Slashdot Mirror
Level of IPv6 Usage Is Vanishingly Small
An anonymous reader writes "The impending IPv4 address allocation shortage has led to a lot of speculation on the future of IPv6 (including here). A new study says that Internet IPv6 migration is not just going slowly — it has basically not even begun. After spending a year measuring IPv6 traffic across 87 ISPs around the world, the study concludes 'less than one hundredth of 1% of Internet traffic is IPv6... equivalent to the allowed parts of contaminants in drinking water.'"
Because it impacts the other guys, not me. It's the people in China and India and everywhere else that need addresses. Me? I've got a whole block right here.
'less than one hundredth of 1% of Internet traffic is IPv6... equivalent to the allowed parts of contaminants in drinking water.'
Like that means anything to me. Can they compare that percentage in terms of the number of pages per Library of Congress?
If people could actually get IPv6 service from their providers instead of having to route everything through congested tunnels, THAT would help.
The biggest reasons:
And probably many others. The bottom line is that right now today, there isn't a 'killer app' for IPv6.
My blog
What's the downside to being ready?
Because it's work. Work takes time. Time is money.
A certain product at a certain company (forgive my being vague, you know how these things are) has a network interface. This interface is currently IPv4 only, no IPv6 support. When anybody asks the design team why not, they say that no customers have asked for it. Somebody suggested that IPv6 was the sort of thing you want to support ahead of need, but these guys have a lot of deadlines to meet and not enough resources to meet them. They aren't about to spend time implementing features nobody's asked for.
Of course, the time will come when their customers realize they've put off changing over to IPv6 much too long, and will start crash programs to make it happen. They'll demand that this product start supporting IPv6 immediately, if not sooner. So the design team will begin their own crash program, and IPv6 support will be added to the product in a hurry. The implementation will probably cost more and be less robust (at least initially) than if they'd planned ahead.
But they have no incentive to plan ahead. It's a common pattern.
Why is everyone so eager to use NAT? I've never quite understood this, once NAT use became widespread things became a lot more problematic, in my first year of college all the workstations in the computer labs (Ultra 5s and older Sparcstation 5s) had public IP addresses and the ISP I used gave all 10 Mbps customers 5 public IP addresses. I've recently started taking a few college courses again, the uni's labs are all NATed (so you can't access /tmp or /var on workstationname-57.lab04.cs.unidomain.tld from home any more, you have to dump the files on your NFS mounted 150 MiB home dir and then access that, great fun) and my current ISP gives each customer ONE public IP address, but I suppose I should consider myself lucky for not being NATed...
Seriously, we need to move back to an internet where a machine connected to the internet can almost always be assumed to have a proper, public, IP address. It would simplify a lot of things. Also, any trolls pulling out the "yuo cant has teh firawalls withouts teh NAT!!!11" crap can please not respond to this as packet filtering does not in any way require NAT. (Not directed at parent post, just tired of trolls and ignorant fools always using that argument).
/Mikael
Greylisting is to SMTP as NAT is to IPv4
I usually do not reply to my own posts (or replies to my posts) on /., but this is one area where I think it may actually be important.
First of all, if I were to guess, I'd say that all those who replied while questioning my background don't actually do network development for a living. While I could start beating my own chest about how most of your traffic right now probably goes through something designed by me, that would be beside the point (and noone knows you are a dog on the Internet :) ).
That said, a few points specifically.
1) "Never heard of structs?". Structures are orthogonal to the size of IP addresses. You can represent IPv4 address as a structure (as original in_addr used to do, exactly because not all hardware supported 32 bit natively). You could do the same with IPv6 (or you can simply stuff it into 16 sequential bytes). What won't change is ability to perform operations directly on the data type. :) ). This is inefficient, prone to error and makes code less maintainable.
