| |||||||||||||
What about IPv6 right here on earth?What about IPv6 right here on earth?Posted Jan 26, 2011 8:51 UTC (Wed) by cmccabe (guest, #60281)In reply to: What about IPv6 right here on earth? by tialaramex Parent article: LCA: Vint Cerf on re-engineering the Internet
I think DJB is arguing that there should have been a transition period where people moved to IPv6, but still continued to use 32-bit addresses. This could have been handled by setting aside a special part of the IPv6 address space for addresses that mapped directly on to IPv4 equivalents.
Then, over time, operating system vendors, Linux distributions, and network equipment manufacturers could have moved to IPv6 "painlessly." IPv4 would have been phased out-- you literally would have been unable to buy IPv4 gear or download IPv4 software-- just like you can't buy IPX or Token Ring gear any more.
Then, when the big crunch came, as it is coming now, we could all look around at our modern IPv6-only gear and have a chuckle at the expense of those old Windows 95 users who still had IPv4 equipment. Finally we would flip the switch, allowing everyone to use all 128 bits of IPv6.
Instead of doing this, the IPv6 designers decided that they would create a completely separate network namespace. If you choose to support IPv6 on your network, the burden of it falls on you. It's measured in terms of time spent, equipment bought, and so on. You won't see any benefit to supporting IPv6, however, until a tipping point is reached where "enough" of the internet uses IPv6 (for some definition of "enough".) At that point, everyone will benefit, probably including the people who dragged their feet during the conversion. So the rational, profit-maximizing strategy is to ignore IPv6.
In fact, address scarcity often helps the incumbent telecoms companies. Having control of a scarce resource, like IPv4 address ranges, is usually considered a good thing.
The working group should not have made ignoring IPv6 an option. They should have pushed for it to become the only choice for newer systems. The only way to do this would have been to have a backwards compatibility mode.
(Log in to post comments)
It just shows how badly even intelligent people may misunderstood the simple problem Posted Jan 26, 2011 10:58 UTC (Wed) by khim (subscriber, #9252) [Link] I think DJB is arguing that there should have been a transition period where people moved to IPv6, but still continued to use 32-bit addresses. What will it change? This could have been handled by setting aside a special part of the IPv6 address space for addresses that mapped directly on to IPv4 equivalents. This is done already. Then, over time, operating system vendors, Linux distributions, and network equipment manufacturers could have moved to IPv6 "painlessly." Operation system vendors? Yes. Linux distribution? Of course. Network equipment manufacturers? No way in hell. You see, IPv4 requires 4 bytes in the routing table, IPv6 requires 16. If the given piece of silicon is used to do lots of things and handle IP traffic as well you can easily add IPv6 support. But if the equipment's primary reason for existence is TCP/IP then switch from IPv4 to IPv6 (without magic function) is expensive. This is why we have situation where hardware all around supports IPv6 with one exception: ISPs don't support it and most have no plans to support it. Intermediate step with 32bit addresses will just increase the mess. Instead of clear IPv4 -> IPv6 step it'll introduce two steps: IPv6/32bit and IPv6. Industry will enthusiastically embrace first step (because of it's PR value) and will delay the second step as long as humanly possible. The working group should not have made ignoring IPv6 an option. They should have pushed for it to become the only choice for newer systems. The working group does not build the hardware, so I can not push anything this way.
It just shows how badly even intelligent people may misunderstood the simple problem Posted Jan 26, 2011 11:14 UTC (Wed) by bojan (subscriber, #14302) [Link]
From what I can tell, DJB wrote his piece in 2002. That's over 8 years ago. How much IP sensitive network equipment is around from then or even before that when IPv6 was designed? Problably little.
It is a simple fact that right now we cannot reach any IPv6 hosts from pure IPv4 addresses that have been around for a while. All the software and networking equipment is way newer than that and it just doesn't work. That is interoperability failure, whether you want to admit it or not.
A brief catch up Posted Jan 26, 2011 12:47 UTC (Wed) by tialaramex (subscriber, #21167) [Link]
Yes, DJB wrote this in 2002. Yes, in 2011 most ISPs still haven't bought IPv6 capable equipment (and those that have never configured it to enable IPv6). This isn't an interoperability failure, it's classic market failure. If IPvDJB actually did anything, and thus cost money, the ISPs wouldn't have bought that either.
