New on LowEndTalk? Please Register and read our Community Rules.
All new Registrations are manually reviewed and approved, so a short delay after registration may occur before your account becomes active.
All new Registrations are manually reviewed and approved, so a short delay after registration may occur before your account becomes active.
Comments
you know iv thought of the same thing but ipv6 is already here so it would make it even more confusing to make yet another standard.
I like how "IPv6 Is a Total Nightmare" and "IPv6 Is A Disaster" articles are written by people who most probably use CGNAT'ed IPv4 over their router with NAT and think this is how it should be done.
Running IPv6 for a decade. It, surely, has a lot of caveats and differences compared to IPv4 in tech and policy ways, but I in no way could call it a nightmare or disaster. IPv6 networks are simpler from many standpoints, and I know companies which run IPv6-only networks internally.
And by the way, IPv6-only networks doesn't mean inability to reach IPv4 addresses, which is usually done with some kind of tunnel or mapping technology. But it is considered as a fallback, exactly for which the articles mentioned above complain. There are IPv6-only cellular networks with 464XLAT.
And there is also the fact that "IPv6 will soon(TM) push out IP4 and be THE protocol" has proven wrong for over 2 decades ...
For good reasons I presume. And corporations and people have done a lot to stick with IP4 and to keep it not only alive but the still ruling IP version.
Funny btw how IPv6 fans quite consistently ignore arguments they don't like.
@Francisco
Plus: creating software using IP4 is easy and addresses a huuuge market. creating software using IPv6 is a PITA and somewhat of a lottery (as described well by you) and addresses a small market
The examples given of problems with IPv6 are silly. If my ISP supports IPv6 and I don't use Github or Datadog (and certainly many don't), these problems vanish.
To draw an analogy: electric vehicles at present are less convenient than gas-powered vehicles because there are far more places to fuel gas vehicles. Does that mean there are a problem with EVs? Would you propose we abandon EVs because they're less convenient? How would any technology ever be adopted?
I don't understand your argument. You're not talking about IPv6 technical shortcomings but rather that at present it's less convenient than IPv4 because IPv6 is not as widely adopted.
"guarantee" no (how could anyone?) but I'd bet it will. That's kind of the point of IPv6. I know you don't like its 128-bit nature, but one of the benefits is that the address space is so vast it's hard to imagine humans ever running out of IPv6 IPs.
I seem to recall "IPv6 should be 64-bit" advocates arguing that a 64-bit address space is so huge we'd never run out of 64-bit IPs. Well, then we certainly shouldn't have problem with 128-bit. A resource that is infinite for all conceivable purposes will remain free because there's never going to be any scarcity.
We're all going to be dead of global warming in far less time than it will take to exhaust ipv6, so don't worry...well, at least not about IP space...
The biggest issue is also that IPV6 isn't a first class citizen in many networks. I've had more than a couple IPV6-only outages in Luxembourg where it'll be down for a day or three while they address it, but it's low priority.
This happens with end user ISP's too, and that fucks things even worse because you got a V6 IP, but the routes dead somewhere along the lines and they don't give a hoot.
Francisco
IF everyone along the way plays along ...
Btw, how much cheaper is DSL or fiber service due to not needing an IP4 or a fraction thereof? (Actual comparisons please, not "my new IP6 contract is cheaper than the former IP4 service from another ISP I used, please.)
(a) AFAIK nobody (significant) states that !EVs will soon (TM) ruuule the world!!!!
(b) ask me again in 15 years because EVs have AFAIK not yet failed a not-made promise for over 20 years.
Nope. Because of the reasons it not widely adopted.
Same would be true for 64-bit address based internet.
Say, you owe me $10 and to pay that is no problem. Well, according to the logic just presented, paying me ca. 40 billion $ is no problem, right?
Another angle: OK - but why just 128-bit addresses? 256-bit addresses would be even more infinite. But then, maybe the whole idiocy, uhm argument that 4 billion times today's IP4 address space is not enough wasn't that smart of an argument in the first place.
The answer to the problem "we'll probably soon not have enough addresses" is not "OK, lets go insane and come up with a scheme that provides more addresses that the whole solar system could ever possibly need!". It rather is "Ok, lets come up with a scheme that provides enough plus a solid reserve for the next say 100 years!" (one reason being is that 128 and even 256 bit processors actually are very likely to exist and be cheap in 100 years)
face palm No, please not the next nonsense wave. And btw, what you said would also be true for 64-bit addresses
Rest of the debate aside I'm a little surprised someone who has never set up an ipv6 capable system until 2023 is referred to as an infrastructure professional.
