"BT won't officially support IPv6 until we need to use it to justify a price increase"

Allow me to paraphrase all of this into a nice tidy little statement:

"The internet system we are rolling out will be completed just in time to be incompatible with the internet. We apologise for any inconvenience caused by spending billions of pounds on a system that will work up until the point in which it is ready to be deployed."

Nice one.

Not that many other ISPs (in reference to AAISP) are embracing IPv6 either mind. S'ok lads... its only "supposed" to be completely switched over by 2012... no rush or anything...

-- Richard

a-one, a-two, a-three, a-one-two-three...

It's time, for one of those things, that nobody bothered thinking through.

Oh look, 21CN is a failure. I didn't see that coming.

Paris could do a better job than sodding BT.

its about time they started supporting IPv4 its certainly a move forward.

That ipv6 stuff is expensive, but I've a mate in the US that's outing his ipv4 kit cheap, so we can save millions for our shareholders.

US was laying fibre in the 60's, BT didnt start till the 90's

UK is the poor relation springs to mind.

I still cant understand why BT moans even tho it never paid for it's original infrastructure, unlike the cable companies, BT's was free gift from the UK taxpayers

Your headline, and the opening paragraph of the article, suggest that BT's hardware does not support IPv6:-

"BT's ongoing multibillion pound "21CN" network upgrade won't offer native support for IPv6"

The quote from BT suggests something entirely different, simply that they don't support IPv6 at the moment: -

"BT currently supports IPv4 on it's broadband products and does not support IPv6."

Which of these is actually the case? It makes a huge difference. The article seems to be implying that BT's ongoing network upgrades are a vast waste of money, since the equipment is incapable of supporting IPv6. Is this true?

>NAT works perfectly well, and there is also plenty of allocated-but-unused space that could be reclaimed, i.e. General Electric, Ford etc - http://xkcd.com/195/

And there's plenty more on the way. Lehmans, Bradford & Bingley, and Iceland don't need theirs any more...

Still, I am tempted to string up any BT exec in sight using the nearest CAT5 cable - they've screwed a vast amount of cash out of the public and failed miserably to deal with the telecoms infrastructure, and it's time they invested it in the future.

Has anybody read original article? IPv6 is not supported because of the bug in Cisco routers. Fix has been around for a while but BT couldn't be bothered to install it everywhere. Thus it "mostly" works. Somewhere, but not everywhere.

>> NAT works perfectly well

Well if "works perfectly well" doesn't extend beyond "I can browse my porn". It's clueless tits with attitudes like that who are responsible for the backwards state of our internet. The reason IPv6 isn't common already is simply because clueless tits think NAT is a good idea - news flash, it isn't a good idea, it never was, and it never will be.

NAT == broken<period>

NAPT as the specification is called was designed to overcome the 32bit limit in IPv4. It's a hack. Simple as. Just like Simon Hobson says.

A genuine thank you to Jack for actually reading the problem details. Still, it's just another little hiccup in the delayed rollout of BTw's much over-hyped 21CN.

Mind you, this is the same BTwholesale that back in the early days of broadband (2001?) had constructed a wholesale network that couldn't cope with any significant workload unless punters made a simple but significant change to their IP congigs (reducing their MTU from the usual factory default). Back then it was because BTw's network architects (using the word very loosely) hadn't realised that Cisco Express Forwarding would grind the punters to a halt if BTw's internal network didn't support "jumbo frames" (which, at the time, it mostly didn't). That took months to sort. http://www.thinkbroadband.com/news/i/904.html

And is this the same BTw that had bandwidth allocation back to front between 1Mb and 512k (and, later, 2Mb), and then priority inversion between business and residential services, on more than one occasion?

And the same BTw that left exchange backhaul undersized (by their own criteria) for month after month after month in many parts of the UK?

Yes, that'll be them. Incompetenece exceeded only by the likes of Bulldog (RIP).

wrt NAT == broken: you may think you have a clue, and you may be right, but 99.999% of folks on t'Internerd don't know (or want or need to know) about NAT, because the things that 99.999% of Internerd users do are working just fine with NAT and without IPv6, and those that do care can mostly work out ways to make the interesting things work in the presence of NAT, broken or not.

IPv6 has been "the next big thing" for more than a decade (my first IPv6 course was in those early days), and in another decade it may still be that way for most folks. Fancy giving us an example of a "killer app" that really needs IPv6? You can't, by definition, because although IPv6 has been around over a decade for the hardcore enthusiast, THERE IS *STILL* NO VOLUME MARKET for IPv6. Yet. And there's barely a datacentre market for it either.

