Slashdot Mirror
IPv6 Ready For A Spin
ibjhb writes: "Sprint and WorldCom are itching to launch the IPv6. This will provided us with the 'zillions' of extra addresses not provided by the current IPv4. There's other capabilities, including increased sercurity. ZDNet carries the story ..." Seems like we've been talking about IPv6 for as long as I've been using IPv4.
The combination of DDNS (Dynamic DNS) and DHCP could be very cool. You'd never have to do any TCP/IP configuration on your machine (thanks to DHCP), but you'd still get the luxury of having the same DNS name all the time (even if the actual IP address changes from time to time).
Actually, typically the difference between residential and business service is quality of service and support, not the bundled services.
For example, if you were to get T1 service and it were to go down at 2 am, you could call them up and have them run a diagnostic on your line while you on the phone and get them to come out in the middle of the night and fix it if there was line trouble.
Also, several providers have different guarantees and quality of service ratings between their business and residential packages. Some go as far as to build a completely seperate network which has a lower saturation rate for business customers than residential customers.
As far as DSL is concerned, consumers typically get ADSL while businesses get SDSL. While bandwidth might be bandwidth, the latency differences between the two types of circuits is much different. Telcos have been running HDSL in place of T1 circuits for some time now when it's cost effective to do so (SDSL is sometimes referred to as HDSL-2).
Also note that more IPs given to a customer implies that the customer will have more machines on their network acting as servers and more bandwidth usage on average. It's a cheap way of doing accounting.
The world is neither black nor white nor good nor evil, only many shades of CowboyNeal.
but if they want a TLA (/48), it's gonna cost them $20,000 per year. For a /35 or anything less, it will cost $2,500.
That's reversed! the number is the number of bits in the network number. The subnet gets the rest. So a /35 costs $20000.
At any rate, a /48 is a huge space to work with (80 bits)! That's 2^48 entire internets!.
In practice, that is sub-divided into 16 bits for subnets (an entire class B sized space) and the rest is per-host (enough to comfortably hold the entire current internet 2^32 times!).
At that rate, they DO grow on trees.
The easiest way to engineer that is to buy a /48 and divide that into 16232 regional offices. Each office then can feel free to accept as many customers as it wants (and has capacity for) since the 64bit host address is guarenteed to be unique worldwide.
The real question is which providers will be stingy and try to screw the customer for every last penny, and which will just provide the bandwidth without hassels. Since it is a compeditive industry, the former won't last long.
I think the real problem is going to be getting the end-user machines upgraded to IPv6. Five years ago it could have been done, but now that grandma and grandpa and all sorts of redneck lusers are having a hard enough enough time just getting their little Windoze machine to read e-mail and browse cnn.com (or nascar.com), how easy will it be to get THEM to switch? You can try to wait them out until they get a new machine, but then someone will buy their old used machine.
Sounds like elitism to me. How many Grandmas, Grandpas, and 'redneck lusers' do you know that actually set their IP addresses? Most get them via DHCP. Most of these machines are Windows-based machines, which soon will quite easily support IPv6 (Windows ME) and may already (anyone know if Win98 supports it?). If anything, ISPs (who these end-users are connecting to) can mass e-mail their customers and say, "Look, we're moving over to IPv6. If you're running this version of this operating system, be sure to upgrade with files found here." Send that out over the course of a prep period (say, 4-6 months) and then when the time is up, just start migrating, leaving one bunch of lines using IPv4 addresses mapped to IPv6. It's not that hard.
And, I mean, that's just one very painful solution. You could also map IPv6 addresses on your end to a block of IPv4 addresses you keep for machines that specifically need them, making the entire process.
The whole, "Older users will be alienated!" is a cry of the alarmist. The true implementers will find a way around this. Yes, with all technology upgrades, a select few will be obsoleted, but don't you trust that the people who want to implement this have thought of ways to get around possible roadblocks?
Not to be too offensive, but you do understand the ideas behind a free market, right? "What's in it" for ISPs is that if they don't get their ass in gear, they are out of business because everyone else has a better product.
But unfortunately, Linux is not an OS. Having IPv6 in the kernel is fine, but it is a *major* PITA to get even the basics (ping, traceroute) recompiled, etc. FreeBSD is way ahead of the game in this case.
Some OS vendors and hardware manufacturers have been delaying IPV6 support for too long.
However even MS now runs with IPV6 if neaded. Cisco, IBM Sun all have support. Linux of course has had it since 2.0.x
--= Isn't it surprising how badly I spell ?