Right now the best thing to do is be dual stack. That'll change in future. That's been the case for at least a decade, probably a lot more.
Someone who hasn't bothered until now - honestly, I don't really see their opinion as all that important.
It appears that providers selling IPv6-only VMs ought to provide a NAT64 service as well, to deal with cases like needing to access GitHub and such from those VMs. Shouldn't expect third-party hobbyists or companies to run NAT64 for everybody, since those services are basically a free anonymous public proxy, and can be easily abused and then will shut down -- unlike when operated by each provider, limited to their customers only.
In general, the main thing this kind of articles should make you do, is to each time take a look at what else can you make IPv6-capable or enabled today. That's the only way to win!
2024 will be year of the Linux desktop IPv6
Since almost 25 years I've been interested in IPv6 and playing with it. I like the standard, I like how it solves some flaws of IPv4. It may not be perfect, but it's better (futurewise) as IPv4.
The issue in my opinion is not the standard. It's the adoptation. Everybody who blames the standard, please clearly state where the issue lies in the standard. It may be a bit more complex, but that shouldn't put off "real" network people.
I always like to draw a parallel in payment systems. You have cash and debit cards. Debit cards solve some issues of cash, but it's not adopted everywhere. As long as you use services that you can pay with debit cards, it isn't an issue to stop having cash. However, when all, or the bigger, supermarkets refuse to accept debit cards, you always have to carry cash with you. And also, as long as you can pay with cash everywhere, why would you switch to debit cards. Then is the issue the debit card or the adaptation? If tomorrow the biggest supermarkets dare to say "we only accept debit cards" it will lead to a much bigger adaptation, but they don't dare because what's the advantage for them...
And this is the situation with IPv6 now for over 15 years. Don't blame the standard. Blame the companies who haven't implement it yet.
No
Have experienced a lot of the same. My home ISP claims to support IPv6, but no articles exist and nobody at their tech support has any idea what I should be using to make it work, so I just gave up. I don't think they actually do, I think they have(had?) some sort of tunnel service with their included wifi router/modem thing that I don't use.
I believe it's similar from a provider's POV. The effort we had to go through to get an automated scalable (routed) v6 solution was immense, and likely will never pay for itself with regards to customers gained we otherwise may not have attracted. Then when you encounter some little bug, you have the minority of extremely noisy v6-only zealots spamming your tickets wanting to get into detailed arguments with support staff because their OS of choice manually installed from 2016 isn't compatible due to some obscure RFC. I really think there is a big disconnect on how you implement v6 stuff at scale (enterprise/scalability) versus just using it as an individual and hobbyist. It's not a linear approach from what you do with v4, especially when the expectation is that every $1 VM must have 18,446,744,073,709,551,616 usable host addresses.
I personally don't have a problem with v6, and I think something should be done with regards to address space exhaustion proactively, but I'm not a fan of how v6 exists today. I'm probably a stick in the mud but I also have been witnessing pretty big drops in IPv4 pricing over the last quarter. We've had purchase offers for /24s drop into the low $7k range and long term leases in the ~$0.15c/IPv4 range (from reputable big companies). Still expensive, but a significant drop from how crazy things went in 2021-2022. As long as Ford can continue squatting on a /8 and DOD with their slew of public /8's and beyond, I think the exhaustion argument is a little bit premature.
Either way, I'm very thankful I'm not responsible for v6 policy/integration/making-it-work for us.
Mostly hate the brokers, RIRs empty because there is a big market for selling IP space.
Yeah, I want to make a video collage of the broker sales rep's increasingly nervous tone in emails over the last 6 months.
It wouldn't be so bad, but the only rep I dealt with that acted as if you were another human left the industry (not surprising, I guess).
That is, in fact, simpler on IPv6 as AS usually announce big ranges (/32 or /29) instead of thousands of /24 IPv4 blocks. Fragmentation is a problem with IPv4.
So you do. People blame IPv6 being hacky and cumbersome yet they insist on doing [double] NATs and limiting ports, which completely breaks end to end communications.
This is why libraries exist to support IP protocols transparently. I can assure you IPv6 is easier than having to deal with NAT traversal.
@rm_
Kudos! That's a respectable position. Certainly not mine, but respectable and constructive.
