[kyledrake]

I don't like to admit this, but at this point honestly I think ipv6 is largely a failure, and I say this as someone that wrote a blog post for APNIC on how to turn on ipv6.

I'll get endless pushback for this, but the reality is that adoption isn't at 100%, it very closely needs to be, and there are still entire ISPs that only assign ipv4, to say nothing of routers people are buying and installing that don't have ipv6 enabled out of the box.

A much better solution here would have been an incredibly conservative "written on a napkin" change to ipv4 to expand the number of available address space. It still would have been difficult to adopt, but it would have the benefit of being a simple change to a system everyone already understands and on top of a stack that largely already exists.

I'm not proposing to abandon ipv6, but at this point I'm really not sure how we proceed here. The status quo is maintaining two separate competing protocols forever, which was not the ultimate intention.

[hypeatei]

> A much better solution here would have been an incredibly conservative change to ipv4 to expand the number of available address space

"And what do you base this belief on?

Fact is you'd run into exactly the same problems as with IPv6. Sure, network-enabled software might be easier to rewrite to support 40-bit IPv4+, but any hardware-accelerated products (routers, switches, network cards, etc.) would still need replacement (just as with IPv6), and you'd still need everyone to be assigned unique IPv4+ addresses in order to communicate with each other (just as with IPv6)."[0]

0: https://news.ycombinator.com/item?id=37120422

[avidiax]

> Fact is you'd run into exactly the same problems as with IPv6.

If you treat IPv4 addresses as a routable prefix (same as today), then the internet core routers don't change at all.

Only the edge equipment would need to be IPv4+ aware. And even that awareness could be quite gradual, since you would have NAT to fall back on when receiving an IPv4 classic packet at the network. It can even be customer deployed. Add an IPv4+ box on the network, assign it the DMZ address, and have it hand out public IPV4+ addresses and NAT them to the local IPv4 private subnet.

IPv6 seems to be a standard that suffered from re-design by committee. Lots of good ideas were incorporated, but it resulted in a stack that had only complicated backwards compatibility. It has taken the scale of mobile carriers to finally make IPv6 more appealing in some cases than IPv4+NAT, but I think we are still a long way from any ISP being able to disable IPv4 support.

[btilly]

I agree with that belief, and I've been saying it for over 20 years.

I base it on comparing how the IPv2 to IPv4 rollout went, versus the IPv4 to IPv6 rollout. The fact that it was incredibly obvious how to route IPv2 over IPv4 made it a no-brainer for the core Internet to be upgraded to IPv4.

By contrast it took over a decade for IPv6 folks to accept that IPv6 was never going to rule the world unless you can route IPv4 over it. Then we got DS-Lite. Which, because IPv6 wasn't designed to do that, adds a tremendous amount of complexity.

Will we eventually get to an IPv6 only future? We have to. There is no alternative. But the route is going to be far more painful than it would have been if backwards compatibility was part of the original design.

Of course the flip side is that some day we don't need IPv4 backwards compatibility. But that's still decades from now. How many on the original IPv6 will even still be alive to see it?

[redox99]

Hardware would catch up. And IPv4 would never go away. If you connect to 1.1.1.1 it would still be good ole IPv4. You would only have in addition the option to connect to 1.1.1.1.1.1.1.2 if the entire chain supports it. And if not, it could still be worked around through software with proxies and NAT.

[morshu9001]

Hardware support for ipv6 hasn't been the limiting factor in a long time. Users higher on the stack don't want to adopt something that makes so many unnecessary changes.

[solarkraft]

You’re focusing on the technical difficulty of implementing it in software. This is not the problem. IPv6 support is now present in almost every product, but people still refuse to set it up because it’s so different to what they’re used to (I’m not arguing whether the changes are good - they’re just changes). IPv4+ would’ve solved this social problem.

[almosthere]

The IPv4+ could pass through a router that doesn't know about it - the cloud host that receives that packet could interpret it in a special way, in fact you could stuff additional data into the next layer of the stack for routing - it's not like many services beyond TCP would need to support the scheme.

[bit_logic]

This whole discussion reminds me of the beautiful design of UTF-8. They used the lower bits to be ASCII which made backwards compatibility so much easier. It also reminds me of the failure of Intels Itanium and the success of AMD x64. Engineers often want to abandon backwards compatibility to make a new "beautiful" design, but it's the design that has full backwards compatibility that's actual impressive.

[x0x0]

It reminds me of python 3. Basically, a huge chunk of people (in my case, scientific programming) get an enormous mess and nothing at all of value until... 3.6 maybe (the infix matrix mult operator). Stunningly, people weren't enthused about this deal.

[krupan]

So well said! Those are great comparisons.