The ISPs chose not to buy IPv6-capable equipment. Cisco and others have been selling IPv6-capable gear for about a decade, and the hardware routers (which need custom hardware or at least FPGAs to process a different unit address size) have been available for more than five years. ISPs haven't been saying "Do you make an IPv6 capable version of the thing we bought last time?" they've been saying "Is IPv4-only cheaper? We are putting off all non-essential purchases to drive down costs.". The ISPs chose not to deploy IPv6 capable end user devices to their customers because it keeps the acquisition cost low. The ISPs chose not to do any configuration work, because that means manpower and technical manpower is expensive. Plus it might mean engineering outages, and customers hate those.
Left to their own devices the water companies in my country don't repair water mains, the train operators never buy new trains, the nuclear power industry take increasingly unsafe shortcuts. So we have to regulate them in order to force them to internalise long term costs. We don't regulate Internet service provision to this extent, many believe we shouldn't have to. Perhaps not, but this time it's going to cost us.
Hardware people, like Cisco, and software people, like Microsoft or Linus came through on this, not always as fast as we'd have liked, but in time at least for the inevitable crunch. That left only one key stakeholder, the ISPs and they're twiddling their thumbs. But by all means blame the engineers, nobody had any doubt from the outset who'd get the blame. If we don't do the impossible day after day it's because we just weren't trying.
The "simple fact" that several of you seem hung up on is a mathematical truth. It's not avoidable by any amount of shenanigans on our part. Repeating that someone, somehow, should have found a way around this is nothing but active denial.
A brief catch up Posted Jan 26, 2011 13:40 UTC (Wed) by bojan (subscriber, #14302) [Link]
Right, the plan that produced something that doesn't work was unavoidable and people that pointed that out 8 years ago are in denial.
Have you tried running DOS programs on Windows? Many of them worked just fine.
Small mixup... Posted Jan 26, 2011 13:47 UTC (Wed) by khim (subscriber, #9252) [Link] Have you tried running DOS programs on Windows? Many of them worked just fine. Well, sure. It's called 6to4 and it works really well. But this was not Bernstein's idea. His idea was to run IPv6 over IPv4 network - essentially run Windows programs on DOS. Can you say that "worked just fine"? If yes then you come from some other universe then me... Even Windows 3.11 stopped using DOS for HDD access and when Windows 95 switched to so-called "virus compatibility mode" it became practically unusable.
Small mixup... Posted Jan 26, 2011 14:02 UTC (Wed) by bojan (subscriber, #14302) [Link]
Actually, his idea was to extend the IPv4 to become IPv6, so that v4 is included in it. This is what happens when you run a DOS program inside any NT based Windows. You are not really running on DOS, it just looks like it to the program.
And given that all the software would have understood 16 byte addresses by now, the transition would just happen.
Practical example: my home network would have been on IPv6, although all my addreses would still be written in 4 byte form in my config files. My effort in this: zero. ISP effort in this: close to zero.
Small mixup... Posted Jan 26, 2011 14:44 UTC (Wed) by khim (subscriber, #9252) [Link] Practical example: my home network would have been on IPv6, although all my addreses would still be written in 4 byte form in my config files. My effort in this: zero. ISP effort in this: close to zero. Wow, great achievement. How it'll be different from the situation today: if ISP supports either SLAAC or DHCPv6 then your home network will be on IPv6 - the only changes needed are new hardware on ISPs side? It's hard to find contemporary OS without SLAAC (and most support DHCPv6, too), you know. The problems with IPv6 deployment are not related to problem described in DJB's article. It's simple economics. ISP economics to be exact. Other problems were fixed long ago. May be not in a way DJB likes, but they are fixed.
Small mixup... Posted Jan 26, 2011 15:13 UTC (Wed) by bojan (subscriber, #14302) [Link]
Aha. And they will buy my IPv6 router? Configure my AAAA records? Setup my IPv6 firewall? Reconfigure all my services? Ditto at my employer that has this multiplied by the factor of at least a 1000?
Please. This plan is an utter disaster. If it was any good, nobody would have to touch a thing.
You know, kinda like this: when I send texts from my mobile, I couldn't care less whether 3G or GSM is used. A phone I purchased almost 4 years ago could do both automagically.
Small mixup... Posted Jan 27, 2011 12:30 UTC (Thu) by cesarb (subscriber, #6266) [Link]
> when I send texts from my mobile, I couldn't care less whether 3G or GSM is used. A phone I purchased almost 4 years ago could do both automagically.