For one, 128 bit vs 64 bit vs 32 bit has nothing to do with processors when it comes to network addressing.
There are several factors for going completely overboard with the addressing.. It is better to err on the side of overkill (when it dosen't effect preformance) in that as our need for the internet starts to expand into more areas of our life, we risk getting into the situation where we are now. That being internet addresses being a scarse resource that must be concerserved.. When your wristwatch, pda, laptop, tv, vcr, coffee maker, etc, etc, are all connected, the last thing we want to have to deal with is lack of address space.
Granted, all that is a damn long way off. And people will have to start to get damn serious about their network security before I am going to be able to start a pot of coffee at home from an atm machine or someting wacked like that.
Remember, when ipv4 was concieved, noone thought it would ever run out of addresses. Yet its right around the corner...
...
. ""The future masters of technology will have to be lighthearted and
. intelligent. The machine easily masters the grim and the dumb."
- My Blog - http://www.memestreams.net/users/rattle/
DHCP is just a tool, and like most tools can be used for good or evil. (heh)
/release' and an 'ipconfig /renew' if you need to push the settings out faster.
It doesn't just set your IP address. It sets your default gateway, your DNS server(s), your WINS server(s), extra routes, time server(s), domain name... all sorts of things. From working quite a bit with Microsoft DHCP, I think that particular implementation offers about 80 possible settings that can be transmitted to the clients. This is most likely documented in an RFC, but I haven't a clue which one offhand.
It's EASIEST from a network-admin standpoint to have everyone use floating IPs. However, the existing DNS structure is not really set up to support the concept of hostnames that can change on an hourly basis. There are modifications to it that are starting to get into production, mostly spearheaded by Microsoft. (and I personally DO NOT suggest you trust your DNS infrastructure to that company -- any foothold you give them into your central infrastructure will someday be used against you as they fight to eliminate all other operating systems from your network.) BIND, the gold standard of DNS, does now support dynamic updates, but I haven't tried to get this working yet.
If you don't have an updateable DNS server, you can get around that problem by using DHCP reservations. I use them a LOT. They do take a bit more work, but I think the payoff is worthwhile. You end up mapping a particular IP address to a particular MAC address -- essentially tying an IP address to an Ethernet card. (which means that IP addresses follow cards, not machines, which you have to be aware of.) But you set the client to full-automatic DHCP. It THINKS it's a floating client. It doesn't know that it will always get the same IP number.
This lets you push big changes out in X/2 time, where X is your DHCP renewal interval. (clients renew their lease when 50% of the duration expires). Instead of having to visit every client to adjust their WINS settings if a server goes down, you just adjust it at the server and the whole network soon knows about it without you needing to do anything else. You can also tell everyone to do an 'ipconfig
Because of all these niceties, it's unlikely DHCP will go away as we transition to IPV6. The concept of automatically-configured network settings is just too useful to go away. It will undoubtedly morph over time, and Microsoft will probably rename it at some point to make it sound new so they can sell it to you again, but the basic concept is here to stay.
Ok, granted, 32 bits wasn't quite enough.. but what is the point of a 128 bit IP address? 64 bits would be more efficient to deal with (since we're getting 64 bit CPUs a plenty, whereas we may never see a 128 bit desktop chip) and would still provide us with 16 quintillion possible addresses. Is 16 quintillion not enough?
Normally I keep up with current stuff like this, but I've never looked much into IPv6... Is there room for backward compability, or is this going to be one of those "please change your ip by august 31st?" type things?
Seems like we've been hearing about IPv6 for forever now.
---
All of this is taken care of for you in IPv6 as well, albeit in a different sort of way. The subnet mask is handled by "Router Advertisements", DNS is handled by an "All DNS Servers" multicast (if I remember correctly, I could be wrong - this may be an advertisement as well), etc. IPv6 is truly a next generation protocol in *many* ways. There is a DHCPv6 spec, but I doubt that many (if any) installations will have to use it as all of the functionality in DHCPv4 is automatic in IPv6 via its 'stateless autoconfiguration'.
For more information, see:
Every IPv6 RFC I've read (that I can remember) has been a good read, so check 'em out. :-)
EthanNASA is already using multicasting on some of their internal networks to efficiently send high-rate spacecraft telemetry streams to multiple destinations.
Mea navis aericumbens anguillis abundat
Right now, NAT is pretty popular since it's a good way for a bunch of machines to share the single little IP address that the ISPs typically hand out. A side benefit of NAT is that you almost always get a firewall as well. They just go together, ya know?
