What About IPv6? How Long Until Widespread Deployment?

Christopher Blood asks: "Over at the register, they talk about the EU adopting IPv6. So what about the USA? When do we get it? IPv6 would solve some and DOS problems and we will need the extra address space. What's the holdup?" While IPv6 may be the cure for all of our IPv4 ills, upgrading the whole internet to the new technology isn't going to happen over night. What has been done to prepare for the jump, and what still needs to happen before it can become a reality?

Comments

Well, it's here already

[Moridineas]

At my university, IPv6 has been deployed since last year, maybe longer. I've been running FreeBSD w/ IPV6 for at least that long. Honestly, it hasn't made that big a difference for me :)

Re:Well, it's here already

[benedict]

That's the worst idea I've ever heard!

NAT causes a lot of problems. It's an ugly hack,
not a solution to the world's ills.

Re:Well, it's here already

[Slak]

As Homer J. Simpson says, "Here's to alcohol, the cause of and solution to all the world's problems"!

Re:Well, it's here already

[jaavaaguru]

I have not seen one benifit for IPv6. I do not say IP for my toaster. There is not a single benfit for the cost or hasle of the millions of machines that need to changed.

The lifetime of a PC is around 10 years usually. Manye peopel will get a new one before that time is up and many people will use older machines, but I'm making a huge generalisation here. If all new PCs came with IPv6 capabilities (insert preferred "put linux on them" phrase here) then there would be no additional cost or bother caused and everyone would be on IPv6 within the 1st 10 years of the 21st century :-)

Not that it makes a huge difference to anyone, but I feel left out not being able to address an increasingly large portion of the internet because my ISP is'nt providing IPv6 to consumers yet. I would be quite surprised if they don't use IPv6 technology elsewhere though, even if it's just for future safety - why use an addressing system that's rapidly running out of space when you could JUST AS EASILY use one that will last for ages before it runs out of space.

I know that "hey my toaster's got an IP address" is a bit ridiculous (assuming you're not counting Color Classics and the like as toasters ;-), but realisticly it would be perfectly possible for all mobile devices (PDAs, Cellphones, etc) to have a unique IP address in the near future.

Re:Well, it's here already

[Gid1]

Hewlett Packard have (among other class B and C blocks), the 15/8 network. That's 15.0.0.0 - 15.255.255.255.

In their corporate manifesto, the "HP Way", they claim "citizenship" and contribution to the communities they operate in to be one of their seven most important corporate objectives.

They operate in the Internet community, yet claim 16 million addresses for themselves, even though practically all of their internal machines are hidden behind a solid firewall system for which NAT would not be a big problem (and possibly a security asset.)

I know it'd be a big job to fix, but it still doesn't seem to be good citizenship to me.

Re:Well, it's here already

[Yokaze]

Ever thought of mobile devices? Especially obiles/cellulars?

Do you expect them to NAT all those?
How will they route their traffic?
All over one NAT-box?
Over several NAT-boxes?

Or do you expect them to assign them a dynamic IP for every connection?
When is the IP free to use for another device?
After 10s of no use? 20s?

How do you achieve a handover between Gateway GPRS Support Nodes?
Or between different telcos?
Or between different telcos in different countries?

Re:Well, it's here already

[invenustus]

I'm not looking forward to that day. If my computer gets 0wned, I might lose my saved email. If my toaster gets 0wned, I might lose my house. Think about it.

Re:Well, it's here already

[Random Feature]

The benefit is to corporations. IPv6 contains a field in the header specifically geared to handle QoS on a global basis. It's for PRIORITY.

If IPv6 is globally implemented, and corporations can get the backbone providers to adhere to the priority fields, their traffic will get priority over yours and mine.

The same could be done using DiffServ or TOS now, but they aren't universally processed by all the routers on the backbone.

There are benefits - but most of them are corporate.

Re:Well, it's here already

[JLouder]

They operate in the Internet community, yet claim 16 million addresses for themselves, even though practically all of their internal machines are hidden behind a solid firewall system for which NAT would not be a big problem (and possibly a security asset.)

My employer has a similar setup -- many class B networks, all allocated to systems that are firewalled off from the Internet, set up this way because it required less thinking than NAT.

When IPv6 is widely used, I imagine we'll see much more of this foolishness.

Re:Well, it's here already

[Gid1]

First thing I did when I took over responsibility for hosting and internet connectivity at a (largish) company I worked at was to replace their existing public IP space (a few thousand addresses) with private IP, hidden behind NAT. It made internal routing *far* easier.

Of course, a few hardcore techies complained. So, I said that if they had a problem with it, they could come tell me why. If they had a good reason for public IP and they convinced me they were trustable as far as security was concerned, I'd happily give them as many of the deallocated public addresses as they needed, and noted them down carefully. After a few months, those allocations would be reassessed.

As far as HP is concerned, something like:
find . -exec perl -pi -e 's/15\.(\d+\.\d+\.\d+)/10.$1/go'
should do the trick! =)

Re:Well, it's here already

[Cramer]

No NAT does not. There are some problems, but they are very specific to stupidly engineered client/server programs where the server attempts to contact the client (using whatever the client thinks is it's address.) Almost every java rmi/corba based piece of shit has this problem.

Next you're going to say firewalls causes lots of problems.

Re:Well, it's here already

[Dwonis]

Some applications have never even been developed because of NAT. For instance, you will never be able to log into your answering machine from across the world to check your messages until IPv6 hecomes widespread.

Re:Well, it's here already

[-brazil-]

God, that's bullshit. There isn't even enough IPv4 addresses around to give one to each person, and static IPs are desirable, and more than one of them per person.

Don't you realize how idiotic it is to avoid the update to IPv6 by instead requiring an update to NAT and an update of every protocol that doesn't work well with NAT. That's more time and money wasted, not less!!

Re:Well, it's here already

[Dwonis]

Yep. Even the whole instant-messaging thing is just a big hack on the fact that most people can't run their own SMTP/talk servers (otherwise, ICQ could have been implemented using these standard protocols, rather than using their own proprietary, client/server model.

Re:Well, it's here already

[Yokaze]

> You could make each tower hand out it own IP subnet. Some of this is handled in the wrieless today and routing to your PC as you wonder the building and single is jumping from access point to access point.

The different access points are all part of a single Ethernet. You surely didn't tried to wander from a different administrative network to a different one while using the Internet.

> The only big problem now is getting a DNS to up the with your current ip/route...

Not really, unless you want to have a server on your mobile.

The real problem is the routing.

Let's say you travel from Strasbourg to Venice.
Someone else is driving, you're bored and you have to much money. So what are you doing? Watching a video via a mobile.

What does that mean?
You initiate a connection to a server, which will send the data to your IP. Now what happens when your leaving the reach of a Base Station Controller?
If I understand you correctly, you'll recieve a new IP.
Wonderful, what happens with your video?

Now let's assume you keep the IP between different BSCs.

Sorry, pal. You're leaving France, here comes Germany, I guess the France Telecom doesn't provide access there.

That would be a switch between two telcos.

Furthermore, several telcos only provide partial coverage and outsource the missing areas to a different telco.
What happens when you leave that area?

Now, let's assume that the telcos don't have a problem with handing out IPs from their pool to a different telco, how will the traffic be routed?

Always over all adminsitrative borders you have passed, since one carrier will never allow a different carrier to mess with their routers.
Imagine what would happen: what work those routers'd have to handle, every millisecond a change in routes, routing tables over several thousand entries.

So, happy travelling from Strasbourg to Venice :)

Re:Well, it's here already

[benedict]

Firewalls cause lots of problems. IMO there's a
more solid engineering rationale for firewalls
than for NAT, but they're definitely also an ugly
hack.

There is something to what you say. NAT exposes
problems in certain poorly-designed protocols.
For example, it interacts badly with the layering
violations in FTP. But even if all other protocols
were designed well, NAT would be a poor long-term
solution. It destroys the end-to-end transparency
of the network. It makes troubleshooting difficult.
It introduces points of failure.

Backbone

[crumbz]

Given that Lucent's CEO said today that he does not see the telecom equipment market turning around anytime soon, a government initiative to upgrade the core routers to IPv6 would help boost the battered sector of the economy. Granted, Lucent shot themselves in the foot last year (several times) and upgrade to IPv6 might just result in a higher volume of spam.....

Newbie question..

[zapfie]

Is IPv6 backwards compatible with IPv4?

Re:Newbie question..

[ColdGrits]

Yup.

As in IPv4 addresses are just a subset of IPv6, so any IPv4-based stuff will still work in an IPv6 network no problem.

Not true the other way round, but then that doesn't matter :)

So yes, they could upgrade the entire Internet backbones etc to IPv6 (and *should* do so asap) and all old IPv4 traffic will carry on as normal.

Re:Newbie question..

[Codifex Maximus]

There is alot of code based on IPv4 functions - upgrading all the programs to use IPv6 would be a mammoth task... hey, maybe there will be some jobs opening up in Socket Programming soon?

Re:Newbie question..

[MenTaLguY]

For _most_ network-aware applications, the only thing different is the address format. Once you have the connected socket, the rest of the network code should remain unchanged.

It's the (non-socket-related) code to handle e.g. address parsing which has to change.

Re:Newbie question..

[Codifex Maximus]

> For _most_ network-aware applications, the only
> thing different is the address format. Once you
> have the connected socket, the rest of the network
> code should remain unchanged.

So, essentially what you're saying is: After you get past all the things that are different then the rest is the same?

