
May be a good thing. In that it might now start to force ipv6 adoption.
The American Registry for Internet Numbers (ARIN) can no longer satisfy requests for new IPv4 addresses and has started a waitlist for those who want more. ARIN warned, in early June, that “It is very likely that we are already processing a request that we will be unable to fulfill [sic].” On Monday this week it escalated the …
But this is NO joke. We've gone from stories of IPv4 running out to stories of IPv4 HAVING RUN out, as in there actually ARE empty shelves now, with only scattered items left here and there. And not just in one major part of the world. Asia's been dry for years, but who cares about them? But now it's both Asia AND North America: TWO key world markets. The IPv4 world is basically overcrowded with only two options left: jury-rig it or move to a bigger world. Thing is, moving to IPv6 has so many growing pains few want to go while jury-rigging will only work for so long. There's already complaints about handling carrier-grade NAT; what happens when someone behind a carrier-grade NAT wants to connect to someone else behind another carrier-grade NAT?
Why would anyone use carrier grade NAT, whrn it's so easy to buy large blocks of addresses on the private market?
No operator in yheir right mind would consider adding extra complexity and cost to thrir solutions without looking into buying more addresses first.
5 years sgo isp's were told by RIPE that addresses were out. Only the crappest decided that meant they coukfn't get any more...
I've been ready and prepared for a while. In fact I used to be with an ISP that offered dual stack. I got everything set up - even my email server although it could only send on IPv6 due to licensing restrictions(*). Then I moved from my niche ISP (IDNet) to a bigger ISP (PlusNet). They had an IPv6 trial. It closed to new applicants shortly before I joined and over 18 months later still shows no signs of going live.
(*)I'm a tight wad and couldn't see the point of paying an extra £10 just so I could receive from Google Mail using IPv6.
I tried to give back my reserved IPv4 address to Demon several years ago - but the service I use has a mandatory reserved IPv4 address. Nowadays they only offer "business" service contracts - which always have reserved IP addresses.
They appear to have recently offloaded web hosting - and the mail "push" feature went years ago.
NAT seems to be capable of handling most people's traffic.
IPv6 would have been adopted 10 years ago if the whole thing hadn't become a holy war based around a couple of sticking points e.g. NAT. I don't see why they couldn't have allowed NAT for those that really wanted it and were willing to put up with the downsides. I can't help feeling IPv6 has got bogged down like a software rewrite with the problem of "it must be perfect this time". It's not going to be perfect but never "releasing" is much worse than having a few issues.
As for the benefits of NAT, what benefits are there that a firewall can't do?
Although NAT does at least mean your computer can't be directly addressed at all whereas a firewall means that when it gets attacked it will block it. A rather silly analogy is this:
Well armoured and on a battlefield. (Firewall)
or
Invisible and several kilometres away from the battlefield (NAT)
Now granted a firewall gives added protection over and above just NAT. For instance it could trap anything - even virus infected attachments or spam email if it's well implemented. But would I rely solely on a firewall? Hmmm.
With IPv4 we can have firewall+NAT = two layers of protection.
With IPv6 we have one layer of protection.
"As for the benefits of NAT, what benefits are there that a firewall can't do?"
Privacy.
As originally intended IPv6 would apparently identify each PC uniquely by forming an address from its MAC address. There have been noises about ways to randomise this address so that tracking of a specific user is not possible. Not sure whether that has become the intended standard yet.
@AC Privacy
Honestly, you could have researched a bit before posting here. There is Privacy Extensions for Stateless Address Autoconfiguration in IPv6 proposed standard (from 2001) which next evolved to draft standard (from 2007). These are widely implemented and available both on Windows and Linux (didn't BSD because I'm not your researcher)
yes the randomness is in place but it's optional, who's to say its not possible to decrypt the obfuscated address?
We need IPv7 with IPv6's issues resolved before we see mass adoption of IPv6.
Having recently had my ISP formally enable IPv6 on their network, I was genuinely impressed that all my Windows 7 and Ubuntu systems supported RFC 4941 out of the box. The randomness may be optional from the RFC perspective, but it looks like OS vendors have enabled it by default.
There won't be an IPv7. At this stage, the only thing really preventing widespread adoption is enterprise upgrades, as I know firsthand how many organizations are not prepared for the handling of IPv6 via routers, firewalls, load balancers and VPNs. I would add DNS to that mix, but I think that's more of an issue training the admins than an infrastructure upgrade for most.
@ Preston Munchensonton
its your router dishing out IP addresses not your OS.
no there won't be an ipv7, it'll likely be called ipv10 but we need something beyond ipv6.
enterprise upgrades are not stopping ipv6, they are the ones for whom moving to IPv6 would be easiest as they control more readily ingress/egress and can easily put in ipv4 to ipv6 gateways at their borders. Many enterprises are running dual stack internally right now.
NAT is great for enterprises as they get to hide all their hosts behind a handful of public IP's, IPV6 can expose all their hosts addresses to the net.
There are many many issues with IPv6 that enterprises don't like but they are not holding up adoption.
This post has been deleted by its author
There is nothing wrong with IPv6 and it has been "released" many years ago. Yes it had some improvements over the years, but that's steady evolution which happens to any Internet protocol, not just IPv6.