Yes, the same way you can use IPv4 on top of either wired ethernet (802.3) or wireless (802.11). A laptop purchased more than 4 years ago could do both automagically.
What we are talking about is more like changing the phone number format or the SMS packet format, not switching between technologies lower in the stack (3G/GSM and 802.3/802.11 are all link-level), so your analogy is a red herring. Changing the link (which affects only that link) is a local decision, and thus much easier than changing a higher-level protocol (which affects everyone).
Small mixup... Posted Jan 27, 2011 1:36 UTC (Thu) by daniel (subscriber, #3181) [Link] His idea was to run IPv6 over IPv4 network - essentially run Windows programs on DOS. Can you say that "worked just fine"?Why yes it did, it was called "Windows 95" then "Windows 98" and finally began breaking down with Windows Millenium. However it bought Microsoft enough time to migrate their application software to an operating system worthy of the name. Yes, Windows on DOS worked out very well indeed, unlike IPv6.
It just shows how badly even intelligent people may misunderstood the simple problem Posted Jan 26, 2011 13:36 UTC (Wed) by khim (subscriber, #9252) [Link] From what I can tell, DJB wrote his piece in 2002. That's over 8 years ago. Yup. How much IP sensitive network equipment is around from then or even before that when IPv6 was designed? Problably little. Wrong question. Right question: how much IP sensitive network equipment without IPv6 support are still produced today? Probably a lot. People are delaying IPv6 as much as possible - and without "magic function" you can not change anything there. This is well-known phenomenon and you can not change much there. Bernstein explains how to solve non-problem: how to marry endpoints if the routing infrastructure does not need an upgrade (as was in the case of MX records: you needed change on Internet endpoints, but it didn't affect routers in the middle). IPv4-to-IPv6 transition does not need to solve such problem because it does not exist: most endpoints have IPv6 support (and had for many years) already. The real problem (the need to upgrade routers "in the middle") is silently ignored in the article. It is a simple fact that right now we cannot reach any IPv6 hosts from pure IPv4 addresses that have been around for a while. This was never in the plan - and rightfully so. All the software and networking equipment is way newer than that and it just doesn't work. Sure - but this is because IPv4 addresses are still available! They are expensive, true, but they are available! That is interoperability failure, whether you want to admit it or not. No, this is application of "Internet only just works" principle. IPv6 will be employed when it'll be needed - and not before. IP, CIDR and other network-wide specifications were introduced under extreme pressure when other options were exhausted - because any other way just does not make any sense from economics POV. I suspect that LTE which will introduce hundred of millions IPv6-only clients will be the required push, but then again, may be not (proxies will mitigate the problem for a while). But Bernstein's plan is just stupid: the fact that you need new kind of IP number supplied by ISP never was a problem (and still is not a problem). The fact that ISP must replace it's expensive hardware was and is a problem - and article kind of ignores that problem completely.
It just shows how badly even intelligent people may misunderstood the simple problem Posted Jan 26, 2011 13:52 UTC (Wed) by bojan (subscriber, #14302) [Link]
Sorry, but you completely misinterpreted what DJB wrote. He was talking about upgrading everything on the net with software that understands both types of addresses. This includes routers. This was 8 years ago - plenty of time.
The reason this wasn't done is that IPv6 in its form (as proposed) was useless and not interoperable with IPv4, ergo nobody wanted to spend time configuring something that had no application.
The whole thing should have happened transparently, so that current IPv4 site didn't have to change a single thing to work with IPv6 addresses. If network manufacturers received that message, there would be no question which equipment to buy. It would be one and the same. And you would not have IPv6/v4 stack combos on OSes - just IPv6 that included IPv4. That's the point that you missed.