[liquidpele]

It would maybe be okay at the router to break some things, but ffs even in software I have to choose? Why do I need both ping and ping6 this is stupid!! They really screwed up by making it a breaking change to the OS and not just internet routing.

[culi]

The only solution is a gov't mandate. China went from almost no adoption to leading the world in adoption (77% of all Chinese internet users) in a few years because they explicitly prioritized it in their last 5-year-plan.

The ISPs aren't gonna do it on their own.

[p_l]

US government has finally learnt from how vendors break the mandates and there's now IPv6 mandate if you want to sell to federal government, and waivers are only available for buyers not vendors, and individually every time.

[ajross]

I wouldn't say "failure". There are many, many IPv6 client devices out there, mostly on mobile networks. And it works great and they do well and the tools all support it very well.

But IPv4 will never, ever die. The rise of NAT as a pervasive security paradigm[1] basically neuters the one true advantage IPv6 brought to the table by hiding every client environment behind a single address, and the rise of "cloud everything" means that no one cares enough about reaching peer devices anyway. Just this morning my son asked me to share a playlist, so of course I just send him a link to a YouTube Music URL. Want to work on a spreadsheet for family finances with your spouse in the next room? It lives in a datacenter in The Dalles.

[1] And yes, we absolutely rely as a collective society on all our local devices being hidden. Yes, I understand how it works, and how firewalls could do this with globally writable addresses too, yada yada. But in practice NAT is best. It just is.

[JeremyNT]

> I wouldn't say "failure". There are many, many IPv6 client devices out there, mostly on mobile networks.

Honestly it's a huge success due to this fact alone.

IPv6 is failure only if you measure success by replacing IPv4 or if you called "time" on it before the big mobile providers rolled it out. The fact that all mobile phones support it and many mobile networks exclusively deploy it tells you what you really need to know.

IPv6 is a backbone of the modern Internet for clients, even if your servers don't have to care about it due to nat64.

[MagicMoonlight]

Yep, just call it IPv8 and make it double the length of IPv4.

Ultimately, an address system that replaces “1.1.1.1” with “JEDBSO:7372B6D6A:727:8:72829:762927” or whatever just isn’t viable.

Even AWS doesn’t let you use IPv6 with anything… and they charge you for using IPv4 now.

[dawnerd]

I toyed with using ipv6 in my local network just to learn it and what a headache that was. Ultimately not worth the hassle. I can remember most of the important device ipv4 on my network, I can't say the same for v6.

[olowe]

AWS supports IPv6 on a number their services now: https://aws.amazon.com/vpc/ipv6/ For example there are options to use their hosted memcache/redis service IPv6-only: https://docs.aws.amazon.com/AmazonElastiCache/latest/dg/netw...

Shocking it took them so long but, hey, it's there now.

[free_bip]

This is the first time I've heard this critique. I think most people don't care if their IP address is easily human readable/memorizable. In my experience when people do deal with ipv4/v6 addresses directly, they just copy-paste.

[onionisafruit]

Circa 1999 I was working for Cisco as a sysadmin. I got my CCNP through internal training and considered making a career of network administration, but ipv6 changed my mind. It seemed so much more difficult and unpleasant to deal with. I didn't want that to be my day to day work.

I think the same thing happens on a different scale with ISPs. They don't want to deal with it until they have to for largely the same reason.

[sgjohnson]

> It seemed so much more difficult and unpleasant to deal with.

In my experience it’s much easier and much more pleasant do deal with. Every VLAN is a /64 exactly. Subnetting? Just increment on a nibble boundary. Every character can be split 16 ways. It’s trivial.

You don’t even need to use a subnet calculator for v6, because you can literally do that in your head.

Network of 2a06:a003:1234:5678::555a:bcd7/64? Easy - the first 4 octets.

Network of 10.254.158.58/27? Your cheapest shotgun and one shell please.

[UltraSane]

At first I though so too but IPv6 is actually easier. instead of CIDR you always have 64 bits for network and 64 for host. You get a public /48 IPv6 prefix that allows for 16 bits of subnets and then the host addresses can just start at 1 if you really want. So addresses can be prefix_1_1 if you want. And the prefix is easy to memorize since it never changes.

I DO think using 64 bits for hosts was stupid but oh well.

[thayne]

I actually think it would have had a better chance of success if ipv6 had embraced the breaking changes to add some killer feature that would have made it worthwhile to upgrade even for entities who didn't need to worry about running out of ipv4 addresses.

I'm not sure what that feature would be though.

[imoverclocked]

This is actually unwanted in a dual-stack world. Once you have divergent behaviors on your networks, you have a complex/weakened security model.

Networking should be boring.

[bigfatkitten]

IPv6's failure was mostly caused by the IETF's ivory tower dwellers, who seem to generally have no practical experience or understanding whatsoever of how networks are actually built and run today, especially at the small to mid scale.