The problem is with applications/appliances which do not properly support IPv6 (thus forcing users to IPv4), because their respective vendors do not have a clue. For example, finding good SIP phone with IPv6 support is tricky etc.
Hopefully when American ISPs are forced to join 21st century and give IPv6 to users, that will in turn prompt software/hardware vendors to give this protocol more consideration. But I won't hold breath, probably a whole generation of software/hardware/network engineers has to die out first. Still, there is a good chance this will happen before end of 21st century ...
The only fundamental issue with IPv6 is that humans are scared of it. They are used to easy to say and remember 4 octet addresses. And they dont like change.
ipv6 configures the same way in routers as ipv4, easier in places thanks to link-locals, and has far better rsnges predefined. I haven't seen anything wrong with ipv6 yet. Only with application layers and humans that haven't adapted to it.
We could crank the version number up by a couple of notches to make it sound really new, and we could fix it so that instead of using 32 bit addresses, we could use 64 bit addresses. No wait...! How about 128 bit addresses? That ought to keep us going for the next thousand years or so...
"That's almost as long as the stories about IPv4 addresses running out."
The press reports often seem to concentrate on the raw IPv4 addresses - rather than how they are being used to support a much larger population of users.
The designated "local" address blocks and Network Address Translation (NAT) are used at several stages in the path between a device and the internet. That allows users and ISPs to multiplex many users' devices onto their available IPv4 addresses.
At the web host farm several customer's web sites can share an internet facing IPv4 address. Each site's requests are differentiated by information other than the external dedicated IPv4 address on which it arrives.
"At the web host farm several customer's web sites can share an internet facing IPv4 address. Each site's requests are differentiated by information other than the external dedicated IPv4 address on which it arrives."
Which then kinda falls apart when they get a request that contains ONLY an IPv4 address. Some protocols are like that.
"At the web host farm several customer's web sites can share an internet facing IPv4 address. Each site's requests are differentiated by information other than the external dedicated IPv4 address on which it arrives."
Sadly this stops working if you want to use HTTPS. Which is increasingly required for new web technologies.
>Sadly this stops working if you want to use HTTPS. Which is increasingly required for new web technologies.
Works just fine (server side) if you're using F5 Big-IP to front the websites and certificates based on domains... However, there are some issues client side as when they do a lookup for a certificate based on IP address...
That's too optimistic.
There is (almost) no market for IPv4 addresses and these cannot be freely traded (yes, there is some trade in IPv4 addresses but it is very limited). Also, once a vendor owns a range of IPv4 addresses, there is no ongoing cost of using it and there is no serious proposal to change it (although I would support such a change - after all, I pay to my ISP monthly).
What needs to (and will) happen is increased deployment cost of IPv4, due to increasing network complexity forced by handling of various (static or dynamic) NATs at network stages where previously they were not used. But it will be very slow process.
where do you get your facts from?
http://www.bbc.co.uk/news/technology-32826353
ISP Networks are layered by design with loads of expensive kit already balancing, proxying, billing, inspecting and securing traffic flows, if they needed to NAT its likely they have the kit and expertise already. I doubt so called carrier grade NAT will ever be deployed as its not needed especially with companies handing back addresses they don't need.
"companies handing back addresses they don't need."
Exactly. I heard a while back that some of the bigger corps and uni's have ipv4 blocks reserved that are never used. Stanford was on the list, MIT, Apple and of course M$... I may have even read here on El Reg.
Ever been to Asia? You have a large population combined with a comparatively small IPv4 allocation. Carrier-grade NAT is already deployed there, creating instances of double-NATting (a NAT on top of a NAT). Let's just say double-NATting can make using various protocols pretty ugly.
I can't help but wonder if one of the big reasons people have been avoiding adopting IPv6 is because it's such a big jump from IPv4 and 'looks scary'. IPv6 addresses for example look alien to people who are familiar with (but don't really *understand*) IPv4 addresses (think PHBs). Perhaps if there'd been an IPv4-XTRA as a stepping stone with addresses like FFF.FFF.FFF.FFF it would have been less of a culture shock and easier to swallow.
Maybe that would never have been workable, and maybe looking too different isn't really a factor at all.
I don't see how letters would make too much of a difference, as we're using to seeing letters on our license plates and some places use letters in their postal codes. They even try to be accommodating by creating shortcuts when the quartet is 0000 (the :: shortcut). I personally see a max of eight quartets easier than trying to memorize up to 16 different numbers.
now start looking through routing tables or firewall rules thousands of lines long and see if you can spot a problem. Heck, this is difficult even with IPv4 and usually those network guys have Windows backgrounds so awk and grep are not among their favorites.
This post has been deleted by its author
No, you don't lose. You dual stack your network so that your old hardware can still talk to what it needs to. But you'll need to provide an IPv6 addressable interface device for IPv6-only systems that need to access data on your old IPv4-only stuff. In the meantime you call the person who recommended you buy IPv4-only stuff and bollock them for not being farsighted enough to advise future-protecting your investment with IPv6 support.
"In the meantime you call the person who recommended you buy IPv4-only stuff and bollock them for not being farsighted enough to advise future-protecting your investment with IPv6 support."