I personally will never consider IPv6 a good solution, because it isn't, it's insane from the premise.
BUT: Should the day ever come when IPv6 - for the vast majority of people - comes down to setting a flag in the OS config, and then have everything just working without any significant hiccups, I will stop to strongly advise clients to avoid IPv6 like the pestilence it (still) is. Kindly note that this includes reasonably priced (about like IP4 stuff) equipment, ISP service, etc. Should it even be cheaper I'll give IPv6 the first '+' point (next to all the '-'s).
Boring. That's my wife's standard "argument".
So you dislike NAT, OK. But that's based on a false premise, namely that both IP4 and IPv6 can be actually used and do work.
Also kindly note that the OP (linked) blog post is from an IPv6 fan who came to realize what a pile of crap IPv6 actually is.
What a nonsensical non-argument! (a) Libraries have nothing to do with that, and (b) I do work, among others, in a language that offers a "super class" IP_Address, which - in theory - works with both IP4 and IPv6 ... and quite reliably the software fails in "IPv6 mode" due to diverse factors.
And btw. most software doesn't even care (or even just know about) whether NAT is used. A program (almost always) just sees an interface, possibly an IP, usually a port - and that's it.
Do you think that so many developers ignore/circumvent/avoid IPv6 because there's a conspiracy going on? Nope, they avoid it because the burden and troubles are large but reward is very small.
Btw: I've seen quite a few clients who superficially asked about IPv6 - but none who actually seriously requested it.
Exactly. I never said IPv4 doesn't work. But it doesn't work they way it should (aka no NAT). They should coexist for a while until IPv4 isn't deemed necessary.
Both NAT and IPv6 main goals exist as a solution to overcome IPv4 shortage. This varies from country to country, where they have more unique IPv4 addresses than population.
But having proper IPv4 support costs me money, as I have to pay an extra to get a public IPv4 address so I can open ports. With IPv6 that problem is non-existent.
Then these "diverse factors" should be reported to the developers, which may or may not fix them, but at least they are aware of it and maybe an "IPv6 enthusiast" can submit a PR to fix it.
From a user standpoint, you are correct, they might not even notice it. (Unless they have to do port forwarding, which is a "virtue" of NAT)
From a service provider standpoint, I do notice when clients can't connect with FTP active mode or how they are unable to reliably host game sessions or join WebRTC calls.
That is, of course, until I set up a "hacky solution" to overcome NAT inherent problems. Be it a range of ports for PASV mode or an STUN server.
I wish my ISP would have ipv6 connectivity. When I inquired about it's availability, they simply said "We have nothing planned for it in the future".
Get a different ISP that does have IPv6
In India, that's not possible. You have no options if you're not in an urban City.
And even then, I think there's only 1-2 ISPs which give v6 by default.
Sadly that's not always feasible; in the UK at least, most of the major providers have been "working on" providing IPv6 for literally decades; they simply don't see the need (presumably because all the upper management has the same enlightened attitude as jsg).
Proper IPv6 support seems to be the domain of smaller, more specialist ISPs who are inevitably more expensive.
To be fair to the major providers; your average broadband customer has no clue what IPv6 is and really doesn't care as long as Facebook and Netflix works, so there's no mass market interest in it. That will only change when it's cheaper / simpler to deploy IPv6+NAT64, rather than IPv4+CGNAT at ISP scale.
>
I get it and I agree to a degree. Who doesn't like free ...
But: A dollar or two per month isn't a high price for easy riding and a bit of comfort; I wish all problems could be solved for a tenner or two a year.
Also, for the vast majority of users we talk about a tiny fraction of that amount (one of the positive sides of NAT at ISPs, but granted, there are unpleasant sides too).
And for the rather tiny fraction of people who actually and really need their own full server with their own IP a tenner or two a year is a nuisance but not a problem.
That's one possible way. Another one and the one bloodily evidently chosen by a large majority is to simply say "f#ck that crap"".
(a) My statement was from a developers perspective, but if you want to see it like you wrote ...
(b) I have/had both dynamic IPs (and some form of NAT) and static ones. As a normal user (not running public services from my office and home) both work fine. If you insist I can call on billions of witnesses. It works fine.
As a service provider one should put the services on public IPs.
Likewise in Canada, where IPv6 support still hasn’t landed with the country’s largest ISP.
The only other provider in my area uses cable, which is a significant downgrade from my current fibre connection (just to get V6).