Who will pay - this is the question... Posted Jan 26, 2011 14:20 UTC (Wed) by khim (subscriber, #9252) [Link] Sorry, but you completely misinterpreted what DJB wrote. He was talking about upgrading everything on the net with software that understands both types of addresses. This includes routers. This was 8 years ago - plenty of time. To throw good money after bad? All pieces where it was cheap and easy to add IPv6 support were upgraded in these 8 years, only routers are the hold-out - and they are real problem. The reason this wasn't done is that IPv6 in its form (as proposed) was useless and not interoperable with IPv4, ergo nobody wanted to spend time configuring something that had no application. No, the reason it was not done is that it's more expensive to install IPv6 on router and gives you no benefits. DJB does not address this issue at all. In his plan ISPs will magically decide to be altruists and install more expensive and useless hardware for the sake of the future. This is not how ISPs operate if they want to survive. The whole thing should have happened transparently, so that current IPv4 site didn't have to change a single thing to work with IPv6 addresses. Have you read the article? This is definitely not what I'm seeing: (2) I control the operating system and the applications. I am ready and willing to make various changes to the code. This asinine dilemma does not change anything WRT to real problem. If network manufacturers received that message, there would be no question which equipment to buy. It would be one and the same. How come? How exactly you propose to make IPv6 router as cheap as IPv4 router? Remember: IPv4 routers are highly-optimized pieces of ASIC which are optimized for particular bit-layout of packets (if you use optional flags in IPv4 they slowdown by factor of 10x-100x, these packets are dropped early, etc). This is critical question - and both you and DJB keep to ignore it. And you would not have IPv6/v4 stack combos on OSes - just IPv6 that included IPv4. There are very few OSes without IPv6. The problem lies with networking hardware. You know: FPGA, ASICs - things which are expensive and hard to change. That's the point that you missed. I've not missed it: as I've said it's irrelevant. The problem which this approach was supposed to fix either does not exist (we can change OS and everything else, but can not ask for a new IPv6 address - WTF? why no?) or impossible (we want to participate in IPv4 network using only IPv6 address: how?). The real problem is not discussed at all: DJB presumes that it's easy to change hardware/software on ISP side and hard on the client side while in reality it's the other way around.
Who will pay - this is the question... Posted Jan 26, 2011 14:36 UTC (Wed) by bojan (subscriber, #14302) [Link]
So, IPv6 is designed to embed IPv4, standards are witten, all software manufacturers start implementing IPv6 that includes v4 (i.e. understands 16 byte addresses as well), but network equipment manufacturers (according to you) do not implement this at all because they cannot redesign their ASICs to do that in almost 10 years. And on top of that, ISPs do not buy a single new router in that time.
Ten years ago all Cisco routers were routinely accessed via telnet. These days folks mostly use ssh. Things change when the right signals are given.
IPv6 transition is being handled rougly like the 2000 bug. At the last minute people are scrambling to cobble together workarounds. At least old programmers had a good excuse - space was at a premium.
Who will pay - this is the question... Posted Jan 26, 2011 14:59 UTC (Wed) by khim (subscriber, #9252) [Link] So, IPv6 is designed to embed IPv4, standards are witten, all software manufacturers start implementing IPv6 that includes v4 (i.e. understands 16 byte addresses as well), but network equipment manufacturers (according to you) do not implement this at all because they cannot redesign their ASICs to do that in almost 10 years. Yeah, let's go with strawman. Of course they support both IPv6 and IPv4. For example Nexus 7000 M-Series (found in one minute using Google): up to 60 million packets per second (Mpps) of IPv4 unicast forwarding traffic and up to 30 Mpps of IPv6 unicast forwarding traffic. Price difference between IPv4 and IPv6 is exactly two times. And on top of that, ISPs do not buy a single new router in that time. Sure they do, but they disable IPv6: this gives 2x price saving. They are not stupid: why spend $200'000 when you can spend $100'000 and give the same features to end-users? IPv6 transition is being handled rougly like the 2000 bug. At the last minute people are scrambling to cobble together workarounds. At least old programmers had a good excuse - space was at a premium. It's still a premium and prices for network equipment show...
Who will pay - this is the question... Posted Jan 26, 2011 15:23 UTC (Wed) by bojan (subscriber, #14302) [Link]
The only reason they can disable (or not pick) IPv6 is because they are two different protocols. Which is exactly the problem with this transition plan. They don't need IPv6, because they are not expecting anyone to use it. Because it's useless. Because all hosts are IPv4.
If transition was handled differently, all hosts with current 32 bit addresses (which would also work as 128 bit ones) would also be IPv6 hosts, so having IPv6 on routers would actually be (surprise!) useful.
Who will pay - this is the question... Posted Jan 26, 2011 17:15 UTC (Wed) by khim (subscriber, #9252) [Link] If transition was handled differently, all hosts with current 32 bit addresses (which would also work as 128 bit ones) would also be IPv6 hosts, so having IPv6 on routers would actually be (surprise!) useful. Useful for what? In this plan (the same as with current one) you can still safely disable useless extension on networking router, save half of the money and users will not notice. What incentive will there be for the ISP to support this extension? If Cisco will decide that it's good idea to support IPv6 unconditionally then it'll just lose to Juniper (or some other firm) which will implement "turbo mode" with 32bit addresses only...