Ok, I'll buy that.

Re:Newbie question..

[MenTaLguY]

Yes.. although now that you made me sit down and think about it, a lot of stuff cares about addresses once you have a socket (getsockname, getpeername, send, recv, etc...).

When?

[Anonymous Coward]

I guess not in close future. When free IPV4 addresses run out, large address blocks reserved to big companies etc become very valuable. So, if you want addresses which work 100% of the time, you'll have to cough up money for the companies to get them. It will be that simple. Really.

Re:When?

[furiousgeorge]

true. but if you're not located next door to said company, the main trunk routing tables become ridiculous.

Remember --- M.I.T. has more assigned IP addresses than ALL OF CHINA.

It's not north america thats going to drive IPv6, it's Europe and Asia where they're already starting to feel the address squeeze.

Re:When?

[boopus]

Is that still true? Last I read they gave a large portion of their address space back.... For all I know they could have kept a coupple million though.

Re:When?

[noahm]

Is that still true? Last I read they gave a large portion of their address space back.... For all I know they could have kept a coupple million though.

No, it was Stanford that gave up their class A. What were they thinking? MIT still has ungodly amounts of address space. We have net 18 (18.0.0.0/8), plus random assorted /16s (128.52, for example, is the AI lab). There are a couple others.

The thing is, though, there's a whole lot of "reserved" address space out there. The IPv4 address space shortage is partially artificial. In some ways this is to preven the world from grinding to a screeching halt where there really are no more IPv4 addresses. Another is that maybe it will put pressure on people to be conservative with address allocation, which might make the shortage less pressing. Maybe it will also help to speed the deployment of IPv6.

Most OS vendors are already supportind IPv6 out of the box. WinXP, for example, can be set up as an autoconfiguring IPv6 host very easily (ipv6/install at a command prompt, IIRC). The BSDs support it very well, as do many Linux vendors. I think that it won't be long until IPv6 communication on the internet is very widespread. I don't, however, think the whole internet will be IPv6 any time soon.

noah

Re:When?

[4mn0t1337]

Remember --- M.I.T. has more assigned IP addresses than ALL OF CHINA.

Good! There are already enough spam relays in China (and very few that I am aware of at MIT).

We should make a deal that China doesn't get anymore IPs until they deal with all the spam coming from there. That and finish their dinner...

Re:When?

[Cardhore]

That's because China only needs on IP for its firewall.

Re:When?

[Cramer]

• I think that all new address space should be assigned to two ISPs at a time based on places where they regionaly overlap

And there begins the stupidity of your plan. Physical location doesn't have one flying f*** to do with the internet. There is almost zero coorelation between network proximity and physical proximity.

Re:When?

[thogard]

With the exception of the US (where bandwidth cheap), there are a limited number of ISPs in a position to allow smaller clients to dual home. This is true for most of the world with the exception of North America and thrid world countries.

I don't know of a single backbone carrier that couldn't cope with this plan if it were handed down by the IETF but none will push it because its too hard and will cut them out of exclusive contracts.

Besides IPV6 won't fix any of the problems with limited address space, it just makes the tables grow so large there will never be an efficent way to route (or even plan routes) and we will be stuck with the same problem but now we have 2x as much data it deal with.

Re:When?

[hta]

The Class A that MIT has is 16 million addresses.
According to ARIN, China has more than 20 million addresses assigned.
The crossover point was sometime in June 2000, I believe.

Re:When?

[Cramer]

(I deal with this multi-homed crap all the time.)

Those places with "25 smaller ranges" (which are very few) have all their address space from one ISP so they add nothing to the global route tables -- they are 100% covered by the ISP's summary announcement. When they connect to a second ISP, those address ranges have to be announced by both ISPs verbatum. Suddenly, there's the original netblock plus 25 smaller blocks being announced by the first ISP and the same 25 small blocks being announced by the second ISP. [*] Traffic flows to the most specific route, so all the smaller sub-blocks have to be annouced individually by all the ISPs.

* - The minimum agreed size for any BGP announcement is a /24. The minimum netblock for global routing is a /20 -- anything smaller may not find global coverage.

[There are companies providing products to handle multi-homing without any of the traditional hastles.]

What about the major backbone routers?

[kronin]

I would like to know how close the backbone through the US is to being IPv6 ready. Anyone that knows care to respond?

Re:What about the major backbone routers?

[Raindeer]

Japan and Korea are leading, together with some other countries in the Asia/Pacific region (APNIC-countries). Second is Europe (RIPE-countries). Third is the United States and its neighbours.(ARIN-countries), though the United States is second as a nation.

The reason I name the RIR's is that I base this on the amount of IPv6 space assigned. See:
http://www.ripe.net/ripe/meetings/archive/ri pe-41/ presentations/plenary-globalrir-stats/sld011.html
http://www.ripe.net/ripe/meetings/archive/ripe-41 / presentations/plenary-globalrir-stats/sld012.html
and here for the up to date list of all assignments:
http://www.ripe.net/ripencc/mem-serv ices/registrat ion/ipv6/ipv6allocs.html

Furthermore you might find it interesting that in the RIPE-area, the RIPE community has decided that all Local Internet Registries can apply for a /32, which should suffice for all of them :-)
You can find that policy here:
http://www.ripe.net/ripe/mail-archives/ipv6 -wg/200 20101-20020401/msg00093.html

roadrunner

[Maditude]

I emailed RR a while back about their plans for IPv6, and despite several back-and-forth email exchanges, never did find anyone who had even HEARD of IPv6, much less get details on their rollout plans for it. Doesn't look too promising for cable-modem users.

Re:roadrunner

[LWolenczak]

They RECENTLY got an ipv6 allocation from ARIN.

When do we get it?

[nublord]

When do we get it?

When corporate America determines they can make a profit from it.

Re:When do we get it?

[sabinm]

More right than not. Why in the world would corp ISP want to give you a static IPv6 when that is a constant bandwidth tap?

Joe Public will never "own" ip addresses again. That was made evident after the "great subnet rush" of the ninties.

Having IPv6 addresses mean that anyone can have as many as they want if given away for free-
until there is a way to consistently and legally charge for "per seat" usage for internet bandwidth, with crimial reprocussions (can you say DMCA) corporations will not adopt a standard which basically says, "a static and public IP address is worthless as a marketable commodity".

Re:When do we get it?

[Com2Kid]

Huh? Only ISPs can hand them out, therefore they have an automatic monopoly over IP addresses. Not to mention complete price control.

Hell ISPs should be PRAYING for IPv6 to come out and every device in a persons house + the toaster to have an IP address.

Think $1 per IP address.

Think 30 or more devices per house with IP addresses.

Tada, the average ISPs profit per customer just more then doubled.

(even broadband ISPs would be making an extra $30 on top of the ~$60 or so they already charge, 50% profit increase is STILL great!)

When Cisco decides to...

[sphealey]

There are two factors holding IPV6 back: lack of concensus from those that make the decisions in the networking world that IPV6 solves any problems that need to be solved at anything like a reasonable cost. And lack of push from Cisco for implementation. There are thousands of other facets to the discussion, but let's face it: if Cisco had said a year ago that "oh, IOS 12.x now supports IPV6 and we think you should start using it" the world would have fallen in line. They haven't, which makes you wonder what they know that we don't. The story is that "customers aren't demanding it yet", but that didn't stop them from introducing the router when no one was demanding them, did it?

sPh

Re:When Cisco decides to...

[univgeek]

Cisco road map (pdf) for implementation of IPv6 in IOS.

The same in html from Google.

They say that by/in 2002 (hey thats now) they will have completed implementation of all IPv6 functions in the routers.

Re:When Cisco decides to...

[Arandir]

In my case, there's only one thing holding back IPv6: my ISP doesn't support it. IPv6 is already completely integrated into my OS (FreeBSD), and effortless to set up. But it's useless until my ISP switches over.

Re:When Cisco decides to...

[isdnip]

Cisco knows that IPv6 is a lose; they have to support it, but don't have to push it hard.

IPv6 is a bad job, period. Most Slashdotters probably don't know its provenance. It has been around for about a decade. IETF created it as a compromise. IETF insider Steve Deering had created a poor-quality hack called SIP (Steve's IP) while insider Paul Francis (aka Tsuchiya) created one called PIP (Paul's IP). How bad? SIP, for instance, assigned all addresses by countries, based on population, and thus gave a shorter prefix to North Korea than to South Korea because it was a bit more populous in his almanac. IPv6 is PIP and SIP glommed together.

Just before the time it was adopted, IETF had adopted a different replacement for IP, TUBA (which I think was also called IPv8). TUBA used a profile of the OSI Connectionless Network Protocol (CLNP). Cisco had already implemented it, along with CLNP's routing protocol, IS-IS. CLNP was elegant and flexible -- some of the OSI work stank, but CLNP and TP4 were gems. The only reason TUBA was dropped was because Vint Cerf, the Chauncey Gardner of the Internet (not really so smart, but he's famous for Being There), changed his vote and dropped TUBA support.

Had Vint not been so perfidious, IPv8 would have been phased in before the public Internet boom of the mid-1990s. The code has been in Cisco and other vendor equipment for a decade.

IPv6, on the other hand, has a wasteful 16-octet address field (only 8 octets are useful at a time) and does little else to solve IP's problems. It does NOT provide QoS (that's an urban legend) or security any better than IPv4 with its existing options. And given the inefficient assignment of IPv4 adresses in the past, the 32-bit field has a lot of life left.