Small site multihoming, for example, is an absolute disaster. Good luck if you're trying to add a cellular backup to your residential DSL connection.

IETF says you should either have multiple routers advertising multiple provider-assigned prefixes (a manageability nightmare), or that you should run BGP with provider independent address space; have fun getting your residential ISP or cellular carrier onboard with this idea.

[pigggg]

IETF has a history of being hostile to network operators. I mean actual network operators - not the people who show up at conferences or work the mailing list who just happen to get a paycheck from a company that runs a network (and have zero production access / not on call / not directly involved in running shit). It's gotten better in the last few years in certain areas (and credit to the people who have been willing to fight the good fight). But it's very much a painful experience where you see good ideas shot down and tons of people who want to put their fingerprint on drafts/proposals - it's still a very vendor heavy environment.

[cryptonector]

IPv6 was a total failure of imagination.

The fact is that already in 1993 routing tables were just too big, and the fact is that having a "flat" address space was always going to mean huge routing tables, and the fact is that because IPv6 is still "flat" routing tables only got larger.

The fix would have been to have a subset of the address space that is routed as usual for bootstrapping ex-router address->AS number mapping, and then do all other routing on the basis of AS numbers _only_. This would have allowed us to move prefix->AS number mappings into.. well, DNS or something like it (DNS sucks for prefix mapping, but it could have been extended to not suck for prefix mapping), and all routing would be done based on AS numbers, making routing tables in routers _very small_ by comparison to now. Border routers could then have had tiny amounts of RAM and worked just fine. The IP packets could have borne AS numbers in addition to IP addresses, and all the routers in the middle would use only the AS numbers, and all the routers at the destination AS would know the routes to the destination IPs.

But, no. Great missed chance.

Well, we still could do this with IPv6, but it would be a lot of heavy lifting now.

EDIT: Ah, I see draft-savola-multi6-asn-pi existed.

EDIT: Ah, see also LISP [https://www.rfc-editor.org/rfc/rfc6830]. But LISP is essentially dead.

[nine_k]

> a cellular backup to your residential DSL connection

Hmm, what's the problem? I suppose your home devices should never be exposed to the public internet, and should only be accessible via a VPN like Wireguard. NAT64 is a thing if your home network is IPv4.

BTW what's the trouble with multi-homing? Can't an interface have two separate IPv6 addresses configured on it, the same way as IPv4 addresses?

[Asooka]

I've been thinking we could simply extend the ipv4 address to be 11 bytes by (ab)using the options field. That is, add an option that holds more bytes for the source and destination address, which are to be appended to the address already present in the header.

I am thinking that since an option starts with 2 bytes and everything must be padded to a multiple of 4 bytes, we can add 16 bytes to the packet, which would hold 7 extra address bytes per source and destination, giving us 11 byte addresses. ISPs would be given a bunch of 4-byte toplevel addresses and can generate 7-byte suffixes dynamically for their subscribers, in a way that is almost the same as CGNAT used today but without all the problems that has.

Most routers will only need to be updated to pass along the option and otherwise route as normal, because the top level address is already enough to route the packet to the ISP's routers. Then only at the edge will you need to do extra work to route the packet to the host. Not setting the option would be equivalent to setting it to all 0s, so all existing public hosts will be automatically addressable with the new scheme.

There will of course need to be a lot more work done for DNS, DHCP, syntax in programs, etc, but it would be a much easier and more gradual transition than IPv6 is demanding.

[fruitworks]

I don't think so. It would be more confusion because no one will know if a network is ipv4 or ipv4+, leading to edge case bugs and confusion and people will similarly be lazy and choose to only implement ipv4 knowing it will always be reverse compatible and the cost is transferred to the consumer.

Plus, it's only 2048x the address space. It's within the realm of possibility that we will need to upgrade again once this place is swarming with robots.

[umanwizard]

ipv6 adoption is still steadily rising. Not as fast as anyone hoped, but at least steadily. There is no way it can be abandoned at this point even if we wanted to.

[aurumque]

I wonder if it could still be usurped by another standard that is somehow more popular. If adoption of that leapfrogs over IPV6 then maybe it will have just been a waypoint along the way.

[kmeisthax]

Stripped of all the other baggage that came with it (e.g. SLAAC, IPsec, etc) IPv6 is an incredibly conservative addressing extension. The only thing even more conservative than v6 would have been to drop the lower 64 bits of the address and the associated EUI-64 local addressing scheme. Which... to be fair, that turned out to be a very bad idea, but the length of the field isn't what was holding up v6 adoption.