Who will pay - this is the question... Posted Jan 26, 2011 19:18 UTC (Wed) by cmccabe (guest, #60281) [Link]
Routers have to go through qualification tests-- sometimes very difficult ones, in third party labs. I'm sure there are a lot of features Cisco would like to drop from IPv4 and IPv6, but guess what: they can't. Not if they want to call their equpiment compatible.
You are focusing on the wrong end of the problem completely. Big equipment vendors love new standards, especially if they're complicated and difficult to implement. It creates churn, which means more purchase orders, which means more money.
The problem is that IPv6, as designed, is as useless as a screen door on a submarine until the magic moment arrives-- the IPv6 rapture, if you will. And ISPs are focused on Q4.
Who will pay - this is the question... Posted Jan 26, 2011 21:51 UTC (Wed) by tialaramex (subscriber, #21167) [Link]
So, your approach is to hope that people will pay extra for something they don't want, because you said so.
Why do you need DJB's jabbering for that? You can just demand they pay extra for IPv6. Call it "Super Internet Plus" if you like. Go lobby your representative. See how much that helps.
The idea that because something is "part of" IP therefore it works on all the network gear people are buying? Just more proof you're completely out of touch with the real world. It is completely _normal_ to have stuff that doesn't work. Most of the time there's a config switch, the default is conservative, everything else is a lucky dip. All the features that deviate from the most basic processing of unicast TCP traffic are in the lucky dip category. You know Intel shipped a whole stepping of i386 that can't run protected mode Windows? That's the whole point of the 386, and it didn't work. Vendors have shipped whole families of products where they know there are features that just don't work. But "Oops, yes that is a bug, we'll let you know if we can fix it in firmware" is a lot better than "No, we can't deliver that feature, we won't bid". Nobody is going to sue - network admins in big corporations (especially big tight-fisted corporations) are used to being disappointed.
Who will pay - this is the question... Posted Jan 26, 2011 22:28 UTC (Wed) by bojan (subscriber, #14302) [Link]
> So, your approach is to hope that people will pay extra for something they don't want, because you said so.
No. For something they want.
I cannot request access to a cool new IPv6 site right now. I'm on IPv4. So, my ISP cannot possibly have request for IPv6 traffic from me. Ergo, IPv6 is useless to them (and me).
Have my stacks been upgraded so that my current IPv4 addressing just worked with IPv6, occasionally I would get an AAAA response to my DNS queries (that is also stupid - this should have been simple A - but that's a different issue altogether), which would have a genuine IPv6 address in it. Now, I would not be able to access this site, because my ISP failed to buy IPv6 capable equipment, although my network was already IPv6 ready without me touching a thing in my config (I'm already on the net).
So, I have a choice:
1. Drop this stupid ISP and get one that does IPv6.
In both cases IPv6 wins by default.
Right now, the onus of IPv6 upgrade is on each and every customer. Each and every customer already connected to the only net we have. For no good purpose whatsoever.
You've lost one more possibility. Posted Jan 26, 2011 23:45 UTC (Wed) by khim (subscriber, #9252) [Link] So, I have a choice: Sorry, but you forgot one more choice:
3. Forget about crazy site which by someone's folly only have IPv6 address. Because 99% users sites choose option number 3 (DJB plan or no DJB plan) there are no need to think about these silly AAAA-only sites. ISPs know this full well. Right now, the onus of IPv6 upgrade is on each and every customer. And this is not a problem at all: either you have IPv6-capable OS like Windows7 (where you only need to connect to the IPv6 internet to use IPv6) or you have something like PS3 or XBox360 where IPv6 does not work because developers just decided to ignore it. In both cases DJB plan is not needed at all. Sure, if you have large organization you'll need to do something, but "IPv6 works by default" approach will not help there at all: a lot of such organizations (most of them?) disable direct access to internet and ask uses to use proxy with authorization - and all that must be changed for IPv6 anyway.
You've lost one more possibility. Posted Jan 27, 2011 3:13 UTC (Thu) by bojan (subscriber, #14302) [Link]
> 3. Forget about crazy site which by someone's folly only have IPv6 address.
