IPv6's design was a mistake. There, I said it.

Reactions to this post on Twitter are amazing 😅

- "Shepard might be an adequate alternate profession for you"
- "Your calling is cabling. Go back to it"
- "Are you mentally challenged or something?"

They're really doing Twitter proud there 👏👏👏

Show thread

@fribbledom bigger numbers more better.

the logic is infallible

@fribbledom I'm over 30 now. I don't think I will ever see a L3 protocol newer than IPv6 in production.


Possibly not, yeah. I bet we're still dealing with IPv4 at age 60, though 😄


In my personal experience pretty much *everything* except the address space expansion is _still_ hindering adoption.

If something is still struggling with adoption 25 years after its invention, it's pretty safe to assume mistakes have been made.

@fribbledom as TCP & UDP didn't changed, apart from some API endpoints like getaddrinfo(3) and friends, I don't see why.

The fact that most of the software is US-based and that the USA have plenty of IPv4 may have something to do with it, US ISP are also notoriously pretty bad at changing (except for charging too much).

Windows, macOS, Linux had it by default for the last 15 years.

In many cases, IPv6 makes your network faster...

@bortzmeyer @Keltounet @fribbledom It's an upgrade that people don't upgrade to, so I think it's fair to call it a mistake.

Python 3 is great, but designing the upgrade path so that it took 10 years for the center of gravity to move over is generally agreed to have been a huge mistake.

IPv6 has gone over twice that time and still counting. Actually Python 3 should probably have learned from IPv6's mistakes.

@fribbledom Out of curious, where do you see the hindering at?

I'm on IPv6 as my primary stack, native, and (sit down for it...) I have a US ISP.
My home traffic went over 50% IPv6 a couple years ago.

@fribbledom @Keltounet The reason it hasn't been adopted more widely yet is that until recently IPv4 was enough for everybody, especially with NAT. Yes, NAT can be a nuisance, but most people would never know the difference.

@mansr @fribbledom and this point has nothing to do with the design of IPv6 :)

This is more resistance to change, esp. amongst ISP and the manufacturers of core products (looking at you Cisco) to adopt, helped by the hoarding of IPv4 blocs by the US.

The rest of the world, esp. Asian countries have adopted IPv6 way earlier because they had no choice.

@fribbledom Some aspects, definitely, but it does address some other areas. I actually had a really good idea how they could have solved MTU issues in a way clearer way 🙂

@farhan @fribbledom maybe two decades of RFCs trying to fix it making it a moving target that no vendor could ever hit? It wasn’t until what, 2017 that we finally got an RFC that combined everything (RFC 8200) and now that’s already outdated with ERRATA.

Take both IPv6 and DNSSEC and place them in the burn barrel please

@fribbledom Nowhere near as bad as IPv4's. Choosing 32-bits originally was probably just a failure of nerve but sticking with it so long is borderline racist (an implicit assumption there are only a few billion people, at most, that matter).

A huge number of problems arise from working around IPv4's lack of address space directly (e.g., the need for SNI) and indirectly via the need for NAT.

AFAICS, IPv6's problems come from half-hearted implementation, not the basic protocols.


Of course I'm not advocating dropping IPv6 and sticking with v4. Both were products of their time, one a bit too naive, the other probably a bit too complex.

A middle-ground would certainly have helped adoption and getting rid of v4.

@fribbledom I'm not convinced v6 is complex as such but would grant that there are maybe more differences from v4 than were strictly necessary and skipping them would have possibly helped adoption but I think their contribution to holding things back is minor compared with simple conservatism.

