Hacker News Board Hacker News, browsed like a classic message board
Live data from Hacker News

Using the Internet without IPv4 connectivity

jamesmcm.github.io

Post reply on HN
Print view View original

81–90 of 133 posts

Re: Using the Internet without IPv4 connectivity

#81

Post by tsimionescu »

kstrauser wrote: post #69
crims0n wrote: post #43

Earlier quoted context omitted.

We have a saying in the industry… IPv6 is an academic solution to an engineering problem. The reality is it’s just too damn complicated to implement and maintain at scale while also retaining compatibility with v4… which is never going to go away because other than the address shortage, there are no problems with it.

We have no such saying in the industry. IPv6 is generally easier to implement and maintain. If IPv6 were the incumbent and someone came in proposing IPv4, they’d be laughed out of the room for its ridiculousness. “You have to run a stateful server just to assign addresses? Dynamic header length? A tiny address space? And tell me again about this NAT thing, LOL.” V6 was designed by the engineers who realized what they…

You still need a stateful server to assign IPv6 addresses for most use cases, through DHCPv6. SLAAC doesn't even give you a DNS server yet. And even if it did, many ISPs assign too small address spaces for SLAAC, or your networks isn't so simple that you can just auto-negotiate some address.

Re: Using the Internet without IPv4 connectivity

#82

Post by cesarb »

xacky wrote: post #35

I have strong opinions about ipv4, especially since I'm forced to use an ipv4 isp. The lack of ipv6 adoption should be considered one of the great failures of tech. Who actually is responsible? Is it router manufacturers writing poor quality firmware, ipv4 advocates in leadership positions at isps, ipv4 address speculators, poor training of network engineers and tech support staff? I think we all need to have a much…

> For comparison, the internet mostly transitioned off of TLS 1.0 just fine, why can't we do the same for transitioning off ipv4?

This is a great demonstration of the advantages of the end-to-end principle. The reason the transition off TLS 1.0 (and earlier SSL 3.0) could happen so quickly is that only the endpoints (the server and the client) needed to be updated to understand the new protocol; nodes in the middle of the path (routers, switches, and so on) only needed to care about the IPv4 (or IPv6) layer, which didn't change with new TLS versions.

But that only works for layers above the network protocol; when updating the network protocol itself, every node is affected.

(And the TLS transition also took longer than it should, in large part because a lot of "middleboxes" violated the end-to-end principle by inspecting or even modifying the TLS connection, without taking part in the protocol negotiation. TLS 1.3 had to be modified to pretend to be a resumed TLS 1.2 connection to trick these middleboxes into not incorrectly rejecting the newer version of the protocol.)

Re: Using the Internet without IPv4 connectivity

#83

Post by jcgl »

tsimionescu wrote: post #81
kstrauser wrote: post #69

Earlier quoted context omitted.

We have no such saying in the industry. IPv6 is generally easier to implement and maintain. If IPv6 were the incumbent and someone came in proposing IPv4, they’d be laughed out of the room for its ridiculousness. “You have to run a stateful server just to assign addresses? Dynamic header length? A tiny address space? And tell me again about this NAT thing, LOL.” V6 was designed by the engineers who realized what they…

You still need a stateful server to assign IPv6 addresses for most use cases, through DHCPv6. SLAAC doesn't even give you a DNS server yet. And even if it did, many ISPs assign too small address spaces for SLAAC, or your networks isn't so simple that you can just auto-negotiate some address.

1. You have been able to assign DNS through SLAAC for years 2. Stateless DHCPv6 serves most needs not covered by SLAAC 3. Yeah, some ISPs screw up and don’t assign enough address space. Most likely because they’re still in the address-poverty mindset of v4

> your networks isn't so simple that you can just auto-negotiate some address

I don’t understand what you mean by this…v6 afaik has every tool that v4 does for assignment. If automated assignment through SLAAC or either kind of DHCP doesn’t meet your needs, then there’s manual assignment, just like with v4.