So I think in the end that the question comes down to the technical problem's with IPv6 and the practical problems with IPv6.
Alot of people here are saying that IPv6 is a bad standard and that it should be replaced by a new standard, no offense but your a meme at this point(https://xkcd.com/927/). There are most definetley flaws with IPv6 but most of the flaws noted here have solutions for them. Such as too long to remember, we have DNS for that. While it's not a perfect solution making a new standard for it just doesn't make sense. So for the people saying we should build a new standard instead of just saying "huh duh make new standard" how about explaining what that standard would look like ACCORDING TO YOU, instead of just proclaiming we need a new standard.
For those saying IPv6 will also run out of adresses, yes in theory it could. We said that 2^32 should be enough adresses forever when we got IPv4 so probally in the end the same will happen with IPv6. But you have to realise that IPv6 has 2^128 adresses. To put that into perspective the estimate for the amount of stars in the universe is 1 septillion or 110^24. 2^128 gives us 3.4 * 10^38 adresses. We could assign each star in the universe 3.410^14 adresses or to put it into a normal number 340000000000000 adresses per star in the known universe. So while in theory we cannot know what will happen in the future we SHOULD not run out for a long time.
So let's move onto the practical problems and this actually made me laugh, we have people in this thread saying "IPv6 doesn't work for everything it's bad" while at the same time we have people saying that they by default completley disable IPv6 on there services. Ah yes that will definetley help.
And another person was referencing how some (mostly legacy) programs don't support IPv6, but this isn't a problem with IPv6 this is a problem with the implementers of the program. I think a good analogy here is saying "We shouldn't switch to electric cars because there are more gas stations than electric recharging stations". Your not wrong, it's more convient at this time to use IPv4, but that isn't because IPv6 is a bad standard as some have tried to present it as being.
I personally hope that IPv6 continues to see more adoption, but I don't agree with the people saying we should deprecate IPv4. It should be a choice to what to use, and what's a good way to see it is some companies may continue to use IPv4, but this will become more expensive over time because v4 adresses will become more and more scarce.
That's indeed not a problem. It's the other way around, IPv6 has too many addresses.
Secondly, back then when IP4 was invented, this planet's population was already over 1 billion that is, it should have been recognized that 32-bit addresses (~ 4 bln) risked to be too small an address space. That is where the whole problem story had its beginning and that's a point IP4 indeed deserves to be criticized.
The right solution would obviously have been/be to use 64-bit addresses - but then, to be fair, there's also always the practical side, the engineers perspective. Back then 32-bit processors began to be available and seemed to be - and indeed were for quite a long time - a technical limit.
Considering what is needed (ca. 6 - 10 bln IPs) and desirable (a reasonably large reserve on top) as well as what is easily and at acceptable cost and efforts (64-bits) practically feasible it should be obvious to everyone with a working brain that 64-bit addresses would be the solution.
What also everyone with a brain should understand is the fact that one should not willy nilly make yet other changes without real and urgent need to something that is in use world wide and by far over a billion people.
So, the correct, good, adequate, reasonable - and with very high likelihood easily and quickly accepted - solution would have been/be to create an "extended" version of IP4 with 64 bit adresses and change nothing else beyond pure 32 to 64 extensions.
But no, some weirdos in a mental asylum were let loose to run amok - and IPv6 didn't find the vital acceptance.
Uhm, sorry, but a program doesn't magically go "legacy" just because it doesn't support some weird badly and ignorantly designed trend (which IPv6 is IMO).
OK, if you see it that way, but there are other, certainly not less valid, perspectives. In particular, one could - and actually should - ask why it's not implemented.
Sorry, but your analogy is bent. It should be "because still, after over 20 years there are more gas stations than electric recharging stations".
IPv6 did get a fair chance, in fact, it was massively hyped and often propagandized as "the new standard. If you don't jump on that waggon you'll end up alone in the desert" - yet it failed to really find the vitally needed wide acceptance and uptake.
Your fuel vs EV analogy has another and major flaw: EV cars are often more expensive and deliver less mileage/reach. That is one major reason for not buying an EV.
Unfortunately one reason for increasing scarcity is that 20 years were wasted hyping a crappy idiocy - instead of letting grown up engineers develop a good and actually fully usable replacement for IP4.
That's why I keep promoting "IP4, just extended to 64 bits and as far as any feasible no other changes".