And if the reply is that the device is 20 years old? IOW, deployed BEFORE the concept of IPv6 even existed? Can't really anticipate what's beyond the realm of possibility at the time.
Also, you have to wonder about devices that can't be replaced or upgraded BUT still need to be able to talk to IPv6 clients down the road.
why not run IPv4 inside and IPv6 outside on the internet concentrating your efforts on dual stacking or translating between the two at the border ? By doing this you could avoid impacting your mission critical systems.
Large and medium enterprises are aware that there is a problem with IP addresses on the Internet buy why does this have to impact their internal networks ? Here lies the problem, they just don't want and don't see why do they have to touch their internal network + applications infrastructure for something that happens outside their network.
If we want fast IPv6 adoption, somebody should come up with a IPv4/IPv6 translation gateway and large enterprises will love it.
Note that concept of large enterprise mentioned here does not include emerging or established pure Internet companies like Facebook, Twitter, Google etc. They are rich enough to figure out a way out of this.
@ac said: "why not run IPv4 inside and IPv6 outside on the internet concentrating your efforts on dual stacking or translating between the two at the border ? By doing this you could avoid impacting your mission critical systems."
It's an idea that's been had before. The devil is in the details. A bit of Googling on the topic will give you an idea of what's possible and the work that people have done to make it happen.
Something I think we should all remember: IPv4 only nodes will continue to be reachable as long as they remain up, even after the world has mostly adopted IPv6.
While IPv6 is meant to be a successor to IPv4, adoption of IPv6 DOES NOT require disabling IPv4. If you have IPv4-only nodes on your network, you may configure some of your IPv6 nodes (usually routers) to perform the task of transferring traffic between the IPv4 network and the IPv6 network. I suspect that this is the arrangement that will be in use for the foreseeable future. IPv6 is here, and it is not that much of a skeleton transplant to get it going.
"Something I think we should all remember: IPv4 only nodes will continue to be reachable as long as they remain up, even after the world has mostly adopted IPv6."
IPv6 reaching IPv4 has never been the problem. That's why there's a block set aside for the purpose. It's what happens when IPv4 has to reach IPv6 and the hardware's too old to be able to learn accommodation techniques yet are not in a position to be replaced anytime soon. Particularly if it's the IPv4 side that has to initiate the conversation.
"IPv6 reaching IPv4 has never been the problem. That's why there's a block set aside for the purpose. It's what happens when IPv4 has to reach IPv6 and the hardware's too old to be able to learn accommodation techniques yet are not in a position to be replaced anytime soon. Particularly if it's the IPv4 side that has to initiate the conversation."
If you are currently using a computer running IPv4-only, you are demonstrating that there is no problem the other way, either. The Internet backbone has long ago gone IPv6, and thereby demonstrate support for IPv4-to-IPv6, IPv6-to IPv4, IPv4-to-IPv4, and of course, IPv6 to IPv6. It really isn't that much of a problem.
@1980s_coder: Agree with you 1000%. The only point I'd disagree with you on is the RFCs--those really are very unforgiving for your typical ordinary sysadmin-person. Wikipedia is a good starting point, and several books (Draytek has a very practical one published called "Real world IPv6" or something). People should seriously get moving; it's different, perhaps a little steep, but once you've learned it and understand it you'll be wondering what all the hate and fuss is about. Come on commentards, you can do it! :)
And my sincere apologies to those net engineers living in parts of the world who have already made the leap and are being inconvenienced by those of us in affluent regions who seem to think that neoliberalism solves every problem that needs solving. Those of us with a clue are trying to spread the faith as fast as we can before real damage is done, honest.
Every time we hear these doom stories about IPv4 addresses running out. And although I don't question the fact that there is only a limited amount of addresses available, I can't help wonder... Usually all we're reading is stories about usage. Like this one: now there are only approx. 130,000 IPv4 addresses available for use.
So what happens if someone cancels his subscription, stops hosting his server, or when an ISP applies a change in their subscription scheme where there's a difference between getting a public or private IP address?
I know, I know: in most cases nothing changes (the ISP & data center keeps custody over its address blocks, so does the ISP) but there are also plenty of scenario's where addresses do get freed, up to a point that they become available for public use again.
But I think it's a bit strange that you hardly read anything about that process.
ISPs would request a block of IPv4 addresses. So an ISP that requested and got a /16 block last week probably still has 60,000 addresses available and uses them up with new customers. Reusing them is probably as bad as reusing phone numbers. These 60,000 addresses cannot be used by anyone else but that ISP. So even when these 130,000 IPv4 addresses are gone (is that 2 /16 blocks? ) most ISPs are still good for a while. Until someone runs out and can't get another block, then that ISP has a problem. Which can be solved by using IPv6.
The rather large company I used to work for until very recently has at least 3 or 4 class B ranges they aren't using that would keep ARIN going for another day or so. Plus a couple RIPE could get back.
We won't discuss the multitude of </16 ranges they have scattered about the globe.
I think they could keep a few of the /24 for NAT/other purposes and get along on the 1918 ranges until they convert to IPV6.
If only they were actually converting to IPV6 instead of yet another committee to study the idea. Maybe next decade.