Think about VoIP: With IPv4, the header has 8 address octets, while the payload has to be short in order to minimize delay. And it's bloody inefficient. With IPv6, the header has 32 address octets while the payload is the same. It's a bleedin' joke! IPv6 is just plain wasteful.

IPv6

[SkewlD00d]

It would cost mucho dinero to upgrade all the infrastructure to support it. But, IPv4 and IPv6 can coexist. The prob is that 50% of lan equipment isn't upgradable. Telcos wont want to float the bill, they'll pass the buck to you.

Woot, most common excuse for downtime.... "upgrading."

YAEUU: Yet Another Expensive Useless Upgrade

the bothersome part

[nukey56]

IPv6 will fix a lot of problems, but one nasty side effect is that we're going to end up with addresses that look like 3ffe:400:34:fd01::1, instead of the easily memorizable four octets. When that day comes, it's going to be a lot harder to shout down the IP of the game server you're playing on down the hall.

"Oh, I'm on three-f-f-e-four,four-zero-zero,three-four,f-d-zer o-one,not(?),one. What's taking you so long?!?"

Re:the bothersome part

[Fastolfe]

We could have a cntral database where everybody applies for a unique, easy to remember coputer name.

Something like DNS?

Re:the bothersome part

[blair1q]

DNS. It's not just for breakfast any more.

--Blair

Moving a super-tanker

[iPaul]

IPV6 is better. Autoconfiguration, neighbor discovery, big address space, compatability with IPV4, etc. However, the more hacks we put in to make IPV4 work the harder it is to change. For the most part we're educating people to do "Stupid IPV4 Tricks" rather than moving to IPV6. The more of that we do the harder it is to change. Also, the more ominous the prospect of change, the more people will dread it.

Frankly, I'm thinking we might see another round, like IPV7 (or IPV8 if they make a habit of skipping odd numbers), or it might come very late. Maybe we'll see it on phones and wireless devices before we see wide-spread adoption of IPV6 or general purpose networking.

Re:Moving a super-tanker

[debrain]

IPV7 (or IPV8 if they make a habit of skipping odd numbers)

Odd numbers are development numbers. Same thing with the minor version of the kernel. (2.1, 2.3,2.5 are dev versions).

Re:Moving a super-tanker

[Yokaze]

Not really.
IRC, IPv4 was the first publicly aviable and IP (as defined in RFC760).
http://www.iana.org
IPv5 was taken by the ST-II protocol, which was supposed to be the next Internet Protocol (at least in the eyes of its inventors). But it was based on connection-oriented routing. This lead to a great resitance in the internet community, which is generally opposed to the idea of connections and channels.
It became experimental due to lack of support not by intend.

Have a look at the Version-numbers as assigned by the IANA.

For those to lazy to look it up.
IPv7 is the "TP/IX: The Next Internet"
IPv8 is "The P Internet Protocol"
IPv9 is "TUBA"

But some people are already joking that one will adopt an odd/even numbering scheme.

ISP incentive

[JDizzy]

A reward system needs to be enacted to entice the ISP's to provide unencumbered access to the 6-bone. ISP's that handle dial-up users can tunnel the ip4 traffic on behalf of their customers.

IPv6 will become the standard...

[popular]

as soon as we USians switch to the metric system.

Re:IPv6 will become the standard...

[scorcherer]

> as soon as we USians switch to the metric system.

I thought you're already making progress, inch by inch. Then again, it's not much better here in the Europe. I'll go now and have 0.586L of beer.

It'll happen when "everyone" knows how it works

[Colin Smith]

At the moment, IP VI is just a name to most network and systems administrators. My Linux boxes have VI support but I've never looked at it.

When there's available information about where to get addresses, configuring routes, netmasks, gateways, setting up name services etc. All the admin stuff that's done on a daily basis with IP IV.

At the moment nobody knows what they have to do in order to setup and use IP VI.

The installed base is hard to change...

[Old time hacker]

I think that IPv6 will take a significant amount of time to acheive widespread deployment in the US. Why? There are too many devices (cable/dsl router/firewall appliances) in use that don't support v6. While they may be flash upgradeable, I'm sure that the vendors would prefer to sell a new box which did support v6.

I've thought about running v6 at home and connecting up to the 6bone. However, the list of instructions was long and complex, and it was unclear to me that my existing ipchains based firewall code would continue to protect me. It was also unclear that I could enhance the ipchains rules to protect myself.

I quite like the idea of being able to expose multiple devices on different IP addresses, but it is (still) a non-trivial exercise.

On a side note, I'd like to see more deployment of multicast -- this could help Internet Radio stations significantly in the future. Yes, there aren't good multicast clients at the moment, but that is because there is little multicast to listen to, and no way of getting multicast to the end user. Lobby your ISP for multicast!

p.s. In case you think that I'm an idiot for not being able to configure IPv6 on Linux -- I'll tell you that I was kernel contributer in the pre-1.0 kernels.

Re:The installed base is hard to change...

[mgv]

p.s. In case you think that I'm an idiot for not being able to configure IPv6 on Linux -- I'll tell you that I was kernel contributer in the pre-1.0 kernels.

I know your pain. I was never up to being a kernel contributor, but I wrote a fair bit of code from assembly level (multiple CPU types) upwards.

I still can't get SAMBA to talk with a windows NT share. Its reassuring to know I'm not the only stupid one around.

Michael

Re:Too costly at this point

[ShmakDown]

Hardware implementation is most deffinatly the bottle neck that has to be dealt with, but some of the tunneling solutions that are out there now are not as bad as they might seem.

They basically just wrap an IPv4 wrapper around the IPv6 packet and send it back out across the net. A lot of network edge routers do similar types of things already, and many edge routers are doing IPv6 tunneling now.

Check out info about netBSD's IPv6 packages to see what solutions are already available and starting to become more wide spread.

It's a simple fix

[I_redwolf]

OS vendors just need to start supporting it and network/system administrators need to start implementing it. For instance Solaris comes with the option of enabling IPV6 but keeping continued functionality with IPV4, it allows for migration. Openbsd does this as well I believe and I've started to see some Linux distro's do the same. Now all we need to do is actually implement it. Alot of people seem to be afraid of IPV6 because of the hex but if you spend a month or two with it; it becomes easier. I recommend some solid reading on IPV4 as well as the IP in general.

IPv6 and IPv4 can live together

[wackybrit]

Some people have asked whether we can have both systems and 'switch' between them. Sure you can, but it's not worth it.

As far as I recall (been a long time since I studied this), IPv6 and IPv4 can actually live and work together on the same network.. without being independent.

That is, IPv6 can be used on the backbones and to connect the larger networks, but IPv4 can still be used at a more local level. Gateways can be established that will translate addresses and the benefits of having far more addresses available can be realized.

However, one problem with running both protocols and using a gateway is that the only benefit you get is having more addresses.. but since we're running out of IP addresses with v4, this is kinda important. A local v4 and backbone v6 solution wouldn't help solve local DOS problems, or allow us to use any of v6's advanced features.

But is an Internet wide upgrade to IPv6 really a viable thing to do? It'd be like converting the US to drive on the left side of the road overnight. Even if you did it state by state, you're gunna have major troubles at the state borders.. converting the Net over to IPv6 will be the same.

That said, there is a network called the 6Bone which you can join up to and actually play with IPv6 stuff from your existing IPv4 network. Go, and get your own IPv6 address today!

(Disclaimer: As I said, I studied IP way too long ago, so any updates, corrections or just plain disagreements with my post are welcomed, and indeed encouraged.)

Don't hold your breath

[MeowMeow Jones]

Most of the people I know haven't even upgraded to IPv5 yet!

Come on people, it's 2002!

IP6 might be the death of linux.

[Krapangor]

...unless the development speeds up fast.
My experiences with IP6 and Debian woody:

• inetd is fucked up accepts only connections to ::1, no other addresses supported which makes the box practically unreachable from outside
• netstat/route etc don't support ip6, only ifconfig
• if ip6 is supported then no or only crippled documentation existd

These are only a few issus. Unless these thing get fixed fast then FreeBSD will replace Linux at most professional environments.

America Doesn't Change Standards Easily

[puppetman]

Heck, you are the only first-world nation that doesn't use metric, and that's easy to figure out.

Yup, a ball and chain slowing down progress....

Re:America Doesn't Change Standards Easily

[Arandir]

The US hasn't switched to metric for a very simple reason: those of us living here, regardless of political affiliation, have a very strong individualistic streak. We don't just go change a system just because someone bigger than us tells us to. We spent a decade in the process of conversion and in the end we decided we didn't want the hassle.

The metric system is still taught in schools, still used in industry, and still available on every milk carton from New York to San Fransisco. But we prefer the English system. We're individualists and that's our choice. Just because it isn't your choice is completely irrelevant.

Oh, by the way, we've been using metric currency since day one, far sooner than most other countries did.

Re:America Doesn't Change Standards Easily

[Sri Lumpa]

We don't just go change a system just because someone bigger than us tells us to.

Nah, you just go and tell other ones to change their system because they are smaller than you.

Re:America Doesn't Change Standards Easily

[Zwack]

But we prefer the English system

Ahem...

It's not the English System

In the UK we call it "Imperial" and it's not the same as the simplified variant that you use in the US.

Want proof? One fluid ounce is the same in "American" and Imperial. How big is a pint?

Under the American system a pint is sixteen fluid ounces. Why? Because a pound has sixteen ounces.

