A dispute over IPv4 addresses blew the lid off an effort reshape allocation
theregister.com
theregister.com
You need to go back before they where introduced and see why almost every major city added them.
I'm saying that no, it doesn't do that. Taxi drivers are staying afloat despite medallions, not because of them.
Overtime the system was intended to do various things like hold drivers accountable which it very much did, as well as various things it failed at.
but how? the market would decide how much a taxi driver earns based on the simple law of supply and demand. Unlivable wage for one might be a very livable wage for another one.
The consumers would massively benefit from increased competition.
Maybe you want locals with families to have the opportunity to drive. So you control the supply until they can afford local rents and still have time to spend with their kids.
Of course, cities often turned it into extortion, but they don't have to do that.
The internet might be entirely decentralized, but IP address allocations are very centralized.
Which one is the central authority?
And even if there was no IANA, the 5 regional authorities would still very much be centralized.
I wish it was that easy.
OSs are fine, but the software and systems that we build on top of them are often v4-only.
Having bugs in untested functionality is understandable, but using long-deprecated known-broken APIs, or using the current ones in ways that are clearly and obviously wrong, isn't.
(Also you should be using, and therefore testing, v6 at this point. But that's a separate discussion.)
In fact v4 addresses are not even just an ordinary resource, they're very like Real Estate aka Real Property, or "land". There's a finite amount of it, we can't do anything about that - to some extent it's effectively "occupied" already in many cases, so an imagined "property right" doesn't trump that in practice, and it's very much not fungible.
It's not v6's fault that people refuse to stop using the old ones.
IPsec was bigger, but immediately backported to v4 and nobody uses it anyway, so I'm not sure it counts.
v6 is already designed to be slowly phased in, by allowing deployment of updated software/OSs/devices that support the new addresses in parallel with the old and gradually moving things over. If anything, the reason people can still use the old APIs is because of the approach of slowly phasing it in.
That's what you wanted, and it's still not good enough for you?
What do you mean by "gradually expand the existing header"? I assume you don't mean "keep making new header versions that increase the address length by one byte until we reach 16 bytes", but I can't think of anything that could be described like that which would have made adoption any quicker or easier.
> I still have to maintain the ipv4 stuff functionally forever
Consider why this is. What's stopping you from just not doing that -- and would a different design of v6 help at all with this?
Maybe it wouldn't have made it any better to expand the header and change the version but then again maybe it would have. Why even have the version value in the header to start with if it is always going to be set to 4? What is the point?
I have to keep dealing with ipv4 until companies like amazon and github and every little smb that has a vpn stop. It would have been amazing yes had they designed ipv6 differently and just increased the header and address sizes and added a padding value for the high bits. We could have front padded all of the ipv4 32 bits at the router level and made it almost seamless to the end users.
> We could have front padded all of the ipv4 32 bits at the router level and made it almost seamless to the end users.
This is a bit simplistic. You either haven't thought about it in nearly enough detail to figure out how it would work, or at least you haven't explained the detail to us.
It's easy to take "192.168.0.1" and write "0.0.0.0" in front of it, but then what? v4 devices don't support any way to do that on the network and v4 software can't handle the extra address length, so... what are you actually suggesting to do here? How can whatever it is be seamless if the devices that end users are using don't support it?
I can describe a working scheme that might be what you're thinking of. You could make something that's basically a version of NAT but where the inside addresses are v6. That would allow a v6 client to reach a v4 server by sending packets to a special v6 prefix that represents the v4 space; it would work essentially the same as NAT in v4 but with an extra header conversion step.
But... v6 already has this. I'm using this functionality on my desktop, which has no v4 address, and I can get to any website just fine. That includes Amazon and Github, just to be clear.
Why do you need to keep dealing with v4 when you can just do what I'm doing but at the ISP level? v6 doesn't need a different design to make this work, because it already works.
That said, western governments can, should, and (somewhat) do mandate IPv6 adoption. It's slow, though. IIRC it's something like 2-3% a year and we're at 20-30%, so in a decade we should have a majority supporting IPv6 and hopefully that will cause some sort of feedback loop forcing universal IPv6 support for practical reasons.
Your numbers are off by a factor of 2. We are around 40% now and the growth is more like 4-5% per year. With that we could be at 90% adoption in a decade, and getting to >50% should be just couple of years; many major countries are already past that mark.
IPv6 has one-off engineering costs to roll it out, but decreases ongoing costs for NAT.
PS: The site can get laggy on low end devices, because of the number of records (sorry :(, but I didn't want to spend too much time on this project)
IP address is just a number! There's no reason to charge money for a number. If you are out of numbers, make more numbers. There's an infinite amount of numbers. What's next? 1000$ per social security number, because we are running out of 9 digit numbers?