I'm not sure it is 'spare', as such. More likely, they have designed their network around having a /8. I guess they could renumber everything to a private range and sell it off. However, I don't see many of the firms that have /8s jumping to cash in, so it's probably not as easy or as profitable as it might first appear.
Knowing DWP, they'll probably make less on the sale as they spend on consultants.
https://governmenttechnology.blog.gov.uk/2015/02/19/freeing-up-unused-ip-addresses/
Why sell it now, when it's only going to get more valuable?
That didn't stop Gordon Brown selling off our gold.
1) No NAT. The religious abhorrence of it is just silly. Having it doesn't break the internet, so why not?
2) Too big. The addresses are untypable, and the amount of memory required in a router is now insane.
Oh, and most of the available kit is IPV6-borken, either by accident (ISP routers handing out duff IPV6 addresses over wifi, preventing connectivity) or design (preferring IPV6 in an IPV4 network)
"1) No NAT. The religious abhorrence of it is just silly. Having it doesn't break the internet, so why not?"
Have you tried to negotiate a connection between two endpoints BOTH behind NATs (particularly carrier-grade NATs beyond endpoint control)? Let's just say it's not easy without a third party that may not be available or trustworthy.
"2) Too big. The addresses are untypable, and the amount of memory required in a router is now insane."
People use letters in license plates and postal codes (and letters beyond F, too). This is nothing new. Which would you rather have: 8 quartets or 16 numbers? As for it being too big, the idea is to avoid having to go through the same exercise again down the road. 128 bits should if pressed suffice to connect every last thing in the known universe (which is why ZFS uses that bit count for most of its limits).
IPv6 is not "too big", you "only" get space for some 2^48 public networks. The remaining bits are for internal networks (2^16) and then for individual hosts (2^64). It is a lot yes, but it gives good amount of space for a well organized public network structure with (hopefully) much smaller BGP tables.
"As for it being too big, the idea is to avoid having to go through the same exercise again down the road. 128 bits should if pressed suffice to connect every last thing in the known universe"
Not up on these things for a while (naughty me) but didn't I see that subscribers are dished out /48s or /32s of IPv6? Might not be as unlimited* as people think?
* fair usage applies
"Not up on these things for a while (naughty me) but didn't I see that subscribers are dished out /48s or /32s of IPv6? Might not be as unlimited* as people think?"
But they can always be adjusted as time passes. We can't change the fact IPv4, being fundamentally 32-bit is limited to around 4 billion entries total (not accounting for some specialized verboten ranges). The human population combined with multiple devices per person, many of which WILL need to be directly addressable, will eventually overwhelm the range.
"Not up on these things for a while (naughty me) but didn't I see that subscribers are dished out /48s or /32s of IPv6? Might not be as unlimited* as people think?"
I don't know what you are talking about.
Most enterprises get a /48, which allows them to create up to 65536 subnets with 64-bit node addresses; Most ISP's get a /32, which allows them to allocate up to 65536 customers.
Most individuals and home subscribers get a /56, which allows them to create up to 256 local subnetworks, each subnetwork having nodes with 64-bit addresses. Address space might not be "unlimited", but most subscribers will find what they get is definitely sufficient.
As normally deployed, the 128-bit address space is actually only 64-bits public. If your ISP gives you a /56, they're giving you 256 networks to play with. That's more than the 1 you need, but it is a far smaller chunk of the IPv6 unicast space than even a single IPv4 address is of that space, so calling it "wasteful" is harsh.
Also, and partly in response to another poster's remarks about routing tables, one of the reasons why they went for a 128-bit space was so that they had bits to waste. The idea is that quite a few routing decisions can be made without complex tables at all, just by inspecting prefixes. Ironically, the protocol that now needs massive tables is IPv4, which owing to the balkanisation of the address space now requires brute-force-sized routing tables in some places. There was an El Reg article about this a few months back, but I can't remember enough of the details to google it.
So you're saying that the public part of IPv6 is intended to greatly simplify routing by making say the first x bits be hard-routed, say, geographically to a few levels so that tables only have to come play later on and be of a more-manageable size since the packet's been partially pre-sorted already.
"Can you please explain why my ISP has to give me IP addresses for my internal devices when I only need 1 or maybe 2 public addresses that could cover all my needs ? Isn't this a waste ?"
Your ISP doesn't "have to". If you or your ISP can show that you will never need more than one subnetwork, your ISP is fully prepared to hand you a /64, which will allow you to have a single network whose nodes may take any 64-bit IPv6 address having the same prefix as the one the ISP handed you. If you and your ISP agree that on a given link that there will be no need for more than the single nodes at each end of the link, the ISP is fully prepared to assign a /127 to the link. But those two cases are by far, the less common cases.
And as another commentator replied, that arrangement actually _simplifies_ the routing computation, and in particular, reduces the space requirement.
(WHOOSH!)
We remember alphanumeric combinations more complicated than hexadecimal (because they use the entire alphabet rather than just the first six letters) on a regular basis in license plates, postal codes, even some telephone numbers that employ the telephone letter system.