All those "Ip4 needs a wheelchair" vs "Ipv6 is crappy and not widely accepted" discussions (actually debates) are only useful to analyze the situation and the needs (vs. not really solutions).
The next, important and overdue step is to discuss how a good and adequate solution should look like. How about starting that next phase instead of basically lamenting over two existing protocols one of which will sooner or later suffocate and the other one failed to replace it for over 20 years (and IMO will continue to fail).
if IP used variable length encoding like utf-8, how long address we'd actually using?
Using IPv6 from the very first tunnelbroker days, and as a datacenter and network operator: Yes.
Too much complexity added, simpler is better.
Been proposing for this for a ~decade, and there actually is an half-baked RFC or two about this, but if i recall right both propose added complexity messing with layers (like ipv6) such as content classification.
Keep the current IPv4 but extend it so that the end point target can act as gateway to the extended address range, and getting to that extended IP, it can act for next extended address range. Or at the very least go for 2 extension bytes from get-go if not dynamic depth.
This requires zero change to the backbone // internet routing at all, zero changes to routing table, zero changes to core infrastructure, only end points and mostly software even at that point. Just start by a simple patch to every OS TCP/IP stack, regular update, and it will automatically gather what 95% adoption rate within a decade? The last few remaining are likely industrial etc. and they can do their own gateways if they really have to use IPv4e.
You only need to convince 3 parties: Microsoft, Linux kernel devs, Apple to get mass adoption.
Eventually ofc hardware support is needed for the end point gateways, but initially can be just software gateways.
Packets with IPv4e addressing would travel through the wider internet just like any other packet, no need to convince every single network admin and business to adopt it. Only those 3. If just Linux kernel adopts it, eventually MS and Apple will follow -- so in reality, just 1 party needs to be convinced to include the patch.
A decade later we can start accepting smaller than /24 subnet routes on the internet, perhaps, so in 2 decades people could route something like /26 or even /28 -- so more ASNs can join.
Any other feature creep should be avoided like plague to ensure this gets adopted. There are other things that should happen, but perhaps as separate RFCs. Would like to see the internet start to move to larger packet size for example.
Bottomline is: You absolutely need IPv4 for servers, no way around it, and IPv6 only works for few % of the time in reality. Us providing seedboxes? Someone did metrics etc. it was like 0.5%.
Not worth all the added cost.
Further, i doubt IPv6 will ever be fully adopted because it needs convincing too many people, and making life more difficult for too many people. Where as, something like IPv4e is implemented as simply as: apt update; apt full-upgrade;
I cannot imagine how you can achieve that zero-anything stuff with proposed an additional IPv4 header option - you would certainly need to invent changes for routing table to be able to appropriately handle addresses with equal 4 highest bytes...
Even further - such change will be much harder (comparing to IPv6) for both hardware and software firewalls, as well as other hardware equipment which has ACL functionality.
I would argue that such changes are much harder to adopt than IPv6. Maintenance of such bloated specification is very error prone task.
It gets even worse, stability and reliability issues when IPv6 is enabled is bound to happen because somethings just "strongly prefers" ipv6. Causing things not to work.
Further, you now have 2 stacks to worry about routing etc. connectivity wise rather than 1.
You have 2 completely different routings as well, even more to debug and worry about. twice the debug work to start, 5minutes quickly becomes 30minutes.
If it were the holy grail advertised -> it would reach 100% adoption very quickly. But it is not.
I know of a significant carrier telco which strongly advises their customers not to use IPv6, and sometimes gets their customers to disable IPv6 because of the issues caused.
Unless you are at the extreme bottom end of the market a customer is worth significantly more that the cost of an IP4.
This. So many times this. IPv4 is a requirement. IPv6 is at best luxury.
By definition, IPv6 addresses are being wasted. Large allocation for every single VPS, advocating using new address for each application instead of ports etc.
The IPv6 routing table is already taking nearly as much space as IPv4 routing table. And that memory is hella expensive, think L1 cache level cost. Every single byte is very precious. That is part of the reason proper routers routinely cost 100+k$
I guess you have DHCP everywhere as well, and the MAC address to IPv6 mappings done too. Good for you.
Here in the real life of a sysadmin managing a DC and network: That never happens, you have to keep typing those addresses manually quite often, and even without means to copy paste.
Since you don't have network access, you won't have DNS working neither.
and here we go. This is why you fail to see the pitfalls, you don't actually have to work with IPv6.
Actually, precisely because of simplicity, security and performance it remains the #1 choice by far.