In reply to a post by billford:

(I agree with Ignitionnet- the current system is the least worst way. Run the systems in parallel, when IPv4 addresses run out, site operators have a choice- implement IPv6 or see their hit rate inexorably decline.

But your conclusion is false.

As the plan seems to be that all end users will have an IPv4 address, be it dynamic, static or CG NAT, site operators have no incentive at all to move to IPv6. Only new sites that cannot get an IPv4 addresses will take IPv6, which the end users will be able to access because they are dual-stacked. It's going to be a long time before there are anything like as many IPv6 sites with no IPv4 access as there are IPv4 ones with no IPv6 access.

I've already said I'm arguing from little knowledge of the technology, but logic, human nature, and life experience point towards a right mess with very little chance of resolution for decades.

All of you that argue against that belief seem to be starting from a position of "received wisdom" that the way BT et al seem to be going is the best, without any discussion of alternatives such as the one I suggest. I'm sure there must be others as well.

In reply to a post by RobertoS:

All of you that argue against that belief seem to be starting from a position of "received wisdom" that the way BT et al seem to be going is the best

Flatly disagree, especially with the emboldened bit.

See Oliver341's query and Andrew's response.

If IDNet (and others, I assume) can do it then anyone can. If they don't then they're corner-cutting.

(For the benefit of others- IDNet are my ISP, and thoroughly recommended )

I agree with you about human nature but at some point IPv4 addresses are going to run out, and incorporating dual stacking at both ends is the easiest way to make the transition as seamless as possible.

If I understand the terminology correctly then I'm dual stacked here- without explicit checking I never know whether I'm using IPv4 or IPv6 to connect to a site. It's completely transparent.

Yes the migration from IPv4-only to IPv6-only was always meant to proceed through a series of phases, and allows that these phases can have fuzzy edges and may take the order of a decade.

@RobertoS
Because the IPv6 space is so much bigger, it allows for IPv4 addresses to be embedded with room for a prefix that identifies the transition method in use, of which there are several, partly because it depends which way the traffic is going and which ends have IPv4, IPv6 or both.

The requesting device understands this is a translated address and more importantly so do the routers and know what to do with it.

But in simpler terms it boils down to an endgame where:
- all customers of an ISP are provided IPv6 and
- can contact IPv6 hosts directly and
- can contact legacy IPv4 hosts via address translation or proxying (which can be transparent; this is not new tech as for example ALGs already know how to rewrite IPs and ports on the fly for services such as FTP if necessary)
- crucially these mechanisms can be setup to be automated once configured by the ISP so the customer doesn't need to know how to do any of this

Similarly while If an ISP still has enough IPv4s for each customer, they can continue to map these to each customer, directly with dual stack or via NAT (e.g. on behalf of a host on the internet trying to contact that customer).

However the first step remains to get both the ISPs and hardware manufacturers (or rather their firmware development) using IPv6 (and hence dual stack) and issuing this to customers as standard so that the other phases can proceed.

Edited by prlzx (Wed 08-May-13 01:12:02)

That looks as though it makes sense, but I need to be more awake to take it in properly .

I should have added, in IPv4, people often assume it unusual or even wrong for some network interface to have more than 1 IP address (which is allowed, but for many use cases not necessary).

But in IPv6, that is not only normal, it is to be expected. For example most host interfaces will have (and use) both a link-local address and a globally unique address.

So the ISP could assign several kinds of IPv6 addresses to the customer.
As well as their "main" prefix (e.g. a /48) that the customer organisation can subnet from, I think there could be additional addresses (for NAT purposes) assigned to the customer router without confusion.

This is another reason the addresses seem so long, so that the prefix can identify what kind of address it is.
Being able to determine just from the prefix, what to do with the packet; for routers - whether and where to route it, for other hosts, whether they need to process it (if at all).

And even the current IPv4 addresses could be considered to have prefixes, if you have seen the binary forms it is clear that a host can figure out what class it is (A to E) from the first 3 to 4 bits.

In reply to a post by billford:

If I understand the terminology correctly then I'm dual stacked here- without explicit checking I never know whether I'm using IPv4 or IPv6 to connect to a site. It's completely transparent.

Hopefully I'll be the same when my new router arrives. I'm also hoping to have my FTP server accepting IPv6 connections and perhaps (license allowing) my mail server.

Since my earlier post I've discovered an oddity... given the choice, Snow Leopard would use IPv6, Mountain Lion (and Lion) uses whichever seems to be "fastest" for the site in question. Which is usually IPv4 of course.

Nought out of ten for standards observance, Apple... though I'll concede it's still transparent.

eta- Apparently Chrome and recent Firefox builds have similar behaviour, and Win8 will ditch IPv6 at a hint of a problem.

Edited by billford (Wed 08-May-13 09:51:36)

As an IPv6 ignoramous, I've been following this thread with increased befuddlement. Shamefully, I have to admit I am none the wiser, but have come across tunelling in my searches, has it anything to offer?
http://tunnelbroker.net/?a=26543899873&n=g&pos=1t1&p...

Edited by professor973 (Wed 08-May-13 12:54:12)

As it says "Once you configure your side you will be able to reach the IPv6 Internet"

In reply to a post by professor973:

As an IPv6 ignoramous, I've been following this thread with increased befuddlement. Shamefully, I have to admit I am none the wiser, but have come across tunelling in my searches, has it anything to offer?
http://tunnelbroker.net/?a=26543899873&n=g&pos=1t1&p...

Further confusion really. I noticed a while back that Windows 7 establishes a 'Teredo' tunnel without asking. Doesn't seem a lot of use though. Most IPv6 test sites I've tried say I can ping to IPv6 and that my ISP (IDNet) DNS suports it but then rate my support as 0/10.

Same here at work although that might be the corporate firewall butting in. But I have IPv6, it works within the office and Timico DNS servers are apparently IPv6 capable. But 'ping ipv6.google.com' fails with unknown host and if I use the IP address:

Pinging 2001:4860:b002::68 with 32 bytes of data:
PING: transmit failed. General failure.
PING: transmit failed. General failure.

Damn' General Failure. Always getting in the way. Nearly as bad as Corporal Punishment and Major Disaster