Here in Germany, a major provider (Unitymedia) has started using DSLite/Carrier grade NAT for all new contracts, I reckon many providers will do the same to reduce costs and in the end out of basic necessity
If IPv6 is not widely adopted this means that p2p communication between regular users has to go through hops. Things like Bittorrent but also peer to peer voice/video calling will not work any more, making the grip of MS Azure, Google, AWS, Netflix, etc even stronger, since distributed services are not even possible and everything will require a strong middleman data center.
After growing up in the 90s, I find it sad to see that the internet is seen as a content delivery network for video on demand and thinly veiled advertisement services - sure those things are part of it, but the underlying technology can empower so much more.
I am however looking forward to the wide adoption of IPv6, it will mean a whole new era with regards to possibility, privacy and security - especially with IoT finally getting traction. Hopefully economic reality of expensive IPv4 will push adoption more quickly.
Thanks for your time :)
But ARIN is run by idiots just line other groups and given they NEVER go back and force network numbers users to PROVE they're using them, they'll never get these back.
If I know where 3 Class B's are, and a few dozen CIDR blocks (as I applied, and received them back in the 80's/90's) you can damn sure bet there's a fuckload more out there.
"My company has enough public IP4 addresses to satisfy demand now and for the foreseeable distant future."
It is not whether your company will need IPv6 to connect to their internal devices, or to their existing customers. Presumably your company is doing well and accumulating new customers. What are you going to do when (not if) you gain a new customer who was never able to obtain IPv4 addresses?
What would you do if you choose to adopt a VoIP provider who, in order to assure quality of service and low connection latency, uses native IPv6, rather than Carrier Grade NAT on IPv4?
IPv6 is a part of your future, and you probably can't avoid that.
>What would you do if you choose to adopt a VoIP provider who, in order to assure quality of service and low connection latency, uses native IPv6, rather than Carrier Grade NAT on IPv4?
'Voice' is an application, if the provider can't handle the carriage of end-to-end voice communications over a mixed IPv4/IPv6 infrastructure then I suspect neither can they support the end-to-end carriage of voice between a VoIP phone and a fixed phone or mobile phone...
"'Voice' is an application, if the provider can't handle the carriage of end-to-end voice communications over a mixed IPv4/IPv6 infrastructure then I suspect neither can they support the end-to-end carriage of voice between a VoIP phone and a fixed phone or mobile phone."
Oh, I wasn't talking about carrying VoIP over mixed IPv4/IPv6 infrastructure. I (though I did not make that clear; sorry) was referring to the situation where a given IPv4 customer was not able to get his IPv4 addresses via any other means than by Carrier Grade NAT. In that situation (and I can speak here from experience trying to deploy real-time communications to traverse NAT), there is significant, perceptible degradation of quality-of-service compared to native IPv4, or native IPv6. Even in the case of having to deploy in a 6-to-4 scenario there was better quality of service once you were able to otherwise keep NAT out of the picture.
It doesn't take a honey pot to detect IP ranges that should be pulled and redistributed.
IPv6 isn't more secure and anyone who thinks that has an MS tag to their shirt with a few $$.
This Pro-IPv6 article says it all http://www.ipv6now.com.au/primers/IPv6Myths.php
END OF LINE
But the unused addresses are still "owned" by their holders. You usually can't just take them arbitrarily without some kind of legal precedent (and in most places such a precedent doesn't exist).
IPv6 may be no more or less secure than IPv4 but it can be potentially a lot easier to manage, especially at the top-level topologies. Think about it. A better-structured IPv6 address topology means less need for huge routing tables.
Plus consider this scenario. What if you're only on IPv4 and your client or whatever is only on IPv6 and can't get an IPv4 if he/she wanted to?
I don’t believe this for a second. This A simple tweak of how the system operates can be done to create endless IP addresses. All IP addresses right now are numbers and dots. All one needs to do is insert a letter into only ONE of those places, ie [24.113.158.191] changes to [24.113.158.19r] or add onto the current formula for endless combinations.
Conglomerations have gotten by with stealing so much from the consumer that they now believe we will swallow any lie they feed us. I for one have long wished the Internet had stayed buried. I miss face to face interaction with intelligent individuals like myself. I detest the dumbing down that has occurred to our children, and especially to those who used to have common sense before Bill Gates made computers affordable, and crooks designed another way to scam the susceptible.
Just a reminder: If you read it on the Internet and have not seen it on the news it is probably a LIE.
Amazing how may fish swallow what they hear and see hook line and sinker. I know exactly how Mycroft Holmes feels when he said ‘I live in a world of goldfish’.
That may actually be a better idea than IPv6, though the problem, as with IPv6, is getting all the hardware manufacturers to add support for this. The IP address is a 32 digit binary number, so while the extra binary digits could be represented as a letter, that might not be the best way to do it.
"I don’t believe this for a second. This A simple tweak of how the system operates can be done to create endless IP addresses."