Folly? You mean the address exhaustion. Yeah, let's do that. That'll work as well as the current plan. Which is to say not work at all.
> either you have IPv6-capable OS
I have an IPv6 capable OS. I have a relatively new DSL router with very young software on it. I have a valid net address. I have my DNS configured. I have my firewall configured. I've been connected to the net for a few years now using the same address. And yet, I cannot ping ipv6.google.com. That is what common sense people call "interoperability failure." I'm sure you'll have some funny explanation for this, full of acronyms like ASIC, FPGA etc. :-)
No, I meant simple fact... Posted Jan 27, 2011 8:58 UTC (Thu) by khim (subscriber, #9252) [Link] > 3. Forget about crazy site which by someone's folly only have IPv6 address. No, I meant another folly: someone decides to implement IPv6-only resource in an IPv4 world. This is real stupidity and you can safely ignore such people - their resource will be dead very soon anyway. Of course if it's some kind of underground resource and you absolutely positively need to visit it... you'll find a way. Just like people without access to the Internet had a ways to download some files from it (yes, I mean ftpmail and other similar technologies). That'll work as well as the current plan. Which is to say not work at all. It works perfectly. Just like any disruptive technology it starts from the places where IPv4 just does not fit and goes from there. The only problem: IPv4 address are still not scarce enough so there are few such niches. I have an IPv6 capable OS. I have a relatively new DSL router with very young software on it. I have a valid net address. I have my DNS configured. I have my firewall configured. I've been connected to the net for a few years now using the same address. But you don't have an ISP which supports IPv6 - and that is the problem. Everything else is irrelevent and if you'll not have such provider IPv6 will not work. DJB plan or no DJB plan. And yet, I cannot ping ipv6.google.com. That is what common sense people call "interoperability failure." I'm sure you'll have some funny explanation for this, full of acronyms like ASIC, FPGA etc. :-) And you'll insist that somehow it can be solved by the insane DJB plan. You were asked dozen of times: how exactly this plan materialize ISPs with IPv6 routers. You refused to answer. This means one thing and one thing only: you don't know. And if you don't know how this critical part will be worked out with DJB plan then what evidence do you have that it may work?
Who will pay - this is the question... Posted Jan 26, 2011 22:09 UTC (Wed) by bojan (subscriber, #14302) [Link]
> Useful for what?
Routing of packets from hosts.
Bottom line: lots of people have to do unnecessary work for no benefit at all. They are already on the net. Why do they have to connect again? Yeah, it's that simple.
Who will pay - this is the question... Posted Jan 27, 2011 0:01 UTC (Thu) by khim (subscriber, #9252) [Link] > Useful for what? You mean: DJB plan will magically induce ISPs to throw money for no apperent reason? Hard to believe... Bottom line: lots of people have to do unnecessary work for no benefit at all. They are already on the net. Why do they have to connect again? Yeah, it's that simple. Bottom line: there were more CompuServer users 20 years ago then Internet users back then. They all are gone today. 20 years down the road IPv4 user will be similarly extinct. Switch will happen in the same fashion: people will get "poor alternative" because they can not afford "good one" and eventually the "good alternative" will be useless. Prioces for IPv4 will start raising in the coming years, so there are no need to worry: people will upgrade. They are not stupid. But before that happens price of IPv4 address should become high enough to make these 2x hardware prices cheap by comparison. It didn't happen yet: "white IP" is sold for $2-$10 per month today. It's not nearly high enough to induce change. But times are changing.
Who will pay - this is the question... Posted Jan 27, 2011 2:35 UTC (Thu) by bojan (subscriber, #14302) [Link]
> 20 years down the road IPv4 user will be similarly extinct.
That I can believe. The transition plan to it is still shit.
Well, sure. Posted Jan 27, 2011 9:05 UTC (Thu) by khim (subscriber, #9252) [Link] > 20 years down the road IPv4 user will be similarly extinct. Well, sure. I mean: people spent lots of reasources trying to invent some "perfect" transition plan (DJB's way is only one alternative - and it's just as stupid as all other ones), but in the end there are only two ways: But for the market way to work IPv4 needs to be significantly more painful then it's now. Prices for "white IP" should be around $100/month at least, not $2-$10 like they are now. I'm not sure we'll reach this stage any time soon. More likely some ecosystem will adopt IPv6 first and the snowball will go from there. Will this be an LTE or something else? We'll see.
