4y ago
    > inside every OS and vast amounts of software for 30+ years Ha. Ha. Ha. > that logic has been hard-coded inside The thing is what some things are literally hardcoded. I have a whole bunch of network gear circa 2010 in production. It does support IPv6, but for the most part it is extremely rudimentary. What do you want from a box what implements SSH by running a telnet session over a no-auth SSH session? And while I can run the latest Windows/Linux/Whatever on my server gear, the network gear is not that simple to upgrade.
    - vlan0 4y ago
      >I have a whole bunch of network gear circa 2010 in production I ran into this 10 years ago. I decided to roll out IPv6 to clients. Router docs said it was supported. Then I watched CPU usage peg because every packet hit the CPU. Turns out it was implemented in software; no ASIC support. Was quite the buzz kill.
      - p_l 4y ago
        Now consider that there are network components in the wild that have literal latch on first octet equaling 127 which triggers packet drop, and this proposal could drop hw acceleration for v4 too
  - greyface- 4y ago
    > they'd do better targeting the reserved multicast range https://www.ietf.org/archive/id/draft-schoen-intarea-unicast-240-02.html https://www.ietf.org/archive/id/draft-schoen-intarea-unicast...
- kortex 4y ago
  What in the actual eff. Anything except actually switching fully to ipv6, amirite? Trying to see the most charitable position here - is there an aspect in which redefining 127/8 to 127/16 across billions of devices is actually feasible?
  - accountofme 4y ago
    I used to work tech support for an ISP years ago.... Trying to get people to find their modem and see if it was plugged into the wall was effort. Imagine trying to get the same people to upgrade router firmware...... no chance
    - jSherz 4y ago
      I assume that most ISPs push out updates automatically unless the hardware is user provided (and thus user supported)? I've been through several rounds of "we're sending you a new router" with UK ISPs where you're aggressively reminded to switch.
    - judge2020 4y ago
      This is why the ISP routers and modems have all introduced remote updates from the ISP, sometimes going further with extra telemetry[0]. Not a great situation for those with the technical knowledge, but I don't think we can expect people to upgrade their router FW on their own. 0: https://old.reddit.com/r/ATT/comments/mqfy9a/just_got_our_new_bgw320_installed_however_the/guhaanr/ https://old.reddit.com/r/ATT/comments/mqfy9a/just_got_our_ne...
      - wyager 4y ago
        I was pissed to learn that arris DOCSIS modems do not allow you, the owner, to install firmware updates or adjust most settings. Only the ISP gets to do that.
        X-Istence 4y ago
        That's because bandwidth caps are enforced by the cable modem and not further downstream. When you pay for 10 Mbps down, 1 Mbps up, that is enforced by the cable modem. Earlier versions of cable modems had flaws that allowed consumers to push configs on the ethernet port and uncap their cable modems and gain higher speeds than they had paid for.
- bombcar 4y ago
  A slightly more insane but maybe a bit more practical would have been to try to move 10.x and friends into 127.x.
- Tsiklon 4y ago
  Ohh wow - I didn't think there were any serious proposals for this, this sounds like madness.
- twic 4y ago
  Expires: 12 May 2022 I couldn't find a newer edition, so hopefully this has gone to the great standards track in the sky.
  - greyface- 4y ago
    Here's a newer edition! :) https://www.ietf.org/archive/id/draft-schoen-intarea-unicast-127-01.html https://www.ietf.org/archive/id/draft-schoen-intarea-unicast... Expires: 8 September 2022

1ncorrect 4y ago

Any effort that may be applied to reclaiming v4 address space will always be better directed to expanding global v6 connectivity.

yardstick 4y ago
Wouldn’t it involve mostly different groups of people? IPv6 is now supported in the vast majority of operating systems and routers and firewalls. It’s now down to the carriers/ISPs to configure their networks to use it. Ie the IPv4 issue is a programming issue, while the IPv6 issue is an administration/configuration issue. Ie Ciscos software engineers aren’t the ones holding back your ISP from turning on IPv6. Cisco engineers aren’t toiling away adding much needed missing IPv6 features. Unless I’ve missed something?
- nousermane 4y ago
  It's not that somebody is holding your ISP, it's that your ISP doesn't see IPv6 as something that they should bother doing (very low demand from customers). And if that ISP happens to be one of those that decided that their core competency is marketing, and outsourced everything else, including network engineering... Well, than they view IPv6 as something expensive/risky (in project costs) in addition to useless.
  - yardstick 4y ago
    No disagreement there.
  - dijit 4y ago
    > very low demand from customers This actually pisses me off enormously. It’s not that there’s low (or no) demand, it’s that people are not being recorded as having voiced this concern. I can say for certain that over the last ten years when ISPs call to upsell their packages I always respond with “I’m not going to upgrade unless it gives me ipv6”. When talking to new ISPs I always ask if they support ipv6 and if not: when. The answer is always “it’s not on the roadmap, we don’t have any people requesting this”- the follow up I usually ask is “are you going to write down anywhere that I am requesting this? Because others could say the same” and they always say that they won’t record my interest, even though it costs them an upsell or a sale. So, I figured maybe I’m weird, maybe I am the only one. I have no way of knowing. Except; I met someone else who also does this- we have the same ISP and they told him: “nobody has ever asked for this” which is categorically untrue since I have asked for this. We’re being gaslit by the industry.