@edavies @fribbledom
> Choosing 32-bits originally was probably just a failure of nerve

32-bit was TOTALLY REASONABLE back then. Who could possibly know that ARPAnet and TCP will still exist after 50 years, instead of being another dead-end project? Computer systems usually become obsolete and forgotten quickly, but there are always outliers like C, Unix, or TCP. The only fault is not realizing they're gonna be the lucky one. 🤣


let's just cut to the chase - we all know where this is headed.


sent from a3a0952b8243abc2ede779ba9140190d4feea32f71bf502d03049f4e8ac1ac787715dae309bf679d257383b266e8027ccd47a62abaefb9d7ca3c4c3a16177e18e99dd0f461e9e8d08af25f8ec89962ff46c22c1f855e904e69c5a569a66b09249c16bd4ff55930c

One knows he's getting old when he says that IPv6 is badly designed. There, I said it. 🤭

@fribbledom is anyone even using it? I don’t think the last few ISPs I used even supported it

@thomasfuchs I asked mine, and they said I could have *either* v6 or v4.

So I stuck with v4 :/

@thomasfuchs @fribbledom It's more common in some countries; some UK ISPs are using it, but not others.

@penguin42 @thomasfuchs @fribbledom My ISP gives me a full /48 network. I changes ISP's because the IPv6 support was so bad with my old provider (they gave me a /64 with 6rd).

@thomasfuchs @fribbledom It would seem as though it's around 35%. It's less than I thought. Some countries have come much further than others though.

@thomasfuchs @fribbledom Everyone is using it but Big Enterprise today. Every OS in current use prefers IPv6 if it can use it. Every mobile carrier has nearly required it for years (to the point iOS APIs went “IPv6 only”) as addressing mobile cell traffic is much more possible in IPv6. At least a third of all home traffic today, especially P2P, is IPv6. It’s businesses and old data centers with huge LANs and/or huge static IPv4 leases that are slow to adopt. Home users and ISPs had to adopt.

@thomasfuchs @fribbledom This has been obvious for a while in IPv6 usage charts because usage was lower during the work week than during evenings and weekends.

When everyone started in healthy at home initiatives in March, IPv6 shot way up versus IPv4.

@thomasfuchs Cox won't even give you an IP4 address at all if you're a new customer.

it could have been better.
Maybe something like

@elera_mortis Agreed, I think people are able to wrap their head around a decimal representation of a binary space more than a hexadecimal representation.

And the idea of replacing
:0000:0000: with ::
adds up to the confusion.

@elera_mortis @BalooUriza :: means “As many zeroes as we can fit.” So ::1 -> ::0001 -> 0000::0001 -> 0000:0000::0001 all the way to 0000:0000:0000:0000:0000:0000:0000:0001. If you had 1234::1 that’d be 1234:0000:0000:0000:0000:0000:0000:0001. You can only do it once of course, so 1234:0000:0000:0000:5678:0000:0000:0001 becomes 1234:0:0:0:5678::1 (1234::5678:0:0:1 might also work idk)



"(1234::5678:0:0:1 might also work idk)"
Yes. AdGuard's dns-server ip ends with :ad1:0ff.
This is imho a funny feature, writing funny stuff with ip addresses.

@BalooUriza @elera_mortis I don’t think people wrap their heads around IP addresses at all, whether it’s IPv4 or v6. Their eyes just glaze over and they think “Oh look, numbers. I better go find someone who is qualified to think about that to solve my problems for me.”

@cy @elera_mortis Yeah, I'm talking about the folks who actually have to support this stuff.

@elera_mortis @fribbledom Yeah, that’d last us another whole 3 years. Very future minded. And not word aligned, to boot!

Real talk though, maybe they should have used an 8 byte address space, instead of a 16 byte one. I dunno. 2^2^2 vs 2^3


I'm not really doing anything which requires one or the other, or much knowledge, but I "got" plain IPv4 addresses and subnet masks after a short explanation by someone who knew their stuff, and I still have to understand how IPv6 addresses work, and why fixed addresses and doing away with NAT wouldn't make every single device I have much easier to track on the internet, 24 years (and a few hour-long websearches) after I read this wasn't an issue with IPv6.

@fribbledom I work with a guy who was on that committee at Sun. I'll pass along your feedback. 😆

@fribbledom (I'm afraid you aren't wrong. How can it be so danged hard to do IP addresses that have more digits?)

Sign in to participate in the conversation

Server run by the main developers of the project 🐘 It is not focused on any particular niche interest - everyone is welcome as long as you follow our code of conduct!