Who will pay - this is the question... Posted Jan 27, 2011 7:26 UTC (Thu) by jem (subscriber, #24231) [Link]
"Price difference between IPv4 and IPv6 is exactly two times."
I don't buy your logic. If IPv6 traffic right now is <1 % of the total traffic and probably won't grow very rapidly any time soon, why does it matter that this year's router model is half as slow forwarding IPv6 packets?
Because today's model is the best available Posted Jan 27, 2011 9:09 UTC (Thu) by khim (subscriber, #9252) [Link] If you enable IPv6 you reduce not only number of packets you can process. You reduce number of routes you can process, etc. And about <1%... this is red herring: if the IPv6 traffic is <1% and you expect to have at on this point for a long time - then why bother (do you really expect significant revenuy from this <1% traffic?), but if you expect that this proportion will grow then IPv6 support means real money. The best alternative is to prepare contingency plans and wait till they will be needed - and this is exactly whay most ISPs are doing.
It just shows how badly even intelligent people may misunderstood the simple problem Posted Jan 26, 2011 12:55 UTC (Wed) by tialaramex (subscriber, #21167) [Link]
"This is done already"
More than that, it's been done several times for people who wanted different things.
At first they say "I just want IPv4 addresses to have an equivalent in IPv6" so that's there, albeit largely useless except for some types of software.
Then it's "Oh, I should automatically get IPv6 addresses if I have IPv4". Did that, it's in 6to4, every IPv4-capable node has a /48 in 2002/16 for it to sub-allocate as it wishes, the IPv4 addressed node acts as router for the subnet, and an anycast address optionally provides a tunnelled route to the entire IPv6 space.
But now we get what people really want, what they really, really want. They want to be able to parley an IPv6 address (which they'll have billions of) to get an IPv4 address so they can continue using the IPv4 Internet after the crunch. This is mathematically impossible, and so, more so even than something which violates a law of physics, just wanting it really badly won't make it possible. But that won't stop them blaming the engineers who "failed" to do it for their woes.
It just shows how badly even intelligent people may misunderstood the simple problem Posted Jan 26, 2011 14:17 UTC (Wed) by bojan (subscriber, #14302) [Link]
What "people" are saying is "I have a perfectly good setup, why do I need to change it?" Answer is: because IPv6 plan sucks.
Yeah, I know it will all work itself out in a few years. But pretending that it wasn't screwed up doesn't make it go away.
It just shows how badly even intelligent people may misunderstood the simple problem Posted Jan 28, 2011 6:54 UTC (Fri) by butlerm (subscriber, #13312) [Link] "This could have been handled by setting aside a special part of the IPv6 address space for addresses that mapped directly on to IPv4 equivalents."This is done already. Not on a routable basis it isn't. If the mapping was routable, the entire IPv4 network numbering plan could be routed through IPv6 only routers. But since the designers did not want to route the large number of prefixes from the existing numbering plan, we instead have a new numbering plan that requires the network configuration of a couple of decades now to be dumped and redone from scratch. A brave new world that no one wants to inhabit. There were proposals on the table sixteen years ago that would have avoided this problem and made the transition process more or less transparent. Look up TP/IX sometime.
What about IPv6 right here on earth? Posted Jan 26, 2011 23:16 UTC (Wed) by job (guest, #670) [Link]
Your premise is wrong. IPv6 is much cheaper to switch in hardware because if its fixed header size. Less complexity leads to more performance at a lower price.
Also since the routing logic is simplified I'd even expect routing tables to drop in size, at least initially, despite the addresses being that much larger. I may be wrong on that, but see above.
What about IPv6 right here on earth? Posted Jan 27, 2011 11:44 UTC (Thu) by jthill (guest, #56558) [Link] But it's either route IPv6 on hardware built to optimize 32-bit-address performance or buy an IPv6-optimized twin for every router they have. Looks like doing the former hurts so bad it makes no difference, the switch to IPv6 will make them double their router investment either way.
What about IPv6 right here on earth? Posted Jan 27, 2011 16:16 UTC (Thu) by job (guest, #670) [Link]
Of course you're working against the market here. Any change will meet economic resistance since the perceived markets starts out very small. But this is a truism, and would be the case independent of whether the change is IPvDJB or IPv6.
My point is that IPv6 is easier and cheaper to route than IPv4, not the other way around.
|
|||||||||||||