The premis(e) didn't include the full situation in it.

Without the ability to allocate IPv4 addresses to new equipment some people are not going to be able to tunnel over IPv4 because they won't have an IPv4 address to tunnel from. ISP level NAT is the worst possible solution.

The last /8 in each region is meant to be allocated for IPv6 migration services, not general use. As the number of IPv4 addresses reduces the bureaucracy and expense involved in getting an IPv4 address for any purpose will likely increase. Address traders ( i.e. scum ) will move in to try and make a fast buck and everyone will suffer for it.



This is a routing artifact, you got lucky. It only proves the IPv4 routing between your home and your linode is sub-optimal.

removed; I don't want to rehash this argument yet again

_________________
Rgds
Stephen
(Linux user since kernel version 0.11)

So is the SKY falling or not?

There isn't an argument. There isn't even a discussion. Native IPv6 is a must have. The current tunnelbroker setup is good for light testing but sucks for any serious use.

Everyone who wants to keep their networked applications going for the next few years can either test them on IPv6 or cross their fingers and hope for the best. I want my applications tested and working before people want to use them.

Woohoo! I can't believe I'm the first to post this here, but:

Linode IPv6 Announcement
Linode IPv6 FAQ

It is really bizarre that Linode is only offering "a single IPv6 address" per Linode, when there is no scarcity and even an HE tunnel gives you a /48 for free (which is this many IPv6 addresses: 1,208,925,819,614,629,174,706,176). But I'm sure this is just a hiccup. Right? The FAQ page says you can pay for more IPv6 addresses, but also seems to imply that we'll be able to get a /64 soon (presumably for free??)

Anyway, thanks for the good work, Linode, and I'll be filing my IPv6 address request ticket shortly

Looks like it is using some form of autoconfiguration (based on the config instructions), so getting a /128 by default makes sense. Anything larger than that would require an offering to the routing god and a change to your /etc/network/interfaces.

Even for larger allocations, this is still a necessary first step -- it's a single IP address, like the point-to-point tunnel address for a HE tunnel. Getting the second step (a routed /64 or shorter prefix) necessarily follows from this.

_________________

/* TODO: need to add signature to posts */

For the one node I had this enabled on in Fremont, the assigned /128 appears to be an EUI-64 address based on my Linode's local MAC, producing the global scope address, and, with prefix reset to fe80, the link-local address. So yes, auto-configuration. No guest networking configuration changes made.

As the initial address for the node, I don't see any problