Brace Yourself, IPv6 is Coming (supabase.com) 320
Paul Copplestone, co-founder of Supabase, writing in a blog post: On February 1st 2024, AWS will start charging for IPv4 addresses. This will cost $0.005 per hour -- around $4 month. A more accurate title for this post would be "Brace yourself, IPv4 is leaving," because I can't imagine many companies will pay to keep using the IPv4 address. While $4 is relatively small for an individual, my hypothesis is that AWS is a foundational layer to many infrastructure companies, like Supabase -- we offer a full EC2 instance for every Postgres database, so this would add millions to our AWS bill.
Infrastructure companies on AWS have a few choices:
1. Pass on the cost to the customer.
2. Provide a workaround (for example, a proxy).
3. Only offer IPv6 and hope the world will catch up.
Infrastructure companies on AWS have a few choices:
1. Pass on the cost to the customer.
2. Provide a workaround (for example, a proxy).
3. Only offer IPv6 and hope the world will catch up.
Leave me behind (Score:3, Insightful)
If all of the advertising folks and other BS moves on to IPV6 with the rest, I'll be happy to stay on the OldNet and forgotten.
Re:Leave me behind (Score:5, Funny)
Re: (Score:3)
Hey, I can still list off my FidoNet address
Re:Leave me behind (Score:5, Insightful)
I can subnet an IPv4 address in my head, can't do that with IPv6. IPv6 is functional but it's a beast to implement. Remember the good old days of setting up IPX/SPX to network DOOM and Papyrus Nascar Racing on your 486 computers? Ah, good times.
Re:Leave me behind (Score:5, Insightful)
Re:Leave me behind (Score:5, Insightful)
Another reasonable option would be to use a prefix (perhaps "x") and then write the numbers in hexadecimal notation. So then the range would be x0.0.0.0 to xFFFF.FFFF.FFFF.FFFF. This has the side benefit of making some of the more common subnet masks more intuitive, and makes it easier to count bits when applying
But what they _really_ should have done, is just increase the size of the address space and change as few other things as possible to make that work. Then maybe it would've eventually been adopted into widespread use. Making IPv6 so different and so much more complicated that you need an entire separate college degree program to teach anyone how to work with it, was the big mistake.
Re: (Score:3)
The purpose for the address structure is that the top half of the address is always the path through the network, the lower half always the ID of the device.
This means routers don't need to look at whole addresses, only one byte either side of the current point, although it's rarely implemented that way.
It also means devices are mobile. The machine's unique ID can be used to track the machine as it moves between ISPs.
Re:Leave me behind (Score:5, Insightful)
People say this, but is this honestly the REAL reason people don't like IPv6? The problem with IPv4 is that it has too few addresses, and the reason you claim to like it is precisely because it has too few addresses... This is like saying you don't like 64-bit processors because you can no longer recognize if an address is in the heap or not.
We've got IPv6 at work, because we make devices that use it. Generally you can recognize from the first 4 hex digits if it's the link-local to the lab netwokr versus the corporate site lan, etc. Since I've been using it almost exclusively for over 15 years, I can handle IPv6 addresses pretty easily but I'm getting too rusty on recognizing IPv4 addresses...
Implementing IPv6 is easy. It's simple and straight forward. IPv4 adds a lot of complexity if you need a stack that does both, mostly because there are a lot of oddities added to IP (NAT for example).
Re:Leave me behind (Score:5, Insightful)
I can subnet an IPv4 address in my head, can't do that with IPv6.
Baloney. Figuring the subnet boundary in IPv6 is trivial. For one thing, it's almost always /64, falling exactly in the middle of the 128-bit address.
For another it's hexadecimal, base 16. That means you only have to remember 4 masks, because everything beyond those 4 moves which digit the mask applies to. Moreover, savvy IPv6 users pick subnet masks that always fall between two digits so that they never have to -change- one of the numbers when applying the netmask like they do in IPv4.
What you meant to say is that you can't memorize the IPv6 address because it's too long. That's not a bug, it's a feature. Half the problems with the TCP/IP protocols stem from developers and operators working with bare addresses when they should be working with names and letting the name service find the address.
Re: Leave me behind (Score:3)
So, the protocol that costs more to maintain, making it less accessible to creators and average citizens? That half the world doesn't even have access to? That citizens don't have reachable at their homes due to CGNAT or dynamic addresses?
Enterprise will hoard up ipv4 for the foreseeable future as well as presenting ipv6 options. Supporting universal Internet access means supporting ipv6.
Re: (Score:3)
They're already dual-stack. They're not going to abandon any eyeballs they don't have to.
Any day now! (Score:5, Funny)
How long has it been, 20 years?
Re: (Score:3)
Re: (Score:2)
Yes, but IIRC we first got ipv6 in the beginning of the 2000s. In 93 or thereabouts I was telneting to www.cern.ch and trying out HTTP/HTML. I may be wrong - there Shirley would have been earlier deployments.
Re: (Score:2)
Yes, but IIRC we first got ipv6 in the beginning of the 2000s. In 93 or thereabouts I was telneting to www.cern.ch and trying out HTTP/HTML. I may be wrong - there Shirley would have been earlier deployments.
Whether or not you're right DON'T CALL ME SHIRLEY!
Re: (Score:2)
There were, but stop calling me surely.
Re: (Score:2)
Well, IPv6 was a more academic thing in the 90s. It was originally released in 95 (so missed 'early'), but then significantly redefined in 1998. They tried to ignore the need for centrally managed parameters so DHCPv6 didn't make it in until 2003 or so, with some ISP requirements trickling in over the next decade or so to make it reasonably deployable in a standards based ecosystem. So 20 years is about right for it to be *vaguely* workable, but I'd give it only about a decade of realistically having eno
Re: (Score:3)
NAT bugs the crap out of me because it started as a "well crap, we have address exhaustion, here's a neat trick to share a bunch of client ports on a single address out to a bunch of private addresses" and then became a "security" thing, with a fair argument to be made that a NAT firewall is a "failsafe" configuration of a failing firewall more likely to dump it on the ground than to forward it, though that's a sketchy one. To the utterly absurd 1:1 NAT where you *have* a public IP address and all traffic
Re: (Score:3)
IPv6 was originally done to solve a problem that was forthcoming - that we'd run out of addresses to give every endpoint a unique address. And without every endpoint having a unique address, you couldn't communicate with that endpoint.
That happened, and we now have NAT. Which went on to prove that not every endpoint needs to have a unique address. It also proved that you don't need total connectivity between endpoints.
That makes a lot of IPv6 irrelevant - because you will never have true end to end connecti
Re: (Score:2)
I remember when I was making Ethernet switches back in the early 00's that each and every RFP I ever saw made IPv6 support a hard requirement.
Which then they ignored completely after implementation. I don't recall one single site using it. I visited the NASA facility back then near Moffet Field and they were 100% IPv6 way back then. (We didn't get that deal). Nobody else until Telco back-haul networks developed and they used it a lot.
NAT happened, and that extended the life of IPv4 by at least 4
Re:Any day now! (Score:4, Interesting)
It's because IPv6 is a poorly engineered piece of crap farted out by a circle jerk of self absorbed asses in an ivory tower. It doesn't succeed because it's insufficiently transparent, with pathetically little effort made towards backward compatibility and a completely fucked transition plan. As a high school project would earn a D- or worse.
Re: (Score:2)
IPv6 was definitely discussed in the early 90s, support was added to the Linux kernel in 1996.
Re: (Score:2)
A very early draft of IPv6 was added to Linux in 1996, it was not compatible with the standardized IPv6 from 1998 and it wasn't until the release of 2.6 in 2003 that the default stack was compatible with the standardized IPv6. Microsoft added it to Windows NT in 1998 after it was standardized (and in 2002 with SP1 they added production worthy support in Windows XP).
And the first draft of what would become IPv6 was in 1994, so people claiming that they had courses on IPv6 in the early 90:ies are simply misre
Re:Any day now! (Score:5, Interesting)
Slightly longer than that, as IPv6 RFCs came out in 1995-1998. But the address space advantage of IPv6 only kicked in in the early 2010s [wikipedia.org] when it became much more expensive to obtain IPv4 address space. In 2011 Microsoft piad Nortel $7.5m [networkworld.com] for IPv4 address space ($11.25 per address). It was only a matter of time before that sort of cost got passed on to customers by companies like AWS.
Re: (Score:2)
I know, I know, but I still can't help it ;)
Re: (Score:3)
Many core routers still do not have full feature parity between IPv4 and IPv6, on the latest software release. Even new models.
There are a bunch of ways to get IPv6 transported in an IPv4-only network, but getting IPv4 transported in an IPv6-only network is not nearly as well supported. IPv6 will always be an afterthought if you have to implement a full IPv4 backbone anyway, before you can start doing IPv6.
Re: (Score:2)
I hear most of the cloud infrastructure of the big players is now ipv6 based (I'm not in that business anymore, so no first knowledge except the occasional paper that passes by me), and has been for a while, so I was curious why there aren't more spillovers. Maybe this is the first trickle.
Re: (Score:3)
In AWS it's basically just a checkbox to assign a public IPv6 address to an EC2 load balancer. For individual instances inside your VPC it doesn't really matter since it's all private address spaces anyway. I've never bothered with v6 addresses internally as the 10/8 v4 CIDR has always been big and flexible enough.
But the reality is that most businesses only need a handful of public-facing addresses anyway, and the $4/address/mo has to be weighed against possible compatibility issues and lost customers.
Re: (Score:2)
Depends on the requirements and the budget, I guess. I've seen deployments of banks that were entirely from ipv4 routable blocks and all communication between offices was done over point-to-point leased lines between routers at their offices, and with phone company VPN connections to a private gateway for mobile access. The result was a network completely separable from the internet, yet completely addressable and it was a really large block. But it was a rather long time ago and getting a bit fuzzy why. I
Re: (Score:2)
There's absolutely no reason why you couldn't implement something like that with IPv6 natively. It would actually be easier to do in some ways.
Re: (Score:3)
This is so not true. Lots of ISPs run IPv6 on top of IPv4+MPLS. You hardly have to touch your backbone at all to get that working; it has been supported for a decade on almost all relevant equipment.
It literally is true. IPv4 and IPv6 are separate networks with completely separate routing tables. That people are able to run both protocols over the same wires and tunnel them over other protocols does not change the fact they are separate networks.
Re: (Score:2)
Yes, I had my first course in IPv6 in 2003, still got the certificate!
Re: (Score:2)
Hehe, good times.
The longer you leave it, the worse it will be (Score:5, Interesting)
If you deployed IPv6 years ago like you were supposed to then this would be a non event. The ISP i use here has provided native IPv6 connectivity since 2007.
Keeping legacy IP running is expensive, you have horrible kludges like NAT and the proxy these supabase guys are talking about, and costs for legacy IP are only going to keep increasing. The sooner it dies out the better because even for those of us who deployed IPv6 years ago, we are stuck with the costs, hassle and risks of legacy IP to access antiquated sites that haven't.
Slashdot is a prime example of this, using cloudflare which is fully ipv6-capable but for some reason didn't deploy AAAA DNS records. Thankfully you can manually override the DNS and force it to use IPv6 over cloudflare.
Re: (Score:2)
I completely agree with you.
Re:The longer you leave it, the worse it will be (Score:5, Interesting)
Ask all the ISPs around the world why they haven't provided IPv6 to internet users. According to https://stats.labs.apnic.net/i... [apnic.net], so far India get >80% IPv6 coverage, while all other countries range from a half to nearly none. No serious websites can risk losing the majority of potential visitors just for saving a few bucks.
To the ISPs, they win by stuck behind together. No need for infrastructure upgrade as long as all their buddies don't upgrade either.
Re: (Score:2)
As a result of having NAT, you instantly loose the principle benef
Re: (Score:3)
As a result of having NAT, you instantly loose the principle benefit of IPv6: Every node having a publicly accessible address.
Have you not noticed how insecure all the IoT devices are? Removing NAT and making them directly Internet-accessible will be a disaster.
Yes, we can have statefull firewalls, but these also eliminate most of the supposed advantages of directly-accessible devices.
What's going on is that the gods of the Internet have had an irrational hatred of NAT and the result is the mess that is the transition to IPv6.
Re:The longer you leave it, the worse it will be (Score:4, Insightful)
No one said that youhave to give up your firewall. Firewalling and NAT are two separate things.
Re: (Score:2)
YOU THOUGHT THERE WAS NO NAT FOR IPV6, BUT NAT STILL EXISTS
https://blogs.infoblox.com/ipv... [infoblox.com]
Re:The longer you leave it, the worse it will be (Score:5, Informative)
Have you not noticed how insecure all the IoT devices are? Removing NAT and making them directly Internet-accessible will be a disaster.
I reversed my mod points to reply to this. NAT is not, never has been, and should not be considered a security feature! NAT is a translation layer and nothing more. Security is obtained by a firewall with the correct firewall rules configured on it. Generally that firewall rule set should be "BLOCK ALL" and then individual rules to allow what needs to pass. If you're relying on NAT for security, that is the real disaster.
Re: (Score:3)
With IPv6 you don't need NAT. You don't need to get assigned a unique 64-bit suffix, you take the suffix that comes with your device already, all you get assigned is a 64-bit prefix assigned to your home, that multiplies the number of addresses by billions even before adding the remaining 64 bits. But this disrupts how things are done in IPv4, so there's resistance.
NAT isn't really security here, it's a hack for getting more addresses but so many think it's a security feature. A simplistic firewall does t
Re: (Score:3)
Increasingly, if the home user doesn't have IPv6, there will be places they simply can't reach at all. automatic address configuration is nice as well
In my case, I do enjoy the v6 addressing. It allows me remote access to a few things on my LAN from my phone.
For most people, it's all good as long as their router comes with a simple default ruleset that provides the same level of protection as NAT but with less hassle.
Increasingly, staying with v4 means carrier grade NAT (AKA double NAT) which plays havoc wi
Re: (Score:2)
Ask all the ISPs around the world why they haven't provided IPv6 to internet users.
Want to know something absurd? My ISP CG-NATed me when I first joined up, I had a non-public IPv4 address but they did offer a public IPv6 address. This was nearly 10 years ago. So I did what any sane nerd would do and switched to a business account.
They would *NOT* offer me an IPv6 address. So here I am 10 years later with a perfectly routable IPv4 address that I pay extra for but without the option of having IPv6.
IPv6 rollout is clearly a case of doing the minimum.
Re:The longer you leave it, the worse it will be (Score:5, Insightful)
horrible kludges like NAT
IPv6 needs to embrace NAT, and protocol designers need to stop thinking that they don't have to worry about keep-alives to punch through middleboxes with IPv6. Transparent end-to-end connectivity doesn't exist on the real internet. If there isn't NAT, there are stateful firewalls. And IPv6 doesn't have a way of dealing with multi-homing that won't explode routing tables and won't burden the servers and clients with complex routing decisions. These problems exist because the IPv6 designers thought NAT was only a kludge to work around address scarcity, but it is much more than that, and IPv6 needs NAT. You can do away with port translations, because there really are enough addresses now, but 1:1 NAT is necessary. Forget your end-to-end addressability: inband-signaling of addresses is the actual horrible design that needs to die in a fire.
Re: (Score:3)
I guess I don't see the issue with maintaining IPv4 on the internal network for ease of management and only implementing IPv6 on the edge devices that live with one interface on the web.
Why do I need to go through the pain of implementing IPv6 for internal devices?
Seems like a "best of both worlds" situation
Re: (Score:2)
The key point of IPv6 was to give flat addressing to everyone. NAT implementation defeats that purpose.
The problem with IPv6 remains the rest of the stuff that came along with the flat addressing. It was a wish list of crap that lost relevance over the years anyway.
AWS trying to charge for IPv4 addresses probably results in them losing business long before it drives anything else.
Re: (Score:3)
I agree in general, but if I were running something that was making money, i'd want a pure v4 implementation alongside a v6 implementation. They will charge me for that; that factors in a decision to host elsewhere. It's not a huge amount but it's something.
Re: (Score:2)
I've got to agree. The Modern World doesn't need to know about the Beige G3 Mac running OS 9. IPv6 to the router and let that do the routing to the internal network.
Re:The longer you leave it, the worse it will be (Score:5, Interesting)
IPv6 does support NAT, but NAT is a horrible kludge and is best avoided so it won't get used unless you have no other option.
There are multiple ways to deal with multi homing - ideally you get your own PI space and announce it via BGP, and the only reason that's rarely done with legacy IP is because it's too expensive - IPv6 address space is easily affordable so this becomes a practical option again. You can bring your own address space to AWS for free too.
If you don't want to handle BGP yourself, no reason you can't let multiple ISPs announce the address space on your behalf and route it to your circuit either.
IPv6 also handles multiple addresses per host by design, so even if you have lowend non-bgp connectivity you can multi home by assigning multiple addresses to the devices. This works ootb with the route advertisements.
Also having to use keepalives just wastes battery life on mobile devices, and the world is increasingly mobile.
Inband signalling of addresses is absolutely required for p2p to function, otherwise how do you discover any peers other than the one you're directly connected to? If you want to do away with inband signalling you'll end up with a purely client-server model which is a very bad thing.
Re: (Score:2)
I wish I could convince a last mile ISP to BGP announce my address space. Most of them will even only give a /64 so I can't run VLANs easily. Comcast will give a /60. Smaller ISPs have all acted like I was asking for a class A v4 to get more than a /64.
I don't know where the OP got the idea that NAT on v6 isn't a thing. NAT64 and NAT66 are both around. They have all the downsides of NAT44 and less of a reason. There is also 1:1, AKA NPT. Firewalls can do everything NAT66 can do from a security perspective,
Re: (Score:2)
If you want to do away with inband signalling you'll end up with a purely client-server model which is a very bad thing.
Next up on Slashdot, why does every damn thing come with a requirement for cloud connectivity! NAT is a disease on the network and it has lead to not only horrible network design but given corporations real power over consumers.
Re: (Score:2)
Inband signalling of addresses is absolutely required for p2p to function, otherwise how do you discover any peers other than the one you're directly connected to? If you want to do away with inband signalling you'll end up with a purely client-server model which is a very bad thing.
Sorry to pop your bubble. ISPs are not interested in letting their consumer customers enjoy P2P among each other. When they can mark up the price for commercial customers for a public IPv4 address, why spend money on implementations that hurt their business plan? Legacy CG-NAT is what more profitable to them.
Re: (Score:3)
IPv6 needs to embrace NAT
I wholly disagree. The 32-bit space made the cost of IP such that NAT is a better idea, but the 128-bit space of IPv6 kind of nullifies the whole reason for NAT on IPv4 in the first place.
And IPv6 doesn't have a way of dealing with multi-homing that won't explode routing tables and won't burden the servers and clients with complex routing decisions
Equipment already mostly handles this and hardware with IPv6 is ready to deal with multiple addresses bound to a single device. This is also why a lot of the networking tools require % to denote the interface. But outside of that you can always announce via BGP your own PI, because of the large space that's no longer an
Re: (Score:2)
NAT is what gives what marginal protection there is to be had for the vast majority of idiots using the internet.
The idea of flat addressing for everybody/thing is a utopian fantasy that ignores the hundreds of millions of script kiddies that escaped from a " diesel-smoking bus with hundreds of ebola victims" (https://www.hoboes.com/FireBlade/Politics/highway/)
Re: (Score:2)
A firewall is what gives what marginal protection there is to be had for the vast majority of idiots using the internet.
Re: (Score:2)
Why do you need NAT? There’s no reason for it when you have enough addresses for every cell of every being on earth. You do need firewalls, but the only time I can think of ever needing NAT in ipv6 is if you’re trying to multihome while not running a full BGP stack. That’s a pretty rare usecase.
Re: (Score:2)
Mobile IPv6 was supposed to be a thing, and was going to be a great tool in the use of predictable IPv6 addresses on portable devices like laptops, but while my Debian kernels still seem to have Mobile IPv6 (mip6) active in them, I can't find any current information on how to actually make use of this.
Re: (Score:2)
My ISP gave me IPv6 a year ago... Apparently implemented by someone who doesn't know too much about it:
One static address, and one "static prefix".
No router advertisements, no DHCPv6, no automatic prefix delegation. Came on paper; here are your addresses, set them up as you can.
Re:The longer you leave it, the worse it will be (Score:5, Informative)
If you deployed IPv6 years ago like you were supposed to then this would be a non event.
It's not going to be an event at all. This is simply Amazon screwing itself as their customers leave for Azure and other hosting systems. IPv4 works fine. You can wag your finger at the rest of us all you like. No one likes IPv6, and no one can make us embrace it. There is no address-cataclysm coming. We like the current system. We like NAT, it's useful. Telcos can use 6-bone for phones. No one cares about that. But start messing with our actual computers? Customers will make you feel the pain. If Amazon wants to fuck themselves, by all means, let them. The rest of us will point and laugh.
Original Source (Score:5, Informative)
Does AWS fully support IPv6 yet? (Score:2)
When they made this announcement there were still several features that require IPv4. It looks more complete now https://docs.aws.amazon.com/vp... [amazon.com] but I wonder how sharp the edge cases are.
the internet is ruined (Score:2)
The more people that got on here, the worse things got. Things went downhill fast once it started becoming a market place for spammers, scammers, and selling PII (personally identifiable information). The money grab ended up being about how to exploit users instead of how to sell them a service. And this makes sense, once we all had decent access to the Internet such a service simply became a commodity. And there is no exponential growth possible when you're simply offering a service that everyone needs but
Re: (Score:2)
You cant solve the software problem. Think about how many millions of lines of code got executed rendering this simple text reply to your post (its not even unicode! slashdot dont got no time for unicode)
There is plenty of room in the millions of lines of code that renders this text for some shenanigan's. What does all that code do anyways?
Wouldn't it in fact be surprising if there wasnt an enormous amount of shenanigan's happening? How else
Re: (Score:2)
> Would the Internet be better if it were operated centrally by a minister, soviet-style?
https://en.wikipedia.org/wiki/... [wikipedia.org]
Soviet style not needed, French is enough..
Re: (Score:2)
Would the Internet be better if it were operated centrally by a minister, soviet-style?
This is exactly how the internet operated until the late 90s. The minister's name was John Postel.
Good. (Score:2)
Hope other cloud providers start doing the same. More IPv4 for us.
Re: (Score:2)
IPv6 has been here... (Score:2)
A lot of cell phones are pretty much all IPv6. Depending on your ISP, you may not have an IPv4 address at all and use NAT64/DNS64 to get to IPv4 only sites, or carrier grade NAT with some local IPv4 address.
The turtle dances for a lot of people already (www.kame.net).
Re: (Score:2)
Yes, and accessing legacy sites through the NAT64 gateway is slower (sometimes significantly so) than native IPv6 sites.
Re: (Score:2)
Any added hop will do that. It's not any slower than the CGNAT gateway. Ideally, the ISP doesn't link the NAT64 to the CGNAT and double up. ugh. And you can always run your own NAT64 locally and it will run as good as your v4 does, with one local hop added.
Some ISP's (Score:3)
Re: (Score:2)
Well those ISPs will claim that there's no demand for IPv6. Those users should have demanded IPv6, and moved to a better ISP if their existing one won't provide it.
Based on google stats, something like 45% of the world actively use IPv6 to access google, and most of those ISPs are not at 100% usage so there are many more users for whom IPv6 is available but they're not using it for whatever reason (eg old router, bad configuration etc).
Re: (Score:2)
Therefore, 32-bit ip addresses have a real intrinsic value that ipv6 doesnt have.
Charging a premium for such an address seems reasonable.
economics man
Re: (Score:2)
My issue is my ISP only delegates a /64, and I have 10 VLANs on my home network. If you aren't supposed to split a /64 then you are stuck.
Re: (Score:2)
IPv6 is the protocol of the future (Score:5, Funny)
And always will be.
Re: (Score:3)
Re: (Score:3)
Somebody modded it funny, but it's very true.
coming? (Score:3, Funny)
Bullshit (Score:2)
IPv4 addresses have cost money for basically forever. That one specific service vendor gave them for "free" (i.e. hid the cost so far) does really not make a difference.
$4/month is also rather cheap. I pay $20/months for a static IPv4 address at home and have paid that for a few years now. Of course, that is a static address and from my ISP, so it is more expensive. I could use a vserver with a static address instead (starts at around $5/month), but that would mean tunneling with much reduced bandwidth or s
They aren't really charging enough for real change (Score:2)
The price that they're now charging for a public IPv4 address in AWS comes out to $3.60 a month. By AWS standards, that's a rounding error on your total bill.
It would probably cost you far more to rearchitect your solution to use an AWS load balancer or application gateway than to get rid of your Elastic IP's. Sure, you should do that for better security anyway, but it's going to take a lot more than this to get most people to change.
FiOS (Score:2)
Yeah, so Verizon turned off IPv6 due to buggy ONT hardware.
Their 6 vs. 4 rate went from like 60% to more like 3% (self-owned equipment).
IPv6 was too complicated, sorry. If it were 128-bit v4 we could have been done a decade ago.
Then moved on to other protocol improvements, probably by now.