> IP address is just a number
No it's not. It's exactly 4 octets. No more, no less. Which takes me to the next point.
> If you are out of numbers, make more numbers
You can't make more numbers and expect all systems that assume 4 octets to work. That's how we ended with IPv6, which actually provides us with a near-infinite amount of "numbers" as you call them. Mind you, they're still quite finite though - their numbers are just really really big.
> There's an infinite amount of numbers
No, and that's why they cost - scarcity. You could try to argue that it's an artificial scarcity, but the most likely answer is just that they never thought they'd need as many addresses as we've realised since.
https://www.hacksparrow.com/networking/many-faces-of-ip-addr...
> No it's not. It's some more specific definition of a number
So... it's not just a number, it's a number?
(Four octets put together form a single 32-bit number)
Access to use an IP address is a ticket to go online, not just the ticket's number.
The fact that an IPv4 address can be decomposed into four quads has nothing to do with that point.
My comment, said differently, is that the syntax of an IP address is the wrong focus to make a "more than just a number" point.
Sure, it's not the raw fact that it's 4 octets. It's the fact that it's not any other number, especially the fact that it's a finite resource.
A domain name is just a number, still some cost millions.
With the added benefit that the revenue from that could go to something fair, like a public dividend, rather than to rent seekers
The problem is that companies who have an adequate chunk of IPv4 addresses have no reason or incentive to bother with IPv6. So unless you're going to yank existing alloctaions - which comes with massive problems - I think letting them trade on the open market is the least bad option.
OTOH, there are many class C IPv4 zombie ranges that aren't actively used for routable purposes. Those could be reallocated readily with an transparent, generous, fair process.
With just a handful of countries over 50% I doubt that IPv6 will be adopted by a meaningful enough majority to have IPv6-only external services, unless there is a stronger push for to have at least transition mechanisms deployed everywhere. Something that players with commercial interests in ipv4 trading like the one of the article will be against.
Imagine if IPv6 had simply said:
- An IPv6 address is simply an IPv4 address plus 2 or 4 octets
- An IPv4 device would simply get routed an IPv6 address based on the IPv4 "prefix"
We'd have a clear path forward with some level of backwards compatibility and this would've been over decades ago.
IPv4 header contains checksum field which must be recalculated on each network hop, as the addresses change with each one.
Other than that? i agree. IPv6 is an disaster, and it wouldn't be so if we moved early - but there were no incentives at all to move early.
What about prefixes. The vast majority of ipv4 has been allocated. It we just have more octets then we haven't actually made more available allocations because they will be under already assigned prefixes.
The smallest globally routable allocations in ipv4 is a /24. Routing hardware has a limit on the number of prefixes it can route. Just changing that to a smaller size would ease pressure, but even that is impossible without upgrading all switching infrastructure to handle the space.
The next version of IP was always going to be a brand new protocol. There wasn't even a realistic option besides that.
Think of IP4v2 as essentially embedding a NAT field into the IP header.
This might work or it might not in any given circumstance but the path to make it work is a lot more straightforward.
As for the routing table size issue, it's not really an issue at all. Assuming extra octets, a /24 block holds 16M addresses instead of 256 (or 254 realistically). You don't need a larger routing table.
This would also allow an organization that might've had a /16 or /8 block or multiple blocks to get by with 1 (or a few) /24 blocks, allowing you to reclaim blocks without making external routing tables any larger.
> As for the routing table size issue, it's not really an issue at all. Assuming extra octets, a /24 block holds 16M addresses instead of 256 (or 254 realistically). You don't need a larger routing table.
If you want to subdivide those 16M addresses in any useful way, you're going to need a larger routing table.
I would think an overflow would be a more realistic scenario. Imagine all the fun that would cause.
Not just technical stuff like this, even broader nonsense like "Before vaccines diseases didn't exist" and "The world is actually flat" and so many of these things.
By getting the remedial class in first, from an impassionate machine, there's a chance we rescue them from some awful intellectual mistake they might otherwise feel emotionally committed to, and in some cases from a spiral into Conspiracy Theorist gloom in which the gap between nonsense they believe and observable reality is all explained by a vast all-encompassing Grand Conspiracy.
Anyway, no, you can't "just" do this. What's happened here is that you didn't really understand the problem, and then you've "solved" the problem you'd invented, not the actual problem, and surprise, the invented problem had an easy solution.
On the upside, you're a step ahead of the people who figure hey, just allow up to 999.999.999.999 for an IP address, that would fix it...
Would you mandate that everyone run this LLM on their devices? That an ISP or OS should force it on people?
I think many people wouldn't mind running an AI locally that did the job of "Catch common errors and misconceptions and alert me to them", but you'll run into the usual arguments of "who decides what reality is" we get now with fact checkers and I could see various websites refusing to let people post anything that didn't conform to the AI's approved worldview.