That being said, respect is due to the likes of AAISP for at least making it possible for their punters to give IPv6 a try if they feel like it. Till BTw broke it, anyway.

I thought that was the whole point of 21CN. You had a native IP core, and for all legacy protocols you just tunnel across it: IPX, leased lines, IPv6 etc.

I guess that's why they don't bother to apply the patch to make IPv6 work...

Tom Chiverton said

>>>> NAT works perfectly well, and there is also plenty of allocated-but-unused space that >>>>could be reclaimed, i.e. General Electric, Ford etc - http://xkcd.com/195/

tom, you are a world class fuckwit.

first off, nat doesn't work. it can't scale. nat can't handle lots of the funky p2p and streaming audio and video applications 'cos these tend to signal which (random and often changing) port numbers are used in the actual protocol. unlike web which tends to always be on port 80 and is straight forward to nat.

even if the allocated-but-unused space was reclaimed and reused, there isn't enough ipv4 addresses to go around. 6 billion people doesn't go into 4 billion ipv4 addresses. half of that ipv4 space would vanish overnight if every one of the 2 billion mobile phones had an ip address.

paris icon 'cos she knows a thing or two about squeezing things that are too big into a small space and making them fit.

So, the examples of apps which need IPv6 so far are:

1) P2P

Well, I suspect any P2P user from the era of Kazaa and WinMX to Bittorrent and beyond would disagree with the statement that P2P don't work behind NAT. As would any ISP, and the RIAA, and the MPAA (if they even know what NAT is). Millions of Torrent users can't all be wrong, can they?

Do the BBC and C4 and Sky (etc) program download services [1] use P2P (hint: all use Kontiki), and do they work behind NAT (hint: yes)?

Sorry, P2P is not a "killer app" for IPv6.

2) Server on mobile broadband

This is more promising, thank you for that. But you've reminded me why I still need BT Callsign and a Twintalk box [2]; I use it as a dial-in backup for remote access to my home network in the event that the DSL broadband fails while I'm away. Callsign costs a lot less than a genuine 2nd line, and Twintalk gives me a dedicated number for the modem. Of course if I was serious I could use an actual 2nd line on a very different ISP, maybe even with broadband, and that would give me some resilience against actual line failure. But this is a lot cheaper, and given what's at stake, it's sufficient. So far.

So, all that being said, how many other people actually want to run servers on mobile broadband? Enough to make IPv6 essential for their favourite "killer app"? I can't think of many applications and I'm a bit of a geek (backup dialup in to home network? If that's not geek, what is?)

C'mon, there must be more? What about Internet fridges? Things like that were going to be the thing that drove IPv6, once upon a time [4]... your intelligent fridge complete with barcode reader gets a call from the supermarket every now and again, and there are so many fridges needing their own addresses that it only works with IPv6, right? No, not right. The fridge makes the call, not the supermarket (because it's the fridge that knows when it's re-stocking time). The supermarket, not the fridge, is the server, and therefore it works just fine even behind NAT (in the unlikely event of anyone actually caring about this). You can get IPv4 stacks for cheap microcontrollers that fit in a few kbytes of code and data [3], has IPv6 reached that territory yet, if not why not? Cost (more memory)? Complexity (=>more memory=>more cost)? Lack of demand?

Sorry, a server (any server) on mobile broadband is not a killer app for IPv6. Maybe one day though.

3) IPsec

I'll add this one since no one else has :) NAT breaks IPsec, and IPv6 fixes it again. But even there, there are already plenty of IPv4-based alternatives.

[1] http://telebusillis.blogspot.com/2007/01/4od-some-technical-architecture.html

[2] http://www.twintalkhelp.com/

[3] http://ww1.microchip.com/downloads/en/DeviceDoc/consolid.pdf

[4] http://www.google.com/search?hl=en&q=%22internet+fridge%22+ipv6

I don't really see what the big fuss is about BT's 21CN it's just a fairly standard "next generation network". Other telcos all over Europe and all over the world are rolling out similar infrastructure and not making a huge song and dance about it.

Unbundled ISPs in the UK and Ireland are already using '21CN' type technologies and have been since their launch.

From what I can see 21CN is entirely a marketing campaign. The upgrades to infrastructure are nothing particularly special.