Re: Using the Internet without IPv4 connectivity

#84

Post by murderfs »

bigstrat2003 wrote: post #79
crims0n wrote: post #43

Earlier quoted context omitted.

We have a saying in the industry… IPv6 is an academic solution to an engineering problem. The reality is it’s just too damn complicated to implement and maintain at scale while also retaining compatibility with v4… which is never going to go away because other than the address shortage, there are no problems with it.

IPv6 is not actually hard to implement or maintain. A lot of people have repeated that meme, but it isn't true at all.

Okay, then please reimplement the equivalent of the following code to work with both IPv4 and IPv6:

    int fd = socket(AF_INET, SOCK_STREAM, 0);
    struct sockaddr_in addr = { .sin_family = AF_INET, .sin_port = htons(1234) };
    addr.sin_addr.s_addr = htonl(INADDR_ANY);
    bind(fd, (struct sockaddr*)&addr, sizeof(addr));
    listen(fd, 128);
    int client;
    while (client = accept(fd, 0, 0)) {
        // ...
    }

Re: Using the Internet without IPv4 connectivity

#85

Post by hylaride »

ornornor wrote: post #77

After all these years I still don’t see a compelling reason to spend days pulling my hair out switching all my machines and home lab to ipv6. I just find port forwarding and firewall rules more intuitive vs the prospect of spending weeks troubleshooting everything, reconfiguring firewalls, renunbering my network. What am I missing?

> What am I missing?

Right now, nothing major. At some point the big companies (Google, cloud flare, etc) may very well tire of having to pay more and more for ipv4 address that they may provide incentives for IPv6 (eg they could start throttling IPv4). There are some early moves going on already here. AWS used to only charge for unused IPv4 elastic IP addresses and now charge for them regardless of their use.

Honestly, the next time you upgrade your gateway/router you may as well set it up to be ready, but you're otherwise not missing anything right now. You can also use IPv4/v6 at the same time. You can enable it on your router and your IPv4-only devices will still work perfectly fine. One note, auto-discovery on IPv6 was a bit of a shitshow (SLAAC, IPv6 auto-addressing, and DHCPv6 all were a thing and the original auto-addressing didn't even support getting DNS servers), but things are settling on SLAAC (though ISPs will be using dhcpv6 for a loooong time).

Re: Using the Internet without IPv4 connectivity

#86

Post by School-Cotton »

msgodel wrote: post #68
simonjgreen wrote: post #3

Slightly misleading title, this is more “getting to the IPv4 internet via an IPv6 tunnel through a VPS”. Also just called 4in6. Interesting nonetheless! We find at our ISP that if we break something with IPv4 we experience a very different type of support issue to if we break IPv6. Breaking v4 results in, broadly, a pretty hard “down” state. While folks are unhappy, it is at least simple. Breaking v6 results in weird…

Most ISPs still just block IPv6 altogether because most small businesses seem to try IPv6 once and then forget to eg update their AAAA records so to the user it looks like their favorite niche thing works when they're on but not on the one they're paying for which creates problems. It's kind of a weird issue, I don't know if there are nice solutions other than hoping IPv4 just goes away eventually. Happy eyeballs was…

> Most ISPs still just block IPv6 altogether

That’s increasingly not true, at least in developed countries. Traffic to Google in the US has been majority IPv6 since a few months ago.

Re: Using the Internet without IPv4 connectivity

#87

Post by riobard »

tsimionescu wrote: post #81
kstrauser wrote: post #69

Earlier quoted context omitted.

We have no such saying in the industry. IPv6 is generally easier to implement and maintain. If IPv6 were the incumbent and someone came in proposing IPv4, they’d be laughed out of the room for its ridiculousness. “You have to run a stateful server just to assign addresses? Dynamic header length? A tiny address space? And tell me again about this NAT thing, LOL.” V6 was designed by the engineers who realized what they…

You still need a stateful server to assign IPv6 addresses for most use cases, through DHCPv6. SLAAC doesn't even give you a DNS server yet. And even if it did, many ISPs assign too small address spaces for SLAAC, or your networks isn't so simple that you can just auto-negotiate some address.