b112 4y ago

ipv6 is untenable. We need ipv4.1. Just add another octlet at the start, but stored in the options field. Old stuff won't get confused, and new stuff will know that empty means 1.x.x.x.x. Problem solved.

kortex 4y ago
How does that solve anything? Any hardware is gonna try to route packets to the "original" x.x.x.x and completely ignore this option field.
zokier 4y ago
Over third of the world is happily using that untenable ipv6. That train left the station long time ago and is chugging along at steady pace now.
dtech 4y ago
I fail to see how that is better than the already widely used NAT
Beltalowda 4y ago
The problem with this is that it requires hardware upgrades across the world: your computer, router, all the world's switches, a lot of software (e.g. software that stores IP addresses as uint32, but also native types that map to uint32), DNS infrastructure, etc. etc. etc. This is basically the same problem as IPv6, except worse since we'll start from zero. In hindsight, clearly a better plan than "this is IPv6, it's new, it's better, everyone will adopt it!" was needed for IPv6, and it would have helped if the IPv6 people were a bit more pragmatic from the start in their design (religious opposition to NAT springs to mind), but now it's about 20 to 25 years too late.
- b112 4y ago
  Easy to fix. All we have to do, is explain that ipv6 reduces anonymity online, and thus, makes it easier for predators to track children. And thus, ipv6 endangers children! This will enable a wellspring of governement grants, to roll out firmware updates for switches and routers, etc. See? A non-problem.

deknos 4y ago

i would LOVE it, if there was some project which would test popular appliances and linux at home, if v6(only) presents any problems...

heretogetout 4y ago

What are users of these ISPs missing out on my not having IPv6? They need to rely on NAT for their devices, but that works fine as far as they know. Their IoT devices need to connect to remote servers instead of wait for remote servers to contact them directly, but that won't change how users interact with their devices so they won't care. Maybe this isn't a serious enough crisis and we can just wait til it becomes a real issue for users. When it does, it will probably be incremental rather than catastrophic, so we'll still have time to upgrade networks piecemeal, and with more reason to do so.

throw0101a 4y ago
> What are users of these ISPs missing out on my not having IPv6? They need to rely on NAT for their devices […] This assumes that they can even get an IPv4 address. New ISPs and telcos may not be able to get enough IPv4 addresses to assign each individual connection one. So either (a) the CPE only gets an IPv6 address and they have to use DNS64/NAT64 or 464XLAT to reach the IPv4 Internet, or (b) the CPE gets a shared address space (100.64.0.0/10) IPv4 address and CG-NAT ensues: * https://en.wikipedia.org/wiki/IPv4_shared_address_space https://en.wikipedia.org/wiki/IPv4_shared_address_space * https://en.wikipedia.org/wiki/Carrier-grade_NAT https://en.wikipedia.org/wiki/Carrier-grade_NAT In which case the end-user gets to deal with double-NATing (first with their CPE/home router, and second by the ISP). Case in point someone that is surprised that every customer gets a public IP address and CG-NAT is not done: * https://old.reddit.com/r/networking/comments/uw0qwy/ https://old.reddit.com/r/networking/comments/uw0qwy/

vlan0 4y ago

>So maybe we should just refuse to pile on more hacks and force the ISPs to get off their ass and upgrade their networks. I haven't recently looked at who has deployed what, but is this really it? US ISPs like Spectrum has been handing out IPv6 addresses for years now. My Verizon wireless phone can access IPv6 resources w/o issue as well. However, show me the cloud services folks are implementing that are dual stack. Start mandating new service deployment is dual stack, and it'll happen. Similar to how the US Govt is pulling it off.

WorldMaker 4y ago
Right, consumer ISPs seem to be driving IPv6 adoption not lagging it. It's Big Enterprise, including all the supposedly "Cloud Native" companies, dragging their feet real hard right now.