But when ISPs that handing out IP addresses by the thousands, I think more people (especially people who aren't security conscious -- that's 99% of the population) won't need NAT. So they'll just simpler routing, which will likely pass all traffic.
I have a feeling that a lot of people's LANs are going to be wide open to attacks when IPv6 becomes widely used.
---
As copyright owner of this comment, I authorize everyone to defeat any technological measure which limits access to it.
But they don't just "grow on trees." Have you bothered to look at ARIN's allocation policies for IPv6?
/35 or anything less, it will cost $2,500.
Pacbell hasn't even ventured into deploying IPv6 yet, but if they want a TLA (/48), it's gonna cost them $20,000 per year. For a
But then you need to ask yourself the real question: 'What am I smoking and why do I think that ISPs depend on the revenue from "selling" IP blocks?'
Sprint (NYSE: FON) is also looking to IPv6 as the vehicle for far higher levels of service security and performance
Looks like there's little bit of spin going on all right. When the spin gets intensive it'll look like this:
Sprint is charging ahead with IP v6 technology. IP v6 will let us provide unparalelled performance to our customers. Our leadership in this area makes us the only choice.
MCI/Worldcom will of course issue the exact same press release.
You can gauge the health of a technology by measuring the amount of corporate spin it generates.
--Shoeboy
I don't see v6 happening on the large scale any time soon, either. The ugly truth is that if you upgrade your network to IPv6, you still need to speak IPv4 with the rest of the world. So, your snazzy-new network's hosts have to go through a protocol translator (a NAT box that does v6->v4 conversion) to access anybody (read: most of the world) that hasn't upgraded to IPv6 as well.
Compare this with using IPv4-to-IPv4 NAT. NAT is more widely deployed and understood, and probably easier to put in place than converting your entire network or setting up lots of v6-over-v4 tunnelling. Sure, you don't have globally visible addresses for each of your hosts. But new services have to be designed taking NAT into account anyway.
Until there are enough IPv6 hosts to make speaking IPv6 useful, there's no incentive to upgrade networks and hosts. The claims of increased security and QoS over v4 are, in my opinion, vastly exaggerated--- just because v6 has a "flow identifier" field doesn't mean anybody's decided something useful to do with it.
See "I'll stick with IPv4 for now, thank you" from Network Fusion for a "testimonial" from a network administrator who's just as happy not upgrading his network...
While I understand that every ISP out there would like to have a virtually limitless number of IP addresses to alocate to their subscribers, won't this damage their business model?
If I pay PacBell nearly twice as much so I can have 5 IPs instead of one, or pay five times as much for a 32-address subnet, will they still be able to justify charging more for multiple IP addresses when they grow on trees?
Will the switch to IPv6 end up costing them a good deal using this revenue model, or will we all switch to (horror of horrors) a per-byte revenue model?
Kevin Fox
Kevin Fox
Sound familliar?
Boffoonery - downloadable Comedy Benefit for Bletchley Park
Yes, but for all the places out there who currently have class A's they will be given a block of ipv6 addresses based on their ipv4 address, which means that instead of the class A having their usually 16 million addresses, they will probably not have trillions upon trillions of addresses, mainly for the sake of simplicity in the switch. Once this is switched over, the new ip addresses will be assigned in a totally different way, and won't be waisted. Plus its highly possible for those Class A's to give back a trillion or so addresses once things get cleared up.
true. short of routers rejecting ipv4 packets from *all* the router macnufacturers, it isnt going to happen. heck, i have ipv6 enabled on my linux boxen and i cant get routed because the ISPs router doesnt support it.
I am actually doing a report on this and found some really good articles and other tidbits at the following:
- Paper.html#CH2
5 6 addresses.
http://playground.sun.com/pub/ipng/html/INET-IPng
http://www.6bone.net/misc/case-for-ipv6.html
As far as compatibility goes, they (Internet Engineering Task Force (IETF)) have really worked hard on making a solid set of standards for this. The Sun.com resource I posted goes into pretty good detail on how this will all come together.
Before reading this article, I was totally under the impression that the ONLY need to go to IPv6 was due to lack of IPv4 addresses but that is not the case at all. IPv6 has a ton of nifty add-ons as far as the extension headers and the size of this new header is really only about twice the size of the IPv4 header. The new extension header includes:
Routing - This is considered Extended Routing which is based somewhat on the IPV4 source routing steps.