Re: (Score:3)
This is a price increase through 'added charges'. (Score:2)
IPV6 is coming any century now.
Let me fix that headline (Score:2)
Re: (Score:2)
you don't need to switch off ipv4, you can use one IP and proxy that to your internal network... you know, like the 3R, to reduce waste:
Reduce, Reuse, Recycle
'Bout Freakin Time, Too! (Score:2)
I remember earnest talks back in 1996 about IPv6. I've heard those same conversations over and over and over and over and over and over and over again over the years. I was "ready" for IPv6 back in 2016, providing real world access to real people. And here is is, 2024, and we're STILL talking about the same da**ed things over and over and over again. CG-NAT and NAT have taken a LOT of US carriers farther than they ever could have imagined and because of that they STILL haven't implemented IPv6, so don't
Hypothesizing... (Score:2)
Hypothesizing about the technology infrastructure of a company he co-founded?
"my hypothesis is that AWS is a foundational layer to many infrastructure companies, like Supabase..."
Either he is using legalese or does not have a clue about the infrastructure...as Inspector Gadget would say "wowzers" or "go-go gadget brain..."
Yet not surprised Amazon would start squeezing more blood, err profit out of its users. Reminds me of the time when banks were charging all these fees and making more money than anything.
J
Nobody cares (Score:5, Informative)
If someone's using 1,000 IPv4 addresses on AWS, they're already being ridiculous. But what they'll do instead is just NAT/proxy that traffic to one IPv4 address and save $4,000 a month. But nobody's going through recertifying all their applications, retraining their devs and others, and whatever else to work on IPv6 when 95% of the world is still primarily using IPv4 and 100% of the world still fully supports IPv4 just because Amazon decided to charge a few bucks a month for it.
IPv4 will still be the primary 10 years from now. Probably 20.
Re: (Score:2)
Money grab is the correct answer. I will only be switching to IPv6 when forced to. My ISP only presents me with IPv4 here.
Re:Bollx (Score:5, Insightful)
AWS have spent quite literally millions of dollars buying legacy address space, they can't just go giving away those addresses for free, they have to recoup those costs. If AWS customers keep using them at the current rate they will end up having to buy even more, which is going to be even more expensive and eventually will hit the point that there is no legacy address space still for sale.
They also have to compete with MS, Google and others who are also trying to buy up the scraps of legacy space that's for sale.
Re: (Score:2)
not money grab, some people do abuse ipv4 address... multiple ec2 instances with public IPs, instead of just one public ELB connecting those, k8s clusters with public IPs, public DBs, and even reserved IPs, but unused...
not all people/companies have good admins and many times is just laziness... this will force many to fix those setups
Re: (Score:3)
Exactly the sort of comment I scrolled down to post!
"While $4 is relatively small ... this would add millions to our AWS bill"
How many IPv4 addresses (that aren't RFC1918 addresses) is the blog author using?!?!? If it adds millions to their bill, that means they're using at least half a million addresses. WTH man? Do they have 500,000 postgresql instances on 500,000 EC2 instances?
Re: (Score:2)
Re: (Score:2)
Scarcity adds value, and valuable property gets rented.
Re: (Score:2)
Imagine you own a parking lot. The marginal cost of letting someone park in a space is effectively zero. Your costs are fixed - even the miniscule amount of wear and tear on the lot's surface is made irrelevant by the damage done by other factors, such as weathering. Do you charge someone who wants to park for a day the same amount as someone who wants to park for a month?
Re: (Score:2)
sure, hide your head in the sand, ipv6 isn't global not because of users, it is because of ISP laziness ... if the ISP setup is done, users don't even know they are using ipv6
Most of my bandwidth is ipv6 and i and my family don't notice. i still have ipv4, of course, so when a remote server don't have ipv6, it fallback to ipv4 without issues. My ISP have ipv6 since 2017 and i was using sixxs tunnels before since maybe 2004
Re: (Score:2)
Re: (Score:3)
1. I don't think it's a pricing decision, more of a business continuity issue. If they can't move their customers away from v4 then AWS will hit a limit on its growth, because there aren't infinite v4 addresses around for AWS to use. They can't keep buying them on the open market because the price will just spiral up and up until AWS starts to cost them money rather than earn money. They'd eventually have to turn away clients, which isn't a good place to be in.
2. DNS. You shouldn't be remembering IPs in the