Stop spreading misinformation.

> You still need a stateful server to assign IPv6 addresses for most use cases, through DHCPv6. SLAAC doesn't even give you a DNS server yet.

DNS now comes in Router Advertisement per RFC 8106. No need for DHCPv6 anymore.

> And even if it did, many ISPs assign too small address spaces for SLAAC, or your networks isn't so simple that you can just auto-negotiate some address.

Most residential ISPs allocate in /48, /52, /56, or /60. Even if they allocate in the smallest /64, it's still perfectly fine for SLAAC for most home users utilizing a single subnet.

Re: Using the Internet without IPv4 connectivity

#89

Post by elcritch »

jeroenhd wrote: post #31
simonjgreen wrote: post #3

Slightly misleading title, this is more “getting to the IPv4 internet via an IPv6 tunnel through a VPS”. Also just called 4in6. Interesting nonetheless! We find at our ISP that if we break something with IPv4 we experience a very different type of support issue to if we break IPv6. Breaking v4 results in, broadly, a pretty hard “down” state. While folks are unhappy, it is at least simple. Breaking v6 results in weird…

When my IPv4 died last time, I noticed it mostly because Github didn't work anymore. These days, most consumer websites just work on IPv6. That said, people whose routers were only provisioned IPv4 DNS servers did have a full outage. If Microsoft would get off their incompetent assets already, my biggest concern would've been remembering the mDNS hostname I've assigned to my router so I could log in and see if IPv4 i…

The Meta Quest software also sucks at this. You’d expect an essentially new platform struggling with this. Valve is basically all IPv4 afaict too.

Pretty annoying and lazy if you ask me.

Re: Using the Internet without IPv4 connectivity

#90

Post by nyrikki »

bigstrat2003 wrote: post #78
ornornor wrote: post #77

After all these years I still don’t see a compelling reason to spend days pulling my hair out switching all my machines and home lab to ipv6. I just find port forwarding and firewall rules more intuitive vs the prospect of spending weeks troubleshooting everything, reconfiguring firewalls, renunbering my network. What am I missing?

You're missing that it's not that difficult. Unless you have a very complicated home network, setting up IPv6 should take you an evening tops. For my network and my ISP (Comcast), it's literally just turn on IPv6 in the router, it will pick up a prefix from the ISP and advertise it locally, then I add a firewall rule for whatever I want to be accessible from outside (which isn't much).

The problem goes way back to the IPng days. With the project being designed by a committee who would choose random hills to die on, wasn't adaptive to new needs and used sticks more than carrots.

Some of those have been changed but typically after trying to implement it broke so many things that people just quit trying.

Some were well intended like the 'no NAT' in the days of FTP and before reverse proxies etc.

Others were intentional pain points to for adoption like when resolvers were not permitted to return A and forced to return AAAA records even when you ISP didn't support IPv6 etc...

Mix in problems like the max prefix size being too large for scanning a local network space to be practical etc... and people have been trained for decades that the pain is worse than any benefits.

Yes, today it isn't hard on small home networks where IPX will an IP gateway would also work fine, but things break as things get more complex.

Somewhere someone probably has a copy of the mail lists where I pointed out in around ~1996 that forcing globally unique IPs was a leaky abstraction and that there was more nuances and tradeoffs that needed to be considered.

It was obvious to me because I was stuck on a Altavisa firewall, but I was roasted.

On the IPv4 side, user needs were addressed through CIDR, carrier grade NAT, FTP passive connections etc...

I still tried to move companies to IPv6 a few more times and was bitten ever time.

Almost every time it was due to arbitrary global decisions, when they should have been focused on maintaining good will and making adoption as easy as possible.

The effort depended on a collition of the willing, and just changing 'must' to 'should' in key RFCs would have dramatically improved the chances of adoption.

I am actually glad you have an ISP that allows you to even do this, mine still does not.

Post reply on HN