Every HackerNews post about IPv6 has some of the worst, most privileged, idiotic, vibe-coded, proprietary, ignorant, 16bit, capital-guzzling, unicorn takes I've ever known on the subject:

- IPv6 addresses are
too hard to remember.
So? You're not meant to remember addresses, that's why we have DNS, write it down, literally a non-issue.

- IPv6 is confusing and I don't want to learn something new.
That's a personal issue buddy, either start reading or get left behind, that's what you said about AI right? More things than you depend on this transition.

- NATing has solved the IP limit problem so there's no point.
NATing is a plaster slapped onto brain bleed, easy and cheap, but ineffective, it causes a wide range of usability problems, such as blanket IP bans, restrictions on self-hosting, connectivity issue for VPNs both private and corporate.
To make matters worse, the effects are significantly worse in poorer countries, while Europe, China and the USA have a bounty of IPv4s to use (though China's still aren't enough), India has been on critically short supply for a while now with reports of
multiple NATed network layers being issued. Imagine if you got banned from Valo because your neighbour 4 districts away got caught cheating.

- We've been trying for 40 years and it hasn't worked so let's give up.
OK, we're going to give up on solving world hunger too then because that's clearly not getting anywhere, and the energy crisis too while we're at it, just shut it all down.
Just because you personally haven't seen the progress or felt its effects doesn't mean its not happening, people smarter than you have been working on this before you were born, and at this rate might continue to work on it after you switch careers to Goose Farming.

- IPv6 hasn't worked so let's just make IPv7.
Insane take, despite how it looks, IPv6 support is extremely widespread and ready to go, the reluctance of big tech and ISPs is purely due to the cost implication and lack of enforcement, creating a brand new spec now would enforce another 40 year delay just to assuage your own personal opinion.

- IPv6 is a security risk because the router isn't NATing.
Misunderstanding of what NATing does. Even with a public-facing IP on every device, ports are still protected by the router's firewall.

- IPv6 is a privacy issue because now you can easily identify every device in a home by its public IP.
A valid concern,
if it hadn't been identified and resolved with the Privacy Extensions to SLAAC that randomises your IP address after a set time period, mitigating the problem to that of your NATed IPv4 Public IP, if not making it more private by muddying the telemetry waters.

#ipv6 #networking
Privacy Extensions for IPv6 SLAAC - Internet Society

Whereas IPv4 had two basic methods for obtaining an IP address, IPv6 has three. Static configuration is basically the same in both protocols, although less relevant for IPv6 given the length of the address. DHCP is also there for both protocols, and IPv6 DHCPv6 is described in RFC 3315. Introducing SLAAC The new method that […]

Internet Society

@Baa
I agree in general and I cheer for IPv6 becoming more widespread

however

> you're not meant to remember addresses

The only cases when I type IPv4 addresses from memory is when I don't have working DNS - either when debugging DNS, or when manually configuring the first interface of a host, on a network with no DHCP.

Unfortunately, these are the cases where I can't rely on DNS to remember IPv6 for me.

It's not a deal-breaker but it does add some friction.

@wolf480pl @Baa
Why?
That's likely a private network and in that case one is free to decide on which address and -type to use.

What's so hard to remember 2001:db8::1 as a default local DNS? 2001:db8::/32 is the "documentation" private IPv6 prefix - much like example.com is for DNS.

I remember a time when I mixed up 192.168 to 198.162 for our home network. Yay! Public IPv4 /16 mapped at home!

Pro-tip: Have a printed cheat sheet.

Reminds me I need to step on my ISP's feet.

@wolf480pl @Baa
Coming to think of it:
Remembering 2001:db8 is 32 bit for IPv6.
Now there's
10.0/8
172.16/12
192.168/16
for IPv4 private networks. Makes 9+13+17=39 bits for IPv4. Whoops?
IPv6 in fact requires _less_ human memory for "common" addresses. Even more so when starting to count characters, not just bits.

"I can't remember IPv6 addresses" is just another word for "I don't want to learn new stuff."
Here's your two-weeks notice, thanks for your services.

@syn_rst
1. since IPv6 has no NAT, you use public addresses in LAN, and you can't control the first 64 bits of it

2. using the "documentation" prefix as a private prefix is wrong

3. actual private IPv6 addresses come from fd00::/8 and you're supposed to pick the next 40 bits randomly

4. I also need to remember a public DNS resolver, ideally one that responds to pings - like 1.1.1.1 for IPv4

5. I'm sorry to hear your country's law allows such a short notice period

@Baa

@wolf480pl @Baa
1. Since I may choose from fd00::/8 for private purposes (see 3), I have at least some control over the first 64 bits. Also, if my ISP handed me a prefix which is shorter than 64 bits - what is what an ISP should do.

2. sure is, so is using "example.com" locally.

3. so, my random choice just happens to be my postal-code.

4. Why is there no DNS resolver on the LAN already?

5. Mine actually doesn't, however, refusing to learn on the job may be a reason for termination.

@syn_rst
4. Because I put it there, but somehow it doesn't work, so I need to figure out if it's broken, or if my internet is down.

Or maybe there isn't one yet and I need working DNS to apt install unbound.

3. https://datatracker.ietf.org/doc/html/rfc4193#section-3.2.1

2. which is why nobody does that

1. sure but the part you can't control is more than 32 bits

I'm sure with enough use I'd memorize the few IPv6 prefixes I use at home. But it will take some time.

RFC 4193: Unique Local IPv6 Unicast Addresses

This document defines an IPv6 unicast address format that is globally unique and is intended for local communications, usually inside of a site. These addresses are not expected to be routable on the global Internet. [STANDARDS-TRACK]

IETF Datatracker

@syn_rst
Also, re your math - that's not how entropy works.

I have 192.168/16, 172.16/12 and 10/8 in my long term memory, so it's only 2 bits to remember which one is used on a particular network.

Then there are the remaining 16, 20 or 24 bits. Even in the worst case of 10/8, it's 2+24=26 bits. But if we assume router gets .1 on the last octet, it'd be 2+16=18 bits.

@syn_rst
Meanwhile with IPv6, even if I get a whole /48 from my ISP (unlikely), the part fixed in tbe standard is 2000/3, so that leaves 45 bits I can't control that I have to remember.

@wolf480pl @syn_rst literally just do NAT and bind your outer address to something short. my local net is fdaa::/64, default gw is fdaa::1, and (for now) i’m statically assigning addresses

it’s necessary on my dual-ISP setup (so I have my devices available under two prefixes at once from the outside), but nobody is stopping you from doing it with just one ISP if you want to use short IPs.

like 75% of your complaints are only valid because you’ve been used to IPv4 for all your life :|

@domi @syn_rst
I didn't mean these to be complains.

> fdaa::/64

so you're saying fuck the RFC?

@wolf480pl @domi
RFCs exist to ensure interoperability, which they're fairly good at.
However, one paragraph below that famous "MUST" for randomized 56 bits in the private local prefix the reasoning says it is to be so to ensure network merges don't generate the issues they do while everyone freely distributes 10/8 addresses all over the place.

No, don't fuck the RFC. But nobody stops you from ignoring the part which creates more hassle than it actually solves.

Just be aware of consequences.