Fragmentation- This will allow headers to be fragmented and be able to reassemble itself back together.
Authentication- This will include integrity and authentication checks to enable better security over the IPV4 standard.
Encapsulation- This also deals with security and enables the packet to be kept Confidential.
Hop-by-Hop Option- this will allow hop-by-hop processing.
Destination Options- Optional information that will be examined by the destination node.
Also some cool facts:
The issue of the number of addresses availible for IPv6 works out to be around:
340,282,366,920,938,463,463,374,607,431,768,211,4
And this breaks down to about 1,500 IP addresses for each square meter on the surface of Earth and that an estimated fifteen percent of the address space for IPv6 would be used in the initial switchover from the old standard to the new standard.
-Gregg
------------------------------------ Step into my Office... WhY? Cuz your %$#$ing Fired...
Guys, this will probably get moderated into the dirt and have me coming up with a -4 score, but I honestly see IPv4 as the metric system in America, how long now as the US been switching the the Metric system? 50 years? I think we're looking at the same with IPv6.. everyone is talking about how great it's going to be, and how wonderful all these new features are going to be.. but honestly if you look at ALL the servers all the everything on the internet that's using IPv4 (well.. lol that IS everything) for us to have to stop all that and change to different protocols and stuff, it isn't just going to happen anytime soon or at all.. I don't care if major companies DO jump on the IPv6 band wagon, I just don't see as becoming big as some people would like to see it happen.. of course this is coming from the mind of a 17 year old systems admin and I haven't been in the field that long so who am I to say that? but thats just my .02.
-Thanks bubu_.
Some links: /pages/ipv6.html /public/784/packet/july98/9.html
Sun http://www.sun.com/software/solaris/ipv6/
IBM http://domino.ngi.ibm.com/patrick
Cisco http://www.cisco.com/warp
W3C W3c site and mailing list search
IETF http://www.ietf.org/html.charter s/nat-charter.html
___
I'm clueless in this area, so treat me like I'm 2 years old, please (not by slapping my hands). Anyways, was there ever an IPv5? If there was, assuming we don't use it because it was essentially broken, why was it dropped and not re-engineered to become what we know as IPv6. If there wasn't, then why not?
--Drew Vogel
IPsec (encryption on the network layer) is mandatory for IPv6. See for example RFC 1825. Linux has an IPv4-only implementation of IPsec, see FreeSWAN.org. Hum, tracing,.. IPv6 by default builds host addresses from a prefix (kind of subnet mask) and the 48 bits ether hardware address. Since the latter are supposed to be uniq, voila, even better than intel's pentium serial numbers. But IPv6 allows many addresses per interface, so you could use the automatically generated ones only for, say, booting from the net, and use chosen or random addresses otherwise. Have a nice Day.
I can't help but wonder who'll be first to get stuck with 029A:7734:029A:7734...
Well, 0x029A = 666. so I see where that comes from... but where the fsck does 7734 come from?
Tarsnap: Online backups for the truly paranoid
When IP4 came out people KNEW that, unless the internet collapsed under it's own weight (a recurring prediction at that time) that the address space was going to 'fill up' mostly due to sparse address assignments
Examples:
- IBM, the CIA, and various other 'important' entities of the time each have 16 million address tied up in class A subnets... Guess how many they're using?
- I know somebody else who has a 'personal' class C routable address (from long before address space started getting 'tight') of which he's probably not using more than 16 addresses.
- I personally know of one company with less than 200 machines that has 4 Class 'C' subnets from two different ISPs.
These are not unexpected events. The issue at the time (as I remember it) was really around machine capability. This was at a time when a SUN workstation with 1/2MB of ram was considered HOT (actually, they hadn't been built yet), 2 MIPS was considered decent for a mainfraime, and the work it would have taken to disassemble a variable-length address would have been noticable.For those (and, I would expect, other) reasons, it was decided to go with 'only' a 32 bit address space and work on a more 'realistic' address space for later when machines wouldn't care about the extra cost. That time is now.
A 128 bit address space isn't just frigin huge. it's close enough to infinite for just about any practical purpose. Physics is probably going to get in the way before we run out of IP6 addresses. Designing a system that was extendable beyond 128 bits would have been ASKING for trouble. There wouldn't have been need for most people to implement it. This would have meant that we would have these atrophied pieces of code that would be generally unused, untested and misunderstood for years because 'we really don't need that'.
It would also introduce unnecessary coding complexities.
I guess that the short answer is: KISS . . . Keep It Simple, Stupid.