Not if the system on which the IPv4 system is operating assumes (and it was designed this way) that the digits you read are simply representative of a 32-bit binary code (each number only goes up to 255—8 1's in binary). It's not going to be as simple as tacking on a letter or another number to the system. If that were true, IPv6 (which tacks 12 more numbers to make sure we don't run out again) would've been up already. No, the real problem is getting there: particularly the painful dual-stack migration headaches. For example, how does an IPv4-only location establish a connection to an IPv6-only site when the only tools it has available are IPv4-based? It can't learn new protocols so anything that can bridge the gap has to use the existing protocols under existing conditions: extremely limiting. And you can't just tell people to change their hardware because there are those who can't, usually because the hardware's too difficult to replace (it may be expensive, a custom job, or from a firm who no longer exists and has no obvious replacement).
It's exactly the same with memory on computers. Hardware manufacturers have been lying to us for decades because so-called memory 'addresses' are numeric. Just replace some of the digits with letters and hey presto! More memory for no more money!
AC because now I've let the secret out, they'll be after me.
I wonder when we'll start to see companies, that have class A networks, listing the value of their class A networks on their balance sheets.
I also gotta wonder why a company that sells printers, computers, servers, enterprise services, and networking equipment is incapable of converting over to IPv6, thus permitting it to sell its two class A networks.
In 2011, the value of this unnamed company's two class A networks was placed at $378M. Surely their value has risen since then. That'd certainly make some fine executive bonuses, particularly for a company that is soon to double the number of its executives ... :(
Time's running out, though, as more folks switch over to IPv6...
The problem is that until everyone else converts over to IPv6, there's no point, because there are no benefits from doing so.
One thing Google, Nextflix et al could do is make Ultra HD video available only on IPv6. Then people would have a reason to switch, and ISPs could sell it as a premium service.
"One thing Google, Nextflix et al could do is make Ultra HD video available only on IPv6. Then people would have a reason to switch, and ISPs could sell it as a premium service."
I'm no longer much of a believer in restricting a service to one network to compel adoption of a new network. It is the business of a network to afford connectivity, and things should "just work". By that, I mean that services like Netflix should continue to work without the user needing to think of, or even to know about, which network he uses to gain access to the service. A sort of 'principle of minimum disturbance' if I may.
It's true to argue that you don't get much adoption if there is nothing new or unique about connecting via IPv6, but I would answer that by saying that the potential for accommodating new services and new customers with IPv6 (especially multicast-based services) so greatly exceeds what is possible with IPv4, that that could be persuasive enough all by itself.
"It's true to argue that you don't get much adoption if there is nothing new or unique about connecting via IPv6, but I would answer that by saying that the potential for accommodating new services and new customers with IPv6 (especially multicast-based services) so greatly exceeds what is possible with IPv4, that that could be persuasive enough all by itself."
Except that kind of talk doesn't sink in with the laity. You gotta be able to sell the stuff in simple "buy or die" English. Otherwise, your spiel will just go WHOOSH! over a mob of glass-eyed slack-jaws. Not only that, the usual lay pulls won't work here. Porn? IPv4's got plenty of that. Same with the cat videos. So if you're going to generate the critical mass needed to get IPv6 up to full speed, the question becomes: how do you convince the masses, which will in turn push for IPv6 migration upstream in order to stay connected to them?
"Except that kind of talk doesn't sink in with the laity. You gotta be able to sell the stuff in simple "buy or die" English. Otherwise, your spiel will just go WHOOSH! over a mob of glass-eyed slack-jaws."
Aha! Fair enough. But that is the classic "If we build it, will they come?" argument. IPv6 has already been built. Will they come? Well, based on the uptake statistics, you may be amazed that the uptake has exceeded all your wildest expectations. Or you may be driven to despair. I suppose it depends on where you started from.
But let's turn the argument around somewhat. If they come, will we build it? In the territories (like Asia) where it is difficult enough to come by a comfortable-sized IPv4 address block, people are indeed adopting IPv6, but in many cases, they find themselves needing to employ means like tunneling to obtain IPv6 transit service. Why should that continue? Why shouldn't all the local ISP's offer native IPv6 transit service? The upshot to that deficit of native service is that customers (or potential customers at any rate) might give up in frustration.
I accept that "my spiel" must come out in "BUY OR DIE" simple language, but really, is it true that everyone is so dim that they do not understand opportunities for new services? That they do not understand NEW CUSTOMERS? I posit that the persons who need simpler language than that are not the persons we need to try to influence. Those are most likely the persons who will go along with whatever is already available, and not think too much about how it's delivered. But seriously, I don't know what to do with respect to that.
The average person doesn't know what "multicast-based services" means.
The most common use-case of multicast is live video streaming to lots of people at the same time, eg internet TV or video conferencing. But you can do that over IPv4, and lots of people already do.
If you are to persuade people to upgrade to a new technology, then you have to demonstrate that it can do something they can't do already, or it is better in some way, eg faster, more reliable or whatever. "Upgrade to IPv6 to get Ultra-HD video" is a simple, easy to understand proposition. They already understand for example that Netflix isn't going to work on a dial-up connection, they need something a bit better, and you need the best connection for the best video quality. Of course there are many people who are perfectly happy with normal HD, or even Standard Definition video, but it is something to run with. At the moment we have nothing.
"The most common use-case of multicast is live video streaming to lots of people at the same time, eg internet TV or video conferencing. But you can do that over IPv4, and lots of people already do."
and:
"Of course there are many people who are perfectly happy with normal HD, or even Standard Definition video, but it is something to run with. At the moment we have nothing."