Under the Imperial system a pint is twenty fluid ounces. Why? Because it does.

Sure it's less convenient to have to remember that a pint is 20 fluid ounces not sixteen, and that a pound is sixteen ounces not twenty...

But PLEASE don't blame the English for your screwy non-metric system.

By the way, does anyone know why America is one of only two countries that doesn't use international standard paper sizes? They make perfect sense to me (you can make the next size down by cutting the paper in half, and the next size up by sticking two sheets together.)

Z.

In two words: unsold inventory

[mangu]

With purchases of new hardware shrinking along with the economy, wouldn't these equipment makers be in a perfect position to benefit from adaptation of IPv6?

The problem is that shrinking sales has caused a huge amount of hardware to be stockpiled at Cisco warehouses. IIRC, last year they had over 5 Giga$ worth of accumulated unsold hardware. They need technology to stand still for a while, so they can sell part of that obsolete inventory.

Re:In two words: unsold inventory

[Cramer]

IPv6 capability within the routing world is trivial. There's not that much to change. It'll take a lot more memory, but very little re-coding.

However, IPv6 support beyond routing is a huge undertaking. Every single network aware program in existance will have to be rewritten to deal with addresses much larger than they currently are.

An interesting question

[wowbagger]

OK, I am about to say something that will make many of you who are knowledgable about IPV6 cringe, so take a deep breath and get over it now.

When IPV6 is deployed, how do I prevent the machines on the inside of my firewall from being routable?

Right now, my personal computer is on the inside of a NAT firewall. There is no way you can route a packet to it - go ahead, try to telnet to 10.200.120.4, I dare you.

Now, I know there are those who say NAT CONSIDERED HARMFUL, and I agree in the general case it does break the essential peer to peer nature of TCP/IP.

But what if I want to break it?

How well tested are the Linux kernel modules for firewalling IPv6? Can I still protect my internal machines from the slings and arrows of outragous 5|<197 |<!66!3Z?

Re:An interesting question

[vrmlguy]

Someone asked:

When IPV6 is deployed, how do I prevent the machines on the inside of my firewall from being routable?When IPV6 is deployed, how do I prevent the machines on the inside of my firewall from being routable?

The answer is to use a "Site-Local" address for any device that you don't want seen outside your site. From RFC 2373:

There are two types of local-use unicast addresses defined. These are Link-Local and Site-Local. The Link-Local is for use on a single link and the Site-Local is for use in a single site. Link-Local addresses have the following format: [...]

Link-Local addresses are designed to be used for addressing on a single link for purposes such as auto-address configuration, neighbor discovery, or when no routers are present.

Routers must not forward any packets with link-local source or destination addresses to other links.

Site-Local addresses have the following format: [...]

Site-Local addresses are designed to be used for addressing inside of a site without the need for a global prefix.

Routers must not forward any packets with site-local source or destination addresses outside of the site.

Re:An interesting question

[TrixX]

how do I prevent the machines on the inside of my firewall from being routable?

The same way you do that now: Using unroutable (i.e. private) addresses in your internal network instead of public ones.

I might not remember this correctly, but I think IPV6 had a large set of private addresses for use in internal networks.

Anyway, a firewall is always useful, because somebody at your USP could route to your internal network if you had forwarding enabled (which you have probably if you do NAT), and anyone at the internet can route you through source routing (although source routing can be disabled in Linux, and probably in any serious OS.)

Re:An interesting question

[Skapare]

There are plenty of useful functions for NATv6. See my post above.

Re:An interesting question

[scrytch]

When IPV6 is deployed, how do I prevent the machines on the inside of my firewall from being routable?

Tell your firewall to not route it. The only reason 10.0.0.0 and 192.168.0.0 (I don't remember the class C one) are non-routable is because every single hop has wired into it the knowledge that those aren't routable.

Plus, I have to imagine there are nonroutable IP6 blocks as well...

Re:An interesting question

[docwhat]

I'm confused. Are you saying you have a NAT+Firewall and you are asking how you would set things up with IPv6?

Answer: You keep the firewall and toss NAT.

Or are you asking how good the Linux IPv6 filewall stuff is? I don't know about that...I assume it works. But I don't know.

IPv6

[SkewlD00d]

Interesting moot point... it seems that 3G licensees were going to require IPv6. Search for "IPv6" on various corporate and info sites:

• 3com - no listings (no support???)
  
• Cisco Systems
  
• HP - no listings in network equipment
  
• Juniper Networks - OS support
  
• Linux IPv6 HOWTO
  
• Lucent - interesting
  
• NetBSD IPv6 docs
  
• Windows XP - Installing IPv6
  
• RFC 2492 - "IPv6 Over ATM Networks"
  

This long annoying sentence here to get around an annoying slashcode bug, because it can't count.

ISP's are biggest holdbacks...

[kbonin]

They have tons of money invested in hardware they don't want to replace. Sticking to IPv4 makes it easy to keep user bases behind short-lease DHCP, which helps to keep the average user from mounting a public server that'll eat bandwidth the ISP doesn't want to provide.

Also a few Cisco points: 1) While some routers do support IPv6, the cheaper ones don't, and a decent percentage of older high end routers have routing algs implimented in semi-custom silicon - not software upgradable! 2) The enterprise network management software is lagging behind in IPv6 support last I heard (I used to work there), not much demand.

Re:Tech Issues

[dieman]

Because of the larger space, theres no reason AOL can't be aggragated into *one* bgp announce and be allocated *one* block that they will ever need, forever.

Fear that!

Re:Too costly at this point

[Webmoth]

IPv6 is backward compatible. I believe what you meant to say is that IPv4 is not forward compatible: in order to make IPv6 work, all routers have to be IPv6 compatible.

What's standing in the way is that a large number of routers are not upgradeable to IPv6, and the cost of replacing would be greater than the return on investment.

Let's say that you replace an existing IPv4 router with an IPv6 router costing 3000USD. Labor and downtime costs during the replacement will probably range a couple thousand more USD. So you're looking at 5000USD. In order to justify the investment, you must realize at least 5000USD (plus interest) profit from sales of IPv6 bandwidth over the life of the router. You won't sell IPv6 bandwidth until there is a definite need; with all the various "patches" out there (DHCP, NAT, Dynamic DNS, etc.) it may be quite a while before IPv6 is needed. You may find that IPv6 is not needed until AFTER the MTBF of the router!

Bandwidth providers are waiting for the need. However, I think it would be wise when installing new or replacement hardware, to replace it with IPv6 capable products.

If you can't do that, it's a waste of money, and wasting money is no way to do business.

On a side note, there are only 16^12 MAC addresses available. When will we run out of these? ;-)

Completely off-topic: I've seen only one forward-compatible technology: WordPerfect 6/7/8/9/10/etc. Save something in WP10, you'll be able to read it in WP6 (minus some formatting, but you'll be able to read it, with most of the formatting intact). I may be wrong, but I believe this was designed when Borland owned the product. Kudos to Borland! "Backwards compatible" implies that current technology is compatible with legacy formats and protocols. "Forwards compatible" implies that current technology will be compatible with future formats and protocols, is designed to be expandable, and designed to tolerate unknown features.

How to transition?

[A nonymous Coward]

Suppose I take my home network (2 computers + 1 firewall), all running some form of highly modded Slackware, and switch the internal local net to IPv6 while leaving the connection from the firewall out as IPv4. Thus the 2 computers would be completely IPv6 while the firewall would have one IPv6 nic and one IPv4 nic. I have to change all dotted quad network addresses (such as in /etc/hosts); what else is there to do? Will existing software go along with the change without recompiling? Or even with a simple recompile?

I bet there's some FAQ somewhere that someone will find using Googole. AIA

Re:How to transition?

[Wesley Felter]

You'd also have to make sure that every network app running on those inside computers supports IPv6, and you'd need some sort of protocol translator on the gateway. For those reasons, I wouldn't suggest that scenario.

I would suggest running both v4 and v6 on the inside machines and making the gateway into a 6to4 border router.

Re:How to transition?

[An IPv6 obsessed guy]

On your gateway, run a 6to4 translator (eg NATPT.) To handle DNS, run the Trick-or-Treat daemon, which takes IPv4-only addresses and puts them into v6 form (NATPT then translates it back at the border.)

Now port all your apps to support IPv6. This involves changing IPv4 specific function calls to their IPv6 equivalent. For folks like me, who only use the web & ssh, this isn't even an issue, since OpenSSH and Mozilla support IPv6.

Viola, you're running IPv6 and nothing else.

Breathing life into IPv4?

[Alioth]

One thing I've noticed is that there's an awful lot of organizations (well, certainly a big handful) which have entire allocations of the old Class As. But virtually all their IP address space is hidden and non-public. People like the United States Postal Service (56.0.0.0 - 56.255.255.255), IBM (9.0.0.0 - 9.255.255.255). These organizations have barely a handful of publically-visible IP addresses, but these massive blocks in the IPv4 space. The USPS has 24 million IP addresses in their block, but probably less actually visible than a small Midwestern mom-and-pop ISP.

Why aren't these organizations told that they have, say, 2 years to move to a private 10.x.x.x network, thus freeing many millions of IPv4 addresses, instead of forcing small organizations to come up with huge justifications for a very small number of addresses?

I've tried IPv6 with Windows 2000...

[chrysalis]

A major showstopper may be Windows.

Let's see. To be widely deployed on WAN networks, IPv6 should first be widely deployed on local LANs.