Anyone who thinks the world is flat is going use an AI that supports their own views and they'll just say any AI that disagrees must have been ordered to lie by whoever is behind the round earth conspiracy. Flat earthers are a perfect example of how it doesn't matter how much proof you have that something is factual, or how much you educate people, there will still always be someone who insists on believing a clear lie
Today an LLM isn't suited for this because it hallucinates, so your hypothetical Flat Earther can probably quickly get an LLVM to invent a groundbreaking 1996 Canadian study which proved the entire "planet" is actually within 87.35km of Ottawa, creating plausible sounding details as long as you remain interested.
Ideally it should be easier for such a system to be self-correcting because the conspiracies always need layer upon layer of conspiracy until basically everyone is "in on it" so as to keep up the deceit with a Grand Conspiracy. The largest successful secrets were probably WWII projects like Ultra and Manhattan, maybe Operation Neptune. None very large, most exposed within a few years (obviously Neptune much faster, you can't keep an invasion secret). So if a belief requires these vast conspiracies that's just wrong, try again.
Tuskegee was kept quiet from 1932 to 1972. The guy who eventually leaked it was hired in 1965, learned about it from co-workers, reported it one year later, then again two years after that, and then after four more years finally leaked information to the press. Tons of people knew, others had previously protested it as well. Operation Sea-Spray was hidden for 27 years. It was ~25 before the Pentagon Papers leaked.
Even large numbers of people can keep a secret long enough to do a lot of damage. What bothers me more, is that there's not a whole lot of accountability when the truth comes out, and those who were hurt by those secrets don't seem to ever get anything close to justice.
I think that as long there isn't enough morality and backbone in my country to hold people meaningfully accountable and help victims when atrocities are exposed, asking for there to always be a whistleblower ready to sacrifice everything might be asking too much. Especially since we seem to be cracking down harder and harder on whistleblowers.
Yes, more whistleblower protection is needed, in my country (the UK) too. But I don't buy the theory that because the protection isn't good enough people just don't come forward. That's too transactional. I think they do come forward anyway and we owe them much more.
MAP-T translates IPv4 to IPv6.
You mean like 0/96 prefix or 64:ff9b::/96 or ::FFFF:0:0/9 prefix does?
I'm not sure how you'd solve that in practice. Not only did we run out of meaningful bits, even ipv7 / addext didn't simply extend the address (https://datatracker.ietf.org/doc/html/rfc1475)
The only way this could have been implemented is with a brand new protocol. There isn’t even a theoretical way to have a backwards compatible IP.
> The top net-buyers organisations of IPv4s in March were Chinese and the top net-sellers were American.
> The above produced 1,644 country pairs totalling 10,941,802 IPv4s. Below is the first record.
https://tech.marksblogg.com/ipinfo-free-ip-address-location-...
Full disclosure: I do have skin in this game. I have a /32 and several /40s of IPv6 and no IPv4 at all. For a hobbyist IPv4 is already prohibitively expensive. $14k for a /24? Miss me with that bullcrap.
You're not even supposed to buy an IPv4 address, but a market for IPv4 addresses sprung up due to their scarcity.
Now, I'm not trying to be pedantic here. To obtain an IPv6 address, you need to be allocated one. And that's the job of your ISP, apparently. Which kind of makes sense since 1) that's how your home computer gets an IPv4 address (a dynamic one, these days), and 2) unless your ISP supports IPv6 you'll need to set up some sort of IPv6-over-IPv4 tunnel anyway.
Now, you may say "oh dear god are there any options that don't involve my ISP", and, well, good luck: https://www.apnic.net/get-ip/get-ip-addresses-asn/
By default when you're allocated an IPv6 address you get at least a /96 block and thus you'll never run out of sub-addresses, so you really only need one.
I just didn't have the willpower (or wallet) to see it through. I was merely doing it for fun and burning a bunch of money for addresses that I would barely use didn't sound fun after thinking about it. Still, it was fun chatting with my friends on IRC about it and it is a very interesting icebreaker!
Alternatively you can become a RIPE member and request all of the resources yourself, but that’ll cost you €2550 on the 1st year and then €1550 yearly.
You don’t need to be physically located in RIPE region to go the RIPE path. All you need is _connectivity_ in the region. A $5 BGP VPS counts as _connectivity_.
> or be a multi-homed site (would have required getting a second ISP to my residence, I believe?)
No. You needed to get 2 BGP servers. Another ISP to your home would be useless unless they both offered BGP sessions.
Of course the $35 doesn't include the VPS you'd need for actually announcing, but this is a _very_ cheap way to get into BGP.
Where did you get the /96 from? A /96 is basically useless.