I suspect by "incredibly conservative" you mean "backwards compatible", which... no. You can't make an addressing extension backwards compatible with hardware that doesn't read all of the address. Of course, we did that anyway with CGNAT, and predictably it causes huge problems with end-to-end connectivity, which is the whole point of IPv6. You're probably thinking more along the lines of an explicit "extension addressing header" for v4. Problem is, that'd mean a more awkward version of IPv6's /64 address split[0], combined with all sorts of annoying connectivity problems. The same corporate middleboxes that refuse to upgrade to IPv6 also choke on anything that isn't TCP traffic to ports 80 and 443. So you'd need Happy Eyeballs style racing between CGNAT IPv4 and "extended IPv4".

Also, that would just be a worse version of 6in4. Because they also thought of just tunneling IPv6 traffic in IPv4 links. I don't think you understand how incredibly conservative IPv6 actually is.

The problem with "incredibly conservative" IP extensions is that nothing beats the conservatism of doing literally nothing. IT infrastructure is never ripped out and replaced unless there is a business case for doing so. The current problem with IPv6 adoption is that nobody has yet said "let's stop processing IPv4 traffic", they've only said "let's get more dual-stack hosts online", which is a process that only asymptotes to 100% IPv6, and never reaches it.

IPv4 was not the first version of the Internet protocol. That honor goes to Network Control Protocol (NCP). The reason why we don't have an asymptotic long tail of Internet hosts still demanding NCP connectivity is because this was back when "having a connection to the Internet" meant "having a connection to ARPANET". The US military could just refuse to process NCP packets and actively did this to force people onto IPv4. Now imagine if someone big like Google said "we're going to stop accepting IPv4 connections" - people would jump onto v6 immediately.

[0] Let's say we add a 32-bit extension header onto IPv4

[krupan]

"Stripped of all the other baggage that came with it..."

But that baggage is a huge part of the problem. Almost nothing you know about IPv4 applies when you switch to IPv6, and most of us found that out the hard way when we tried to make the switch. Leaves a pretty bad taste in your mouth.

[WorldMaker]

> The current problem with IPv6 adoption is that nobody has yet said "let's stop processing IPv4 traffic"

Mobile carriers have done that between consumer devices and network towers. That forced a lot of innovation (including tools like better DNS64 and "happy eyeballs" protocols) and network stack hardening.

The roll out of out CGNAT in some cases is "let's drop IPv4 traffic randomly" and "happy eyeballs" in consumer devices is transparently driving a lot of consumer traffic to IPv6.

This is why mobile and consumer devices are leading the pack on IPv6 adoption.

It's maybe not all of Google that next needs to say "we're going to stop accepting IPv4 traffic", it's maybe more specifically GCP (and AWS and Azure) that need to do that to drive the non-consumer IPv6 push we need. The next best thing would be for all the cloud providers to at least start raising IPv4 address prices until their clients start to feel them.

[eqvinox]

> The current problem with IPv6 adoption is that nobody has yet said "let's stop processing IPv4 traffic"…

One of the giant CDNs translates all IPv4 traffic to IPv6 at the edge (stateless NAT46) and is IPv6-only in its core network (for one of its primary product networks; like everybody they have multiple networks.)

[vachina]

Truth is there are too many devices that only speak IPv4 or have untested IPv6 stack. People still can’t even agree on how ipv6 address is represented.

[iov6throwaway]

People have totally agreed on how IPv6 addresses are represented.

[imoverclocked]

I think this is defeatist talk where it’s not warranted. I remember IPX networks in the 90s were still a thing because people believed they could eke out a little more performance for their games. It’s taking a long time to move to IPv6 in some parts of the world. eg: anyone who doesn’t feel the pain of the IPv4 address crunch likely due to having a large chunk to begin with. Many influential organizations in North America definitely fall in that category.

IPv6 is a success IMHO because it is used in so many places. Google’s IPv6 traffic graph shows close to 50% adoption and still trending up. We can’t possibly expect the world to be near 100% overnight… the internet is a big place with the whole spectrum of humans influencing IT; There will always be someone who will cling to IPv4 for dear life.

[alphazard]

> I'm not proposing to abandon ipv6, but at this point I'm really not sure how we proceed here. The status quo is maintaining two separate competing protocols forever, which was not the ultimate intention.

The end game will be a cryptographically large address space allocated based on some cryptographic operation, rather than a committee carving up the space arbitrarily.

Tor already does this, addresses allocation is not a problem. I think they used to use hashes, but now use Ed25519 public keys. Obviously, Tor is not suitable for most tasks. No one should have to pay for the extra latency if they don't need the anonymity.

The real problem is routing in these address spaces, and there have been a few projects like CJDNS which try to solve it.

[morshu9001]

Imagine every address along a major road is 3 digits, and some shortsighted post office code assumes 3. Your business is 845 Oak St. One day they say hey, this road is getting too long, let's update that code to support 10 digits and we never worry about this again.

Oh and btw, your address is now 9245593924 Oak St.