It works very well on Unix systems. My little personal network has a bunch of OpenBSD and Linux boxes, 100% IPv6, and everything works like a charm.

But what about Windows?

I tried it with Windows 2000. Because the OS doesn't support IPv6 natively, I had to download a patch (and it's not very easy to find, I can't remember the exact URL, the link was posted on a ML a while ago) .

Before the patch applied I had a big fat warning "Disclaimer: this is very alpha software, your OS can become extremely unstable. Don't call the Microsoft technical support any more after that, we won't answer" (the words were different, but it was the meaning) .

And indeed. The system went very unstable, even for IPv4 requests. IE worked. *some* command-line tools worked. But third party packages like Mirc, CuteFTP and Opera crashed with no further warning.

It looks like there's no effort in the Windows world to provide IPv6-enabled software. This is a major showstopper.

Re:Get some from the U.S.

[Bullschmidt]

The problem with at least some of these, is that these universities have so many because they were some of the institutions which started it all. Its tough to take something away from someone when they played a large part in its creation.

Cisco is the Microsoft of routers

[mangu]

Whatever Cisco decides, it will be the consensus in the networking world. What they know that we don't is that they are in deep financial trouble. Their worldwide employee layoff figures last year were in the five-figures range. Their troubles started when they implemented a sophisticated market analysis system that predicted increasing router sales throughout 2001. That software was so "advanced" that they refused to believe their sales people when they started telling management that they couldn't possibly sell so many routers.

What About IPv6?

[t_allardyce]

How Long Until Widespread Deployment?

About 15 years.

After the introduction of the SSSCA in 2003, Microsoft dominated the US OS market. While other countries switched to IPv6, America was forced to use the proprietary protocal built into windows (thanks to auto-updates) which included advanced DRM, IP tracking and P2P restrictions - as a standard client, your computer could only connect to a 'server' i.e a Windows machine running Windows Server Edition with a valid federal license. The internet was effectively split in 2 - USA, and the rest of the world (troll: this didn't matter as most US citizens didn't know about the 'rest of the world' lol :)

It wasn't until the great Microsoft witch hunt of 2017, when 4000 Microsoft employees where burnt at the stake after the SSSCA was lifted (well, not lifted per say, actually, someone just blew-up congress)

You can use IPv6 today!

[Wesley Felter]

Even if your ISP doesn't support IPv6, you can use 6to4 to start using IPv6 today. It's much easier and more efficient than the 6bone. Since IPv6 allows a host to have multiple addresses, the eventual transition from 6to4 to native IPv6 will be seamless.

They already have.

[qaggaz]

Cisco released IPv6 IOS images back in June with IOS 12.2(2)T. Note that this was the first commerical release, there was a earlier EFT release about for quite some time that served as a beta. The major features are there: IPv6 routing, support for stateless autoconfig, IPv6 address family support in MBGP, support for RIPng. No other routing protocols yet.

You can check out Cisco's IPv6 page for more information.

Juniper also has IPv6 available, here how to configure IPv6 on JUNOS 5.1.

Never?

[Broccolist]

I'm going out on a limb here, but has anyone considered that IPv6 may never get widespread acceptance?

From the point of view of any individual organization, there are no reasons to switch to IPv6 right now. First movers receive no benefits at all: in fact, it only makes communicating with the rest of the (currently IPv4) internet more difficult. Moreover, I imagine that many businesses large enough to have an impact already have a large IPv4 address block, and have a vested interest in discouraging others from making the switch:

1. There is no reason for them to pay for new routers
2. A crowded IPv4 internet might allow them to loan out some of their in-demand addresses for extra profit.

The various hacks available for IPv4 do the job. I can easily imagine a scenario where Cisco doesn't push IPv6 routers hard enough in the future, and people invest more and more in NATs and so forth, making a global switch harder and harder as time goes on.

The fundamental problem is that IPv6 doesn't provide any short-term killer benefits, and that's what's necessary for an evolution to take place. My prediction (though predicting acceptance of technologies is always risky, so I may well turn out to be wrong) is that we will still be using an IPv4 internet in a decade.

Re:"IPv6-ville"

[Derkec]

Why nuke power instead of solar / wind or even the highest tech of them all: covection tower power?

Try freenet6.net

[MavEtJu]

If you are interested in playing with IPv6, try to get a tunnel via www.freenet6.net.

They're supporting devices running *BSD, Linux, Win*, Solaris, HP-UX and Cisco IOS.

FWIW...

[luge]

Duke has had IPv6 available on resnet since at least some time during the '99-'00 academic year, so at least two and nearly three years. You just had to know who to ask to get an IP address :)

Re:FWIW...

[Moridineas]

My mistake :)

and geez is the 20 seconds rule annoying!

Scott

NAT provides convenience, not security

[pHDNgell]

While it may sound neat to say, ``go ahead, try to telnet to 10.200.120.4,'' it doesn't exactly work that way.

Does this machine on 10.200.120.4 have the ability to make direct outbound connections? Assuming yes, does you realize that the only difference between an inbound connection and an outbound connection is who sent the first packet?

Many people tend to believe that the *only* security risk they have to worry about is inbound SYN packets, so they base their entire security policy on stopping bad inbound packets. The last two sites I broke into, I did so by tricking a machine to come to me. Just for humor, here are the two scenarios:

The first one was quite a while ago, and I did it at contract. A co-worker found a potential hole in a CGI, but nobody took it seriously. By sending the right data through the CGI, I found that I could make it execute arbitrary commands. First, I did some basic stuff (id; ls -lR /; etc...) and had it output the mail to me (couldn't see the output from the CGI). I figured out the web server user had a shell and a writable home directory, and the machine had ssh (client and server installed). I generated a private key and had it mail me the public version of that key, then I added it to my authorized_keys and installed my public key in the web server's authorized_keys. Then I had the web server user ssh to my host with remote port forwarding back into the web server's 22. ssh -p 2222 localhost and I'm sitting in a shell on the web server (192.168.something).

The next time I saw something like this, it was out in the wild. There was a web server that was running a CGI that *seemed* like it was probably just handing the input over to a command, so I gave it a shot. This time, the web server didn't have a usable home directory, so the ssh thing was out, but it did have X installed, so I fired up a VNC server, opened it to the world and opened an xterm up in it. Before too long, I had an entire X desktop running on some guy's web server. I sent the local admin an E-mail (through pine) letting him know what was wrong and recommending he fix it before someone meaner than I am comes along.

Anyway, point of the story. Having an unroutable IP address is good internet security as long as you keep it unrouted. Once you give the thing direct internet access, the unroutability of it becomes much less relevant.

Re:NAT provides convenience, not security

[wowbagger]

Of course only blocking incoming connections is only a part of a security policy.

However, both the examples you gave in your message required you to be able to connect to the target machine via HTTP and issue an HTTP GET request - therefor you had inbound connectivity to the target, just not inbound connectivity to J. Random Port.

There is NO inbound port available to you. Not 80, not 22, not 25, nothing. The only inbound ports would be when I am FTPing down a file, if I am not running passive mode. However, since the firewall only allows traffic from the FTP server, you would either have to spoof that (and then all you would do is corrupt the file I am downloading) or hack the FTP server (same problem).

And as to the other people who pointed out that I could use a site-local address: Of course, what do you think 10.200.120.4 is? However, NAT for IPv4 is very well tested, so my "unroutable" 10.x.x.x address is still able to get to /. (as this very post bears witness to). Would my IPv6 site-local address be able to do the same - in other words, is the state of NAT for IPv6 anywhere near IPv4? Considering the common opinion is that NAT is unneeded in IPv6, I very much doubt it.

The great thing about my workstation being unroutable is that, should I be stupid enough to get a Trojan that announces itself to the 'net and says "I am at $address $port, come abuse me", if $address is not routable, this does very little good for the script kiddie - even if the system reports a traceroute so that he can follow it back, he STILL cannot route a packet to it.

(now, this does not stop the Trojan from connecting to an [icq|http|SOAP|...] server and pulling its commands down, but as I stated at the first of this post, no one aspect of securing a system is sufficient - security is a journey, not a destination).

Re:NAT provides convenience, not security

[cookd]

That doesn't change what the guy is saying. NAT prevents another computer from initiating a connection to the internal network, but it doesn't prevent you from being hacked. A clever hacker can hijack existing connections, or convince you to open connections that aren't friendly.

For example: you browse to www.ima.hacker.net. The page has code to exploit a browser vulerability, and the exploit code initiates a connection back to www.ima.hacker.net.

Another problem is connection hijacking -- a hacker can send extra packets to a firewall that actually get through because they are marked as being from the same port and address as those of a real connection. This is especially easy if the hacker is able to sniff packets en route.

Yes, being behind a NAT does reduce the risk of attacks: you probably only have to secure your client apps, not your server apps. But clients are vulnerable, too.

Overall, IPv6 will be far more resistant to hacking. The designers had the wisdom of many years of IPv4 problems and security flaws to influence the design. Now it is much harder to spoof a packet. Now you can't sniff packet ID numbers. Any advantage that you are currently attributing to NAT can be gotten with a firewall, and much more reliably.

Can't wait can't wait can't wait.

Re:NAT provides convenience, not security

[Kynde]

Before too long, I had an entire X desktop running on some guy's web server. I sent the local admin an E-mail (through pine) letting him know what was wrong and recommending he fix it before someone meaner than I am comes along.