You can natively compare two v4 addresses by using a == b (which will translate into a single assembly instruction). You cannot do that on a 129 bit data item. Your choices are - memcmp, or defined operation (compare first 4 bytes, then next 4 bytes, then next, then next
2) Radix trees. Sure, anything can be stored in a radix tree with appropriately long prefix or appropriately large number of nodes in a prefix. What can't be done, however, is keeping this tree in memory (given current device and system memory sizes, which are in low gigabytes to a few dozen gigabytes). This problem is exacerbated by the fact that IPv4 address space is very compact of necessity (not too many holes, and everything is neatly CIDRed together), whereas IPv6 is of necessity full of holes (and designed to stay that way).
3) Performance is a relatively minor consideration in this.
As far as NAT goes - I firmly believe that solutions (in technology and elsewhere) are of two kinds - "organic", i.e. borne of and supported by needs and circumstances, and "artificial". Organic solutions are not always streamlined or pretty. Humans are a good example. A rock of salt is pretty darn inorganic (though I wouldn't want to stretch this analogy too far :) ) NAT is the former, IPv6 is the latter.
Well that whole 640k thing with regard to IP addresses has been largely negated by the adoption of routers within the home. Back when cable/DSL adoption was first starting, many people would end up with a switch and then have to call up the ISP for a second IP address. And with several computers in every home these days (not to mention other devices that grab IP addresses - games consoles, WiFi cell phones, network printers, etc), that plausibly could have become a very big issue very quickly. I've got at least a dozen pieces of hardware that consume a local IP address (not to mention the two or three VMs I have going at any given time), and it's a very good thing they don't each consume a slot in the worldwide public address space.
For all practical purposes, even an A.B.C.D.E would probably be enough thanks to routers - that still gives us ~1 trillion unique IPs worldwide. Of course if we were to make the switch it would make sense to give us the additional headroom. I'm hardly intimately familiar with the inner workings of IPv6 but assume it has benefits beyond mere address space, but the added complication to sysadmins of dealing with something like "2001:0db8:0000:0000:0000:0000:1428:57ab" (thanks, Wikipedia) is simply a nightmare in the making. Four bytes versus sixteen? I can remember which computer is 192.168.0.11 on my local network easily enough (and could certainly remember my public IP if I were bothered, as it never seems to change despite not paying for static), but you can practically smell the smoke coming out of my head after just looking at that.
It's certainly forward-thinking, but having (estimated) fewer atoms in the universe than IPv6 addresses available is just slightly overkill, doncha think?
How are sites slashdotted when nobody reads TFAs?
1) The world is document centric, not IP address centric. I want to access a collection of named documents and services from "slashdot.org". I dont care if these come to me by IPv4, NetBUI, IPX/SPX, Token Ring or Carrier Pigeon. I want to get "slashdot.org" and I want to make sure "slashdot.org" really is "slashdot.org" and not "somephishingsite.com"
So what you're saying is that you have no real reason to be anti-IPv6?
2) "End 2 End" isn't a selling point. I dont want my home network to be publicly visible.
So stick it behind a firewall that blocks incoming connections to all IP-addresses assigned to you unless you allow them?
3) Protocols that route around my desire for #2 succeed. All good P2P clients support UPnP. 3.1) Protocols that do not work with my desire for #2 fail. See Active FTP and the failed or failing IM networks and IM software that do not transfer files over NAT.
So, you'd rather have ugly workarounds than see the internet work the way it's supposed to work?
4) Those P2P clients are proof that how documents get to me are independent of the underlying link. I have no doubt that BitTorrent could be easily adapted to operate as a wire protocol on 802.11g or on top of IPX/SPX.
See answer to #1
5) If (and a big one) IPv6 got any traction, smart entrepenuers will began creating new services or modify existing ones like BitTorrent to operate and bridge IPv4 and IPv6. Really smart ones will most likely realize that once they abstract TCP/IP out of their design, they can do other "fun" things like implement their file sharing network directly over WiFI or some other mesh type network.
Have you even heard of the OSI model? Why in god's name would you want to have a Layer 3/4 P2P protocol? That's what TCP and IPv4/IPv6 are for.
/Mikael
Greylisting is to SMTP as NAT is to IPv4