[rcxdude]

because it's unnecessary to get the same benefit. Being behind a firewall would have the same effect (and any ipv6 deployment will have this), it's just that NAT requires this. It's like saying eating a spoonfull of cinnamon has health benefits because it hydrates you when you have to drink a glass of water afterwards: you could just drink the water.

[throwaway1105q]

I don't quite understand what you mean by "any ipv6 deployment will have this". When my ISP switched to IPv6, my internal devices were exposed to the internet and the only thing that stopped the incredible amount of bot traffic was my own on-device firewall that I explicitly turned on and configured. Luckily I don't have any smarthome stuff, not sure how I'd configure a firewall on a lightbulb. These devices didn't have a public IPv4 before that. And a bonus - the ISP didn't say anything about this possible consequence, just "we're making some changes".

NAT has more benefits - I don't want anyone to know how many devices I have at home, I don't want anyone to know which one I'm using to access their website, I don't want anyone to try guess the OS and version of my devices, etc. And now I'm scared to have a simple DLNA media server because I can't just install WireGuard on the TV. I'm probably going to buy a router and make my own NAT soon (don't have access into the ISP modem).

I felt better when the whole municipality had a single IP address. A lot of bullshit ads - means the targeting wasn't working. Now they're way too good.

[RulerOf]

> I don't want anyone to know how many devices I have at home

Even if your ipv6 host or border firewall allows pings through, it's not practical to scan an entire /64. There's just too many addresses in it, and your devices will frequently change them.

> I don't want anyone to know which one I'm using to access their website, I don't want anyone to try guess the OS and version of my devices, etc.

They already do this through fingerprinting that operates with higher-layer protocols.

> And now I'm scared to have a simple DLNA media server because I can't just install WireGuard on the TV.

This is very simple to implement. Ensure it's listening on the link-local address. That's the IP that starts with fe80. These are unrouteable by spec.

[throwaway1105q]

> They already do this through fingerprinting that operates with higher-layer protocols.

It's very hard to distinguish my iPhone and Mac from the other dozens/hundreds people have in my building just through fingerprinting. Very easy if they have separate IP addresses.

Ad link local - cool, I'll look into that, thanks.

[crazygringo]

It's actually very easy just through fingerprinting. You might be surprised.

It doesn't matter if everyone in your building has an iPhone and a Mac as well -- there are things about virtually every single one of them that make them unique.

[oarsinsync]

https://www.amiunique.org/ is scary and eye opening

[throw0101c]

> When my ISP switched to IPv6, my internal devices were exposed to the internet and the only thing that stopped the incredible amount of bot traffic was my own on-device firewall that I explicitly turned on and configured.

When my (previous) ISP switched on IPv6 none of my internal devices could be connected to because my Asus did stateful packet inspection and only allowed in replies to connections that were previously initiated.

> NAT has more benefits - I don't want anyone to know how many devices I have at home, I don't want anyone to know which one I'm using to access their website

Given that temporary IPv6 addresses tend to rotate every 24 hours it will kind of hard to track individual devices by IP in a 2^64 address space.

You could rotate addresses 10 million times per second, using each only once, and it would take over 5000 years to exhaust a single /64.

> I felt better when the whole municipality had a single IP address. A lot of bullshit ads - means the targeting wasn't working. Now they're way too good.

I now have to use a ISP-supplied router (for GPON), but when I still had my Asus on the DSL/IPv6 ISP I could tell it to reboot every night and I would get a new IPv4 address and a new IPv6 prefix every day.

[PaulHoule]

My ADSL connection rides on some non-IP network before it hooks up to a concentrator about an hour away. Most location based services, other than Apple, seem to assume I am in Norwich, NY. So I get these ads that say “They don’t like it when seniors use this one weird trick to save money on car insurance in Norwich but they can’t stop it” and “Horny grandmas want to jump your bone right now in Norwich” and such.

Contrast that to using public WiFi in NYC where everybody knows exactly where you are.

[throwaway1105q]

I'm on DOCSIS to the Home / Fiber to the Building, but there seems to be some kind of overlay network and as a result, my PC that's hooked into the modem is on the public internet.

Before IPv6 it was a classic internal LAN with IPs like 192.168.0.1.

[hdjdjdj]

Pro tip: buy a computer for and make it into a router .. There are some great cheap fanless machines out there (servethehome has reviews)...

You could also just use an old pc...

For software opnsense, pfsense, openbsd, freebsd, Linux (openwrt could be used too if you want embedded)

It is a pain to start ... But satisfying when it works :)

[rcxdude]

Well, that sounds like a colossal misconfiguration on the ISP's part. A firewall blocking incoming requests has been standard part of ISP routers for a long time.

[throwaway1105q]

On the other hand... What exactly is the benefit of IPv6 then? I thought the point was to make all my devices addressable on the public internet. How is it useful if the ISP firewall blocks my servers?

And yes, incompetent ISPs are the norm.

[ndriscoll]

Without bidirectional NAT, hole-punching works. Two sides of a p2p connection can coordinate with an intermediary to learn each other's addresses. They send each other a packet, which gets dropped by the other side. Their firewall sees the outgoing packet though, and opens the port. The next time they send each other packets, they will be allowed through. The intermediary is only needed to do the initial handshake instead of for all packets.

With NAT, it doesn't work because the ports get remapped, and the intermediary doesn't know how they will get remapped on the p2p connection, so they can't coordinate to send on the correct ports to open the firewall.

Or UPnP can work. By default, your router drops incoming packets on all ports. If you want to e.g. run a game server, then on startup, it hits a standard API to tell the router to forward that one port. On shutdown, it can tell it to close the port (you could potentially also have the router require keepalives to keep the forward alive. I'm not familiar with the details of UPnP and related protocols).