For the record, I know people that have gotten sued for such actions. Actually for a lot less. So you really should be carefull out there. It may be fun to find security holes, but if you want to be on the safe side, you should just stick to shutting the fsck up about it afterwards. Or at the least, always let the otherside know about the security holes anonymously, in which case you gotta really cover your tracks.

If at any point of finding out the existence of a security hole or otherwise you've might have had the intent of an electronic counterpart of breaking and entering (e.g. scanning, sending crap to cgi bins, etc) you may find yourself face to face with their lawyers. If the company executives are morons, they can and will sue you regardless of wether you actually did any damage or such.

An analogy is lock picking an 1800-century lock to gain an entrance to a bank and then without stealing anything letting the bank manager know that they're lock is obsolete. He could sue you. And in electronic world I know of such cases.

Re:NAT provides convenience, not security

[Havokmon]

I think we all understand the potential insecurities of NAT, but what's with the current attitude?

It's like people are saying, "NAT is just security by obscurity, so just don't use it."

I think THAT'S the message that needs to be abolished. Anything that can be added to the current 'state' of the network to increase security, should be added.

Considering NAT IMMEDATELY prevents stupid errors (such as my old ISP's Accountant sharing her C drive with the world), it shouldn't be discounted because a bunch of techies can come up with convoluted ways around it.

Re:NAT provides convenience, not security

[docwhat]

I think the message isn't it shouldn't be abolished. NAT is useful. It's people thinking that NAT is some form of security that should be abolished.

I'm unclear what the top parent poster's problem is? Why do you need NAT if you are given your own set of IPs to work with? You will still have a firewall at the front (where the NAT+Firewall *was*), right?

Re:NAT provides convenience, not security

[Havokmon]

I think it *IS* a form of security, it's an easy form of security. Just like dead-bolts.

Just because *you* know a way around it, doesn't mean it can't/shouldn't be used.

I'll start using IPv6...

[ewieling]

I'll start using IPv6 when the backbones start using IPv6 and I can get IPv6 addresses from my ISP.

Re:I'll start using IPv6...

[Skapare]

Or we can put together our own IPv6 network with some tunnels to get started, just like 6bone but without the hassle.

Re:no, everyone.

[sigwinch]

sk your dialup isp for a static IP account, and they might tell you about another company who will give it to you for $80/month.

That's to squeeze money out of businesses who absolutely need static IPs. At colocation facilities, who are already charging $100+/month, extra static IPs are only a couple of bucks a month each.

I can't wait...

[jbf]

...for IPv6 because...

• ...I want my IP headers be twice as long
  
• ...I want to go from 50% header overhead in Netmeeting to 75% header overhead
  
• ...I want to include a 16-bit field (Flow ID) in my header that no-one has yet figured out how to use
  
• ...I feel the need to address every atom on the face of the universe, and then some
  
• ...I love IP addresses like 1.0.0.0.0.0.0.0.0.0.0.0.0.0.0.0
  
• ...I like the idea that we'll all have to buy new copies of embedded hardware that are currently IPv4-only
  

IPv6 doesn't; change priority / QoS that much

[billstewart]

You've pointed out that IPv4 has DiffServ / TOS bits, but backbone router ISPs don't universally support them. Adopting IPv6 won't change that - it's a policy issue on the part of the ISP. It will make some kinds of features easier to implement, such as giving people private-line-like performance across pre-defined parts of a single carrier's internet, but they could do that today if they wanted, if they could figure out what to charge and how to manage it. Having more address bits makes it easier to design entertaining features, e.g. a chunk of your address space that uses router filters to create private subnets, but the critical issues are ISP policy.

There are ISPs starting to deploy this stuff, primarily driven by the Voice-Over-IP market. For the most part, what matters isn't prioritization on their 10Gbps backbone, where there's plenty of room for everybody - it's prioritization on the T1 line to your building, or in the oversubscribed DSL network to your house. One of the real issues becomes policy at the interfaces between ISPs - Little Local ISPs care about this a lot, but most of the Tier 1 players have the view that "Why should I provide special support for the connection between me and my competitors - I'd rather sell you the prioritized connections on your whole network where I can manage it all (and get all the money, and provide realistic guarantees of service quality, and get all the money)."

Re:IPv6 doesn't; change priority / QoS that much

[Metrol]

You've pointed out that IPv4 has DiffServ / TOS bits, but backbone router ISPs don't universally support them.

If those routers are IPv6 compliant the MUST support QoS bits. The reason they don't today is because it's not specifically a part of the IPv4 spec.

Adopting IPv6 won't change that - it's a policy issue on the part of the ISP.

No, it's not. The backbone routers will be making these decisions. The ISP's will be able to purchase the right to resell QoS bits to customers. This isn't some wiz bangy feature you can add. It's built directly into the protocol itself to do this! Heck, that's specifically why those bits are in there.

It will make some kinds of features easier to implement, such as giving people...

First off, this has nothing at all to do with giving anyone anything. This is a scheme to prioritize packets on a highest bidder basis. Voice over IP is a red herring tossed out there to get folks into seeing how maybe their web traffic really isn't as important as voice traffic.

Today the telcos provide one very large portion of the Internet backbone. Do you really think their interests lay in providing cheaper voice service? It's just not realistic. What's more, it's not needed! Voice traffic across phone lines is already heavily digitized to make it so efficient as to require a tiny fraction of the employees needed just 10 years ago.

---

As you can probably tell, IPv6 scares the hell out of me. When corporations can prioritize the packets coming off their servers above those of individually owned machines, kiss anything that once may have been considered beautiful about the Internet goodbye. Nothing left but a flashing billboard on the browser highway.

Re:Why "first" world?

[Julian352]

The US is not the only country in the world. As can be seen on many different websites there are other countries that use it, they are just much less globally noticed than US.

In 1990, there were only three nations that had not converted to metric: Burma, Liberia, and the United States. (source: here)

It is however in US best interest to convert, as many of the other countries require all products to be manufactured to the metric system. If you would go to your local grocery store, you'd notice that some products are already done that way, especially if they have global nature. A good example of that would be the 2 Liter bottle of soda that is sold at the store. As that item is sold both in US and abroad, the manufacturing costs are lower if same measurement system is used. That's why the Liter is used. (to conform to requirements of other countries)

What will that do to my LinkSys router/firewall?

[crovira]

I imagine I'd have to upgrade the firmware again.

While my Linux box is configurable and my OS X box is probably configurable, I've got two OS9 boxes that I'll have to wait on Apple to convert.

But I agree, IPv6 is the way to go.

There will be NATv6

[Skapare]

There will be NATv6 whether you like it or not, and regardless of what the RFCs say about it. There are legitimate uses for it. I may want to have the same hostname for a variety of different services, but put those services on different machines behind the firewall. There's a form of NAT for that. I may want to load balance 1000 servers to one name (which DNS will limit to just a few IPs at most ... and don't forget that AAAA records take more space out of the response packet than A records do). I may want to hide my internal infrastructure and make everything appear to be right at the border. And perhaps I just don't want to upgrade some server to IPv6, preferring to leave it at IPv4, and let the NAT present IPv6 to the world while my intranet sees it as IPv4.

Since we already have mastered the logic needed in a variety of forms of network address translation, IPv6 is just a matter of some code changes to accomodate the larger IP address.

Re:There will be NATv6

[Skapare]

But you don't need them. Everything can be handled right in the NAT (if programmed to do it) without anycast.

Re:There will be NATv6

[Skapare]

Doing NAT is still cheaper than upgrading everything to IPv6. And since NAT to services doesn't need any DNS tricks, it is easiest to deploy cheaply. Upgrading a server involves more time and more disruptions. Dropping a NAT in between is so much easier. Remember, the intent in this is not to upgrade the network to IPv6; it's to be able to serve connections from clients with only IPv6 addressess where there is no client side NAT (which requires the DNS tricks if the client software cannot address the massive IPv6 space).

Is the number of IP addresses in use increasing?

[Animats]

The number of domains registered peaked a while back, and I think the number of Internet users dropped a bit last quarter. Is the number of IP addresses in use still increasing, or has that peaked, too.

The cycle keeping IPv4 right where it is

[Guppy06]

The way I see it, there's a four phase cycle keeping IPv4 the standard for the internet for a long time to come.

1.) ISPs want to charge more for sharing a connection and a smaller address space gives ISPs justification to charge more for corporate users than home users. They already heavily frown upon the use of NAT (unless you pay more for them to set up your LAN for you). So why don't the ISP's just separate the concepts of bandwitdth and addressing?

2.) The backbone is overtaxed as it is. Currently the home user's connection speed is limited more by intermediate links than by their connection, even if the user is just using a 33.6 modem. A small address space provides an easy method of limiting bandwidth use. So why don't they just upgrade the backbone?

3.) IP address space is the primary driving factor in connection costs, more so than bandwidth. Most tier 1's more or less own their address blocks and stand to make money hand over fist as the price of using a single address skyrockets. If a tier 1 wants to make more money, it makes better economic sense to buy more address space than to put in faster connections. So why not jump to IPv6 to increase the address space by an order of magnitude squared so the big guys can focus on the bandwidth trouble? Tier 1 folks will make money no matter what, right?

4.) A larger address space opens up the ISP industry to small competitors. While most ISPs are owned or operated by large corporations that can afford the pricey IPv4 addresses, IPv6 stands to give every man, woman and child on the planet a bigger address space than many tier 1's currently have in IPv4. The low-level ISP scene under IPv6 could very well look a lot like the BBS/internet scene of ten years ago. Not to mention all the private little portals that could end up competing with MSN and Yahoo (with or without a DNS name). But still, the little guys could probably make a stab at making that happen with IPv4, using NAT to drive down the cost of a small IP address block. Why don't they do that?