Free Software: Like love, it grows best when given away.
I always kind of wonder whether this will divide the whole internet along monetary lines. When this switch happens, networks in the third world are likely going to take significantly longer to upgrade, if they do at all. Also the transition would create a glut of very cheap, very good IPv4 routers that might be just too tempting to someone who has to create a network on a very small budget.
Since you can browse an IPv4 network from IPv6, but not vice versa, this creates pockets of the internet only accessable to people with the cash to upgrade their routers. Maybe I'm wrong, I hope I am, because I don't much like the ramifications of having two "internet classes"
Such a price the gods exact for song: to become what we sing - Pythagoras
Classful routing died five years ago because the routing tables were going out of control, and because class C was too small, but class B was *way* too big.
So they came up with CIDR (classless inter-domain routing) which uses variable net masks to group routing blocks geographically. They divided up the rest of class C and used that all up, so now they're cutting into the old class A. So lots of cable modem customers have 12.*.*.* addresses now.
For more info on CIDR, see RFCs 1812, 1817, 1860, etc.
However, with 128 bits, it is likely that end-users will get *at least* 48 bits (enough to use Ethernet addresses and reduce the need for site-local DHCP), probably 64 bits, for their own use. So a block with a 64 bit prefix will be treated like a single 32-bit address today.
#naabhaprzrag, #sverubfr-000, #agi-fcbafberq, negvpyr[pynff*=' negvpyr-ary-'] { qvfcynl: abar !vzcbegnag; }
I think the real problem is going to be getting the end-user machines upgraded to IPv6. Five years ago it could have been done, but now that grandma and grandpa and all sorts of redneck lusers are having a hard enough enough time just getting their little Windoze machine to read e-mail and browse cnn.com (or nascar.com), how easy will it be to get THEM to switch? You can try to wait them out until they get a new machine, but then someone will buy their old used machine.
Sure, sites with DNS entries can always be given an IPv4 address for backwards compatibility, but that doesn't solve the problem of all the jillions of potential client machines out there with ancient IPv4 protocol stacks that may never be able to connect to anything with an IPv6 address but no IPv4 address. (Though I suppose a sort of reverse-NAT could be done if you combine a NAT box with a DNS server and have it spit out 10.x.x.x addresses.)
It's also a problem for those with legacy machines, including antique computer collectors.
I personally expect to be ready when it comes, but I'm pretty damn net savvy and I'm still not sure what I'll need to do to get IPv6 compatible.
It could become a fiasco on the level of the metric system, and it could create a new divide between the "haves" and the "have nots".
#naabhaprzrag, #sverubfr-000, #agi-fcbafberq, negvpyr[pynff*=' negvpyr-ary-'] { qvfcynl: abar !vzcbegnag; }
Any kid who grew up in the '70s can tell you that 7734 is the word "HELL" upside-down on a calculator. :)
#naabhaprzrag, #sverubfr-000, #agi-fcbafberq, negvpyr[pynff*=' negvpyr-ary-'] { qvfcynl: abar !vzcbegnag; }
When IPv6 comes into use on the net, will I be able to get my own block of IPs for cheap? How much?
/64 subnets to users.
With 6to4, you can get a whole subnet for free, and you don't even need to wait for your ISP. ISPs will probably assign either individual addresses or whole
How many IPs fall into the lowest level "block?"
The lowest level block is a 64-bit subnet, with 2^64 addresses.
If I buy a block from NSI or whomever, can I "symbolically link" it to say, a SW Bell DSL line?
No, that would destroy the global routing database. You can't do it today with IPv4 and you won't be able to do it with IPv6. I think ICANN is actually the ones allocating IP space these days, although I'm not sure how ARIN fits in.
I'm not an IPv6 expert, so if any of these are wrong, just correct me but flames aren't appreciated.
You don't have to use a MAC address; you can use any address that's not used on your subnet if you're concerned about privacy.
And you don't even have to wait for your ISP to support it. Just turn on 6to4, and with one IPv4 address you can get quite a few IPv6 subnets (each of which contains 2^64 addresses).
Windows already supports 6to4, BSD probably does, and I don't know about Linux.
I understand Mac OS X to support IPv6 natively. Are they positioned to take advantage of this switch like no other "consumer OS"?
--
$tar -xvf
FreeBSD's 4.0-RELEASE branch supports IPv6. If you're interested in getting your own IPv6 Internet address and being connected to Internet6, the Kame project is what you're looking for.