Yeah, but if the world sticks with IPv4 we will have nothing. One of the biggest advantages is the potential to accommodate a collection of new users and services that FAR EXCEED the current size of the Internet. I don't believe that is something to sneeze at. We may find that even in the case of multicast applications (which I gave as just an example), that some applications we currently deploy as unicast, might be more reliably scalable as multicast applications when we start talking hundreds or thousands of times more connecting parties than we have at present, but such scaling is precluded owing to a lack of address space.
But in the end, IPv6 is (in my opinion) something akin to the LASER. Initially, nobody has any idea of what is possible.
Sure, but you need to persuade me why I should upgrade now rather than wait until these unspecified new services are available. If I'm going to spend money and effort on it, there needs to be a tangible benefit that I get in return. That's why I'm proposing a new thing that is in the pipeline be available only on IPv6.
"Sure, but you need to persuade me why I should upgrade now rather than wait until these unspecified new services are available."
The move toward usage of IPv6 is not so much an upgrade as it is a transition. It was never meant to be an _upgrade_ per se. It was meant to overcome several limitations that became apparent even when less than ten percent of the world's human population was connected to the Internet. It was meant to address the question of what should be done with IP when the human population greatly exceeds the 4.2 billion IPv4 addresses. It was meant to address the question of what should be done when each of those human beings desires to walk around with several internet connected devices. It was also meant to enable and support the return to the end-to-end interaction and security model that was the original vision of the global Internet.
As I said in another comment, networks exist to afford connectivity. The value of that connectivity grows with the second power of the number of connected entities. More likely than not, I would never be able to make an argument that could convince you to move immediately to IPv6, but I honestly don't believe I need to. You may begin to use IPv6 any time you like, and you should do so only when you perceive the value of connectivity to the global Internet via IPv6 exceeds that for IPv4. I could never imagine what is in the pipeline that is only available on IPv6 that will convince you to move right away.
I do believe that eventually you will, though, simply because the transition is so easy; simply because you more likely than not, do not need to replace any equipment; simply because when you do decide to make the transition, the environment to support that transition will already be in place. IPv6 does not require the Internet to "take a holiday" to make the transition.
"I do believe that eventually you will, though, simply because the transition is so easy; simply because you more likely than not, do not need to replace any equipment; simply because when you do decide to make the transition, the environment to support that transition will already be in place. IPv6 does not require the Internet to "take a holiday" to make the transition."
That assumption is part of the problem. Reality doesn't hold up to this, as there really ARE plenty of hardware fixed to IPv4 and incapable of being upgraded to IPv6. In addition, a small but significant portion of these "stuck" devices serve linchpin roles that make them difficult to replace. How do you replace such a device when there's no budget for it, when the hardware's so customized that replacing it would be a project, not a chore, or if the only possible source for the device no longer exists?
Then you have the IT people working behind the scenes, the ones who have to work the nitty-gritty of the network: especially when things go wrong. These people need to be able to talk low-level, and in terms of low-level, IPv4 was at least within reach for most: four numbers no higher than 255. Now, what if you have to work on IPv6 at a low level and you now have a complicated address with more than 4 non-zero words? And as others have noted, some networks shouldn't be directly-addressable, not trusting in the filtering capability of the firewall (which they feel can be bypassed), which means that aspect of IPv6 is a liability.
"That assumption is part of the problem. Reality doesn't hold up to this, as there really ARE plenty of hardware fixed to IPv4 and incapable of being upgraded to IPv6."
Yes, indeed, it is not trivial to make the transition to IPv6, but it isn't as complicated as a skeleton transplant either. I really should not have tried to make it sound like it is trivial. I apologize.
That said, IPv6 was meant to coexist indefinitely with IPv4. Various 6-to-4 mechanisms have been provided to ensure that the initial islands of IPv6 can interoperate with the ocean of IPv4. Eventually, when adoption of IPv6 becomes widespread, there will still be (probably very large) islands of IPv4, and those same 6-to-4 mechanisms will be what allows them (the IPv4 nodes) to remain online. Thus I think my main point still stands: you do not need to take down your IPv4 networks to build out your IPv6 network.
I actually think what you say in your last paragraph, Charles 9, is full of misconceptions. I will answer that in another reply to your post.
"That said, IPv6 was meant to coexist indefinitely with IPv4. Various 6-to-4 mechanisms have been provided to ensure that the initial islands of IPv6 can interoperate with the ocean of IPv4. Eventually, when adoption of IPv6 becomes widespread, there will still be (probably very large) islands of IPv4, and those same 6-to-4 mechanisms will be what allows them (the IPv4 nodes) to remain online. Thus I think my main point still stands: you do not need to take down your IPv4 networks to build out your IPv6 network."
Don't think 6-to-4. Think 4-to-6 (as in what if it's the IPv4 device that has to connect to an IPv6 device, not the other way), using only existing IPv4 protocols.
"Don't think 6-to-4. Think 4-to-6 (as in what if it's the IPv4 device that has to connect to an IPv6 device, not the other way), using only existing IPv4 protocols."