Lather, rinse, repeat.

128 bit

[ZigMonty]

It's 128 Bit. You need to double your number a few more times. I get about 3.4*10^38 unique addresses. Which is quite a bit more.

Re:128 bit

[GigsVT]

And we only needed 48 bits in the absolute worst case. And I'd be able to remember my address if it was xxx.xxx.xxx.xxx.xxx.xxx too.

Re:Still paying for IPv4 deployment

[Skapare]

The cost to "upgrade" to IPv6 can be simplified by having IPv6 be translated at or near the border to IPv4 private addresses on the inside. No OS upgrade needed. No application upgrade needed. Just make sure your border routers can do IPv6, add the NAT, and you're live.

IPv6 Header

[germanbirdman]

[Bandwidth used up by bigger headers]

Actually, it doesn't really make that much difference.

An IPv4 header is actually quite difficult to process for hardware routers because it can have a length of anywhere between 20 and 60 bytes.

An IPv6 header on the other hand consists of a main header with a fixed length of 40 bytes and possible extension headers which do not need to be processed on all systems.
The 40 bytes of the IPv6 main header includes the 128 bit source and destination addresses.

The IPv6 headers are actually quite efficient and are designed in such a way that they can be easily processed by hardware.

So no, there will not be a BIG increase in bandwidth because of the headers.

Re:Moron - Actually YOU don't know what...

[autocracy]

OK, just because you gave the challenge - IPv5 was the development version of IPv6... moron :)

Re:IPv6 is easier than you think

[Skapare]

Just put 6to4 capability in, or right next to, your border router. Put some IPv4 private IPs in for the inside pool, and away it goes. It should be easy (but I don't know if implementors are that smart) to take a /104 chunk of your IPv6 space and map it 1-to-1 to the whole 10/8 space by keeping the low 24 bits the same. That should give you plenty of time to transition your inside servers, and all your access customers (if your an ISP) or all your offices and cubicles (if your a business) to working on IPv6.

I dont think MS will have a problem

[xiaix]

As other posters have pointed out, BSD has it...microsoft probably just has not come up with a proper 'embrace and extend' logistic for it... (yet)

OT but needs to be said

[ZigMonty]

The British hecklers in the audience may wish to remember that they are the only first-world nation without a written constitution.

And yet which country's people are currently more at risk of loosing their freedom (DMCA, terrorist pirates, etc)? You're acting like not having a constitution (assuming it's true) is equivalent to being lawless. It's still illegal to murder someone in Britain and, last time I checked, there was freedom of the presses by law. A constitution is just law that's hard to change. You could argue that Britain's system is more flexible and adaptable to our changing world.

Also, there's a reason most scientists in America use the metric system. Guess what it is.

When Microsoft supports it.

[duffbeer703]

When 75% of active Microsoft Windows hosts fully support IP v6 out of the box, IPv6 will begin to appear.

Otherwise, forget about it.

CLNP/GOSIP politics were a different problem

[billstewart]

Background: The OSI protocol suites had a wide variety of choices of protocols at different layers in the stack. CLNP (ConnectionLess Network Protocol) was roughly equivalent to IP. GOSIP was the late-1980s US Government OSI Protocol stack, a specific set of protocols from the OSI suites covering Layers 1-7, and looked more like the TCP/IP world than the ISDN/X.25/EuroTelcoBureaucrat world. Padlipsky's "The Elements of Networking Syle" (ISBN 0-13-268111-0) is the classic critique.

The important differences between the OSI protocol stack people and the TCP/IP people weren't at the transport layer - they were mainly the application layer and the availability of working implementations on Unix. Multi-Protocol Routers were becoming available at the time, driven by the widespread use of IPX, the Not-Dead-Yet-ness of Appletalk and XNS, the Routing?-What's-That? bridginess of DEC LAT, and the Hadn't-Taken-Over-The-World-Quite-Yet-ness of IP, so there were routers with CLNP available at costs not substantially different from other multi-protocol routers that also did IP. While the TP4/CLNP stack wasn't much clumsier than TCP/IP, the set of application services was - X.400 was MUCH heavier-weight than SMTP, and FTAM was somewhat more bureaucratic than FTP, VT was more general than Telnet, and 4.2BSD UNIX came with TCP/IP and sockets and such, with well-written relatively-open code that was usable on Vaxen and ported to Suns and other popular computers. If you wanted to write stuff, you could just write stuff.

First, second, and third world explained

[yerricde]

Gah. You're not "first world". America is the "new world" (aka 2nd).

No. The first world was countries that fought on the U.S. side in the cold war (U.S., Canada, western Europe, etc). The second world was the Soviet Bloc (no relation to Soviet blocks). Countries too small for either superpower (USSA or USSR) to notice came to be known collectively as the third world; after the cold war ended, "third world" continued to refer to developing countries.

Poll: Which world will achieve 50% adoption of IPv6 first?

Re:First, second, and third world explained

[puppetman]

No, the origins of the terms First, Second and Third World actually originated in 1952, created by a French guy named Alfred Sauvy, and relate more to pre-industrial France than the world after WW2.

He actually called them estates, and drew a parallel to the demographics of France before and during the French Revolution. The first and second estate are the nobles and priests. The third estate are the unwashed masses that want to be something, and are exploited by the first and second estate.

Yes, it was drafted during the cold war (though the cold war officially started in the late 40's, and this was 1952). But the concept of have-and-have-nots has been around since antiquity.

Anyway, it's a crappy analogy. Now we have agrarian, industrial, and information based economies (and everything in-between). We also have democracies, autocracies, monarchies, totalitarian-states, military-based govenments and dictatorships. Way too complicated to sum up in three broad, misunderstood categories.

Regardless, the French are crazy.

Respond, don't moderate

[jbf]

Interesting that none of the major issues with IPv6 are addressed by an idiot moderator who doesn't know what to do with his/her moderation points. Guys, when you get moderation points, only moderate the stories you're an expert on. Even the people who go to the "universal adoption of IPv6" Scotch BOFs at the IETF would agree with many of these points. In particular, Steve Deering, the author of SIP, which later merged with a number of other things to become IPv6, would agree that 128 bit IP addresses is overkill, and SIP had only 64 bit addresses originally.

CmdrTaco: if you let moderators do stupid things like this (and yes, there are more stupid moderators than smart ones), you're going to alienate all your clueful posters.

Re:Respond, don't moderate

[jbf]

However, I wouldn't mark this particular moderator down. Your post was a list of complaints about what other people are doing to solve a real problem with no solution of your own to offer or any reason why its not a problem. Personally I wouldn't mark it as a troll but I think it's borderline.

Does criticizing a solution mean that I have to try to solve it myself? If I said "I want to go to the moon, I'll buy a couple D rockets from the hobby rocket shop," would someone be mistaken to say "that'll never work" without providing a solution? Would a "that can't work, moron" response be a troll, or even justified as one?

I'm just bothered by the general "rah rah rah IPv6" crowd, plus the "it'll never happen because of evil M$" crowd. There are real, technical issues with IPv6. There are real, nice benefits of IPv6. One of the places the IETF really messed up, though, is the increase of address space. Sure, it's not a big deal on modems, since compression will help a lot (IPv6 headers compress much better than IPv4 headers).

But the big thing you're leaving out of the picture here is the mobile/wireless explosion that has happened of late. Yes, it's great that my UMTS phone will have IPv6 and actually be addressable. Hoorah. But all those extra bits have to go over the (sparse and expensive) air.

Worse yet is the problem with ad hoc networks. People are putting router IDs (RIDs) in their routing protocols, so they can squeeze a 128-bit header into 32 bits. Problem is, you need to pick unique router IDs, and you need to advertise the correct associations. This is a major pain. You can't just waste the bits, because they're going over potentially slow links, every bit transmitted costs battery power at all receiving nodes, and increases congestion in that area of the network.

A major advantage of 128 bit addresses is that it makes things like SUCV (statically unique cryptographically verifiable) addresses possible. But that's only necessary (at least now) because the IPsec WG screwed over Mobile IP's IPv6 authentication scheme (with certificates for each address).

If you accept that the earth's population will stay under 2^40 (1099 billion) for the forseeable future, then each person will have 2^24 addresses. Even allowing for inefficiencies due to things like CIDR, each person will still have over 16000 addresses (16 million with perfect efficiency). I just don't think we'd ever run out. Also, considering the address allocation scheme of IPv6, it's not clear that a better allocation scheme for 64 bit SIP wouldn't last longer. 64 bit interface indicies? Please. I'm not going to have 2^64 interfaces anytime soon.

See, this is what you should have put in the original post.

Why, do I need to quote Steve Deering to carry enough weight to criticize IPv6? I pointed out in my original posts the disadvantages of the IPv6 addressing scheme.

Re:Respond, don't moderate

[GigsVT]

I'm glad someone else is sane. 128 bits is way way way too many. 64 or 48 bits are enough to carry us well into the future.

AOL's Too big for 10.x.x.x :-) China/India too.