Without a public IP, you need intermediate servers to relay all traffic to you, which centralizes the web. With p2p working, you can e.g. have high quality video calls with friends/family instead of dealing with the garbage quality tech companies allow. Or I can share with my mom photos of her grandkids with effectively unlimited storage; for 2 years of 2 TB Google storage, I can buy 20 TB of disks.

[Spooky23]

The biggest benefit is exponentially higher complexity, assuring continued job growth for network engineers.

[bzzzt]

The point is not needing a NAT translation table and running out of ports on your router. My provider also delivers an IPv6 configuration with all ports closed. I can enable incoming traffic for the devices that need it.

[throwaway1105q]

Running out of ports how? Someone is hosting 65536 public services in their home network? Why not just pay for an additional public ipv4 then?

I can't configure anything technical about my internet. Any change is paid, and often simply not possible.

[toast0]

Running out of ports is usually a misunderstanding, but a device doing stateful NAT will have a limit on how many states it can manage, and it's usually not fun when it goes over the limit.

[tflol]

This looks like the usual ipv6 kool aid batshit. I don't want a bunch of kids and enemy states poking at and port scanning my laptop directly, regardless of whether or not I have a firewall enabled.

And, no, I don't think it's practical for everyone and their grandma to "just set up a bastion"

[scrps]

This is also spec for IPv4, it was intended to be as publically routable as IPv6 is. NAT is just a consequence of everyone realizing circa early 90s (iirc) IPv4 addresses would run out at the rate the network was growing. Yes NAT acts as an inbound default-deny firewall but that isn't it's purpose.

You have a router, it has a firewall, that is meant to be used to control access to the network, you don't have to assign rules to every device you can assign default interface rules that apply to any connection.

Just because you get a publically routable address doesn't mean the internet defines physics and hops over your router and firewall.

Also as an aside - perimeter security is a very outdated way of looking at security, yes the perimeter is still important but if it is your first and only line of defense you are gonna be in for a bad time, defense in depth as it is called where you look at your systems and networks as layers to an onion is the more modern standard and NAT as a security mechanism has never been standard in either because it isn't.

[rcxdude]

I mean, they'd need to figure out your IP address beforehand, something that's a lot harder with ipv6. You've also got a much better chance of punching a packet through a NAT than an ipv6 firewall (and it's now expected behaviour for a lot of applications, as NAT makes it too difficult to just make connections directly).

[rainonmoon]

They wouldn't need to figure out anything. The "kids and enemy states" are just hosing address ranges. I don't agree with the above commenter that NAT offers any meaningful security in this regard (now they're just hosing your consumer router instead which is probably less secure than the average updates-installed Defender-enabled Windows box). But you're both making points about security through obscurity in different ways.

[throw0101c]

> The "kids and enemy states" are just hosing address ranges.

If you could scan one million addresses every second it would take about 500,000 years to scan just one /64. Not sure how practical that would be.

When I was still with an ISP that did IPv6 my Asus would block any incoming connection attempt unless it was a reply (SPI firewall), though it may have (IIRC) allowed pings in by default.

[tflol]

SPI firewall looks interesting, appreciate the education.

[rainonmoon]

Yeah that is an absolutely bonkers amount of time so you're probably right in that the approach of low-effort wide net-casting attackers would have to change. I'm curious to know how Shodan etc. deal with this.

[kalleboo]

Shodan ran an NTP pool time server on IPv6 and harvested the addresses of machines that checked in to get the time. Pretty clever.

[tflol]

> now they're just hosing your consumer router

There is a dramatic difference in effort between ( owning a device ) and ( owning a router, configuring network access to the device, then owning the device ).

Also psychologically: If I was a rock hard piece of shit and I knew I was at the doorstep of a personal device, I would treat it much more aggressively than a router. I suppose maybe that's just me and not the kids and enemy states.

[rainonmoon]

I mean, I don't know why you would when the router potentially gives you a foothold across many devices instead of one and the router is likely running multiple services. Yes, that is just you; the threat model I'm describing is widespread automated attacks, not individual or particularly motivated.

[tflol]

You're saying there's less incentive for widespread automated attacks on personal devices?

edit: Changing the subject to insulting me is a bad way to conclude. You're creating an illusion the debate is concluded in your favor instead of responding to points. I don't think any of my points had a sound argument against them.

[LegionMammal978]

Wouldn't IPv6 firewalls configured for typical users (i.e., denying unrecognized incoming connections) pose a similar barrier to making direct connections reliably on the application level? Not every user will be willing or able to open a hole in their firewall for every shiny new application that wants one.

[utensil4778]

Yeah, I think it is very explicitly a bad thing for all devices to be directly exposed to the entire internet- firewall or no. NAT is a pain, sure, but it does have the benefit of forcing you to have a network isolated from the internet, and only allow external access when explicitly configured to do so.

I have exactly one machine which needs to be accessible from outside the local network. The rest of them should never be. Do I want to spend extra time ensuring that each and every single device on my network is secure, or do I want to do the inverse and assume all devices are secure and only spend effort to make the one machine exposed?

I can't imagine anyone who would actually want or need their WiFi toaster to be publicly routable, WiFi cameras, every computer. There's absolutely no reason for it. Instead of relying on network isolation, we expect users to just implicitly rely on who knows how many different firewall implementations. Hopefully your router configures it by default.

[semi]

Are you sure about that 'never'? that no device will ever try to use p2p fonnections?

Even then id still rather ensure every device is appropriately firewalled. 'not worrying about it's sounds like a hardened shell with a juicy center. What happens when a device does get compromised and tries to spread to your local network?