That's just it: 6-to-4 provides for bidirectional communications, initiated from either side. The same mechanism works in both cases, and only the IPv6 networks need to be adjusted to accommodate the IPv4 nodes.
Not if there's a fundamental disconnect between them: as in at least one end of the conversation can't understand IPv6 at all. And old IPv4 hardware is likely to be IPv6-unaware. Meaning it can't see an IPv6-only node unassisted.
https://docs.oracle.com/cd/E19683-01/817-0573/6mgc65bd2/index.html
(Note the X's between IPv4-only stuff and IPv6-only stuff)
"And old IPv4 hardware is likely to be IPv6-unaware."
Well there is much new hardware that is also IPv6 unaware: many consumer-grade DSL routers, current generation IoT devices etc.
Additionally, I would be cautious of assuming that an early implementation of IPv6 will work nicely with whatever finally gets deployed in the coming "year of IPv6".
"Then you have the IT people working behind the scenes, the ones who have to work the nitty-gritty of the network: especially when things go wrong. These people need to be able to talk low-level, and in terms of low-level, IPv4 was at least within reach for most: four numbers no higher than 255."
Not at all. That is just pure nonsense. In what way is a 16 octet string (the IPv6 address) more complicated than a 4 octet string (the IPv4 address), other than being longer? I work with troubleshooting networks every day, and besides rarely needing to go low-level and dealing directly with the bit-level data on the wire, or dealing with its hexadecimal equivalent, I am usually able to trace a connectivity or performance problem without needing to go so low-level. Are you honestly going to assert that it is prohibitively more difficult to deal with a 16-octet string represented with quads of hexadecimal digits than it is to deal with a 4-octet string represented with triples of decimal digits? I actually find it easier to convert hex to binary and back than I find to convert between decimal and binary. Things might be different for you. But if you are honestly making that assertion, aren't you just admitting that some workers don't like to consult manuals and other documentation? In what way is that a problem of IPv6?
Another idea implied in your statement quoted above, is that the IPv6 header is more complicated than the IPv4 header, and that the IPv6 datagram is more complicated than the IPv4 datagram. Again that is not completely true. The IPv6 header is actually _simpler_ than the IPv4 header, because several of the options and flags that would have appeared in the IPv4 header do not exist in the IPv6 header. If options are needed, they may be placed in IPv6 extension headers. True, IPv6 defines several different extension header types, but IPv6 defines them to have simple structure, and defines simple rules for extension header processing. You process the IPv6 headers in the order that they appear. Thus IPv6 datagram processing has definite semantics, based on the headers appearing in the datagram. IPv4 header processing is definitely not that straightforward.
"And as others have noted, some networks shouldn't be directly-addressable, not trusting in the filtering capability of the firewall (which they feel can be bypassed), which means that aspect of IPv6 is a liability."
To the extent that a statement like "some networks shouldn't be directly-addressable" has meaning, it reflects a serious misconception. If by "directly-addressable" you mean routeable, the only network that really fits that description is an isolated network. The instant you enable transit of datagrams from that network to topologically exterior networks, whether via NAT, or simple forwarding, that description breaks down. A network is either "directly-addressable", or it is unreachable. That is, datagrams addressed to nodes on that network, will reach that network intact, or it will reach that network modified, or it will not reach at all. If it does not reach, then the address used as the destination was not valid; no need for further consideration. If it reaches intact, then the network was directly addressed, and the datagram was not changed in flight. If the datagram reaches its destination modified, you are now left to decide on the trustworthiness of the datagram and its contents. I don't know about you, but I like to have my data come intact.
Net 10 IP addresses (private IP addresses) is ultimately just a convention about router filtering rules that says that datagrams with one of the private IP addresses as a destination MUST NOT be forwarded onto the global Internet. Besides the fact that IPv6 does indeed specify such address types (globally unique local addresses), there is nothing set in silicon that makes that necessarily so. In fact, there is indeed leakage of such IPv4 datagrams onto the global Internet. This usually arises from misconfigured border routers. The only advantage of Net 10 IP addressing is that sites using that can maintain their internal addressing without considering changes from their upstream transit providers. IPv6 makes provisions for that also.
"I also gotta wonder why a company that sells printers, computers, servers, enterprise services, and networking equipment is incapable of converting over to IPv6, thus permitting it to sell its two class A networks."
Anybody who has publicly routeable IPv4 addresses will probably hold on to them. At least until adoption of IPv6 reaches maybe the seven nines percentage range globally. The primary reason for that as I see it, is that the easiest way to maintain reachability between IPv4-only networks, and IPv6 networks is to maintain some number of dual-stack hosts where the IPv4 address of that host is one of the globally valid IPv4 addresses in the pool you have. And that's just it. You keep the IPv4 addresses in a pool and you allocate from that pool when an IPv6 node needs to communicate with an IPv4-only node, and then release the allocation when it is no longer needed. Initially, configuration may be done manually, but as things progress, you should see movement towards automatically maintained address allocations and pools, and then towards globally shared pools once IPv6 adoption becomes widespread enough. That is when I suspect you would see current holders of IPv4 address allocations releasing them back to organisations like ICANN.
If that ever happens, one possible benefit would be a much more coherent arrangement for IPv4 routing infrastructure.