[billstewart]

If all of those 33 million users were connected at once, they wouldn't be able to have unique IP addresses in the 10.x.x.x address space, which only holds 16 million :-) Fortunately, they probably don't have that many at once, and even if they did, they almost certainly do some sort of tiered connectivity, either with NAT or proxies, that keeps them from needing whole-internal-network-routable IP addresses. I don't know if they do it regionally, or per dial POP (something like 1000-10000 POPs, which could contain all the address space for their local users as well as having some wide-area address space for the whole-company-visible or Internet-visible parts), or by not having more than 16 million modems (the easy approach for most of us :-).

For China and India and in general the rest of the world, the choices are either to get on the stick and do IPv6, or else to use some other tiered-local-addresses-proxy-NAT system. By then it wouldn't be surprising if cheap mobile devices (phones or otherwise) were the big driver, and IPv6 means you just don't need to fix the addressing problem again.

Re:The era of the 4 to 6 Gateway

[Skapare]

if you mean "4 inside, 6 outside" then it has some limitations. If you're on a 4-only box inside and want to connect somewhere, it has to have an IPv4 address, or you have to have some IPv4 address mapped to the IPv6 address with corresponding DNS change (I hear this is what the BSD folks are doing). Doing servers this way is easier as the client (outside) is connecting to a specific IPv6 address, and the NAT can translate that easy enough to (inside) IPv4 (no DNS juggling needed).

Such network address translation should at least prevent any delays in upgrading servers from delaying IPv6 deployment to the backbone. Clients will seriously need to be upgraded, and if Microsoft drags their feet, that can set things back really bad. But we also need solid IPv6 router code for the backbone, and I gather that Cisco is not moving very fast on getting it widely implemented solidly. Maybe when the economy picks up they might be able to (if they see the demand for it).

Re:IPv4 doesn't really seem that close ..

[Skapare]

If ARIN (and APNIC and RIPE) would assign portable space in smaller pieces and make some kind of rules requiring ISPs to route them when there is one prefix in an AS, regardless of size, then I suspect ISPs will find it a lot easier to do NAT. Right now a small ISP doing "the right thing" and deploying NAT for all business customers, instead of giving them each a /29 (and thus using only 1/8 the IP space), is delaying their ability to reach the holy grail of ISP-dom: portable IP space and an ASN. And while this is happening, larger ISPs are still flooding BGP with hundreds or even thousands of prefixes for gobs of discontiguous IP space.

Re:The real problem

[Skapare]

With IPv6, address space assigned to large ISPs will be a lot better aggregated. Of course that won't stop them from breaking it up into a lot of prefixes, but hopefully that will only be for their really sub-autonomous networks. Lots of big ISPs were given some big chunks like /16, but they got those many times. So that means many prefixes announced even if they could aggregate them if adjacent. At least with IPv6 they can be given all the space they will need for 100+ years right now.

Routing in IPv6 is also different. The low 64 bits as I understand it won't play any part as that is pretty much going to a single LAN, so the routing announcements shouldn't need any more than the high 64 bits, and maybe even just the high 32 bits. Here's the list of RFCs that match string search for "ipv6" and "route". I haven't actually read them, so maybe you can easily find where I'm all wrong.

rfc1752|rfc1809|rfc1825|rfc1883|rfc1884|rfc1887|rf c1888|rfc1933|rfc1970|rfc1981|rfc1999|rfc2000|rfc2 019|rfc2080|rfc2101|rfc2185|rfc2199|rfc2200|rfc229 2|rfc2300|rfc2353|rfc2373|rfc2400|rfc2401|rfc2460| rfc2461|rfc2465|rfc2466|rfc2473|rfc2491|rfc2492|rf c2500|rfc2526|rfc2529|rfc2545|rfc2546|rfc2590|rfc2 600|rfc2626|rfc2700|rfc2710|rfc2711|rfc2740|rfc274 5|rfc2746|rfc2765|rfc2766|rfc2767|rfc2772|rfc2799| rfc2800|rfc2874|rfc2884|rfc2893|rfc2894|rfc2899|rf c2900|rfc2956|rfc2983|rfc3000|rfc3002|rfc3053|rfc3 056|rfc3068|rfc3089|rfc3111|rfc3132|rfc3162|rfc317 5|rfc3178|rfc3234

IPv6 :: OSI

[igb]

The problem IPv6 has, confirmed by its enthusiastic reception by the EU, is that is
the OSI of the 21st century (following on from
ATM, the OSI of the 1990s). IPv6 solves a
problem of 1992 --- proliferation of subnets,
exhaustion of v4 space --- while other, incremental, changes did the job just as well.
NAT and DHCP mean that huge ISPs don't need
huge blocks, and the falling price of RAM means
that large routing tables just aren't the problem
they were. The Internet simply isn't a bunch
of LSI-11s linked by 56K lines anymore, and I
recall ``look, doing that will mean every router
has to have a megabyte of RAM'' being used as
an argument-ender.

To compound things, IPv6 suffered from feature
creep (see also: ATM, X.400, Modula 2 standards)
and tried to solve a bunch of other problems as
well, such as QoS. But _those_ were being
solved in v4 land, too, with RSVP, and it's
compatible and interworking with existing
code. Those over 35 should compare the complex
``look, we need multi-part mail'' solution
proposed by the X.400 lobby, which requires MTA
support all the way, with MIME, which will pass
transparently through any MTA.

The final nail in v6's coffin is that, largely,
it's not had the attention of the A team inside
vendors, and has been seen as another add-on
protocol, like OSI, ATM, etc.

I think Vernon Shryver said a few years ago that
he didn't expect universal IPv6 in his working
lifetime. I don't (I'm 37), anymore than I ever
expected my email address to because /O=...

ian

Marx win through IP V6

[alephnull42]

Some info gleaned from a conference i attended recently:

- U.S. military is completely on IP V6

- Big japanese government funding for "pushing" IP V6

- 900 Trial customers in Japan, including 3 cars (all this for just ~$20 Million in Public money)

- The same body received approx $9 Million from European Comission to push IP V6

- Less that 20 commercial broadband IP V6 customers worldwide...

Two scenarios: Like with UMTS, governments pushing a standard they don't understand will result in failure. IP's success was based on market success, and theres nothing like 10^7 dollars for turning a good concept into a bloated Frankenstein monster

OR

The governments are pushing this because it will give them the infrastructure they need to come out with true "big brother" scenarios... Unified protocol with full control

Paradoxically, in this day of "global liberalization of markets", this major infrastructure development is not being driven by market forces, but by centralized government bodies like in the best days of Communism... weird

Just a note: As long as they were nationalized, not a single telegraph, telephone or telecoms company made any profit. Strangely, the same industries started blasting out profits almost immediately after privatization

Re:RFC1918 address are ROUTABLE!

[vrmlguy]

I get so irritated seeing someone refute something that I never said. I didn't say anything about IPv4, I only talked about IPv6. If you are posting a reply the the original post, then do so, please don't post a reply to me unless you are discussing something that I said.

Yes, see that's exactly what's needed

[Colin Smith]

And a way of getting proper addresses, not just test ones.

Keep in mind, though...

[artemis67]

The Windows Update feature would make installing IPv6 a fairly painless operation. You need it? Just go grab that Critical Updates package you've been neglecting for two years.

It's not a showstopper; I wouldn't even say it's a bump in the road, provided MS thought it was important enough to put in Critical Updates.

Re:Well, it's here already (slightly OT)

[Etyenne]

It made internal routing *far* easier.

Not always. A big problem with private adress space appear when two business (or dept, or whatever) bridge their LAN with a VPN and they are using the same private range. Most LAN use etheir 192.168.[0|1].0/24 or 10.0.0.0/8, so this happen often (it happen to me all the time). Hopefully one or the other use DHCP so they can be migrated to an other adress range (almost) painlessly.

Re:Well, it's here already (slightly OT)

[Etyenne]

I am not sure I get your point, and by the little I understand you don't get mine either. The type of scenario I was referring to is not a client connecting to a gateway, it's a gateway connecting to another gateway to make both LAN look like they are local to the client. In this scenario, the VPN connection never get NATed; it is initiated by each gateway on their outbound (Internet) interface. Routing become an issue in this scenario: how are gateway supposed to route if both side of the VPN have the same subnet ???

Just to clarify my thought about DHCP: migrating adress that where statically assigned "by hand" is a lot of work since they must be changed on each workstation separately. If you use DHCP, you just have to edit /etc/dhcpd.conf, wait for the lease to expire (at night, on the weekend, whatever) and bingo! all (or most) of your machine now use the subnet wich hopefully you can route thru the VPN link. Get it ?

Right now, outbound PPTP connection are a real pain to NAT with iptables. There is an iptables connection tracking module but it has not yet been integrated in the base patch. Hopefully it will in iptables 1.2.6

Re:DynDNS

[Ben Hutchings]

Well my cable company leased me the same address for about 6 months and, er, doesn't ban servers. The main reason for using DHCP on cable is, as everywhere else, to enable central management of network configuration, reducing network administration and support costs.

Re:Well, it's here already (slightly OT)

[Pii]

Not to nitpick but...

I take that back. My intent is to nitpick.

RFC 1918 sets aside:

• 10.x.x.x (Class A)
• 172.16-32.x.x (Class B)
• 192.168.x.x (Class C)

as "private address space."

Re:Well, it's here already (slightly OT)

[frog51]

It gets even more fun when using SQLnet and NBT authentication etc if the two merged organisations each use a significant part of the 10.0.0.0/8 space and aren't prepared to DHCP everything.

Believe me the NAT rules are a joy to behold (as long as someone else is supporting them) and usually means having to use other (non-approved) class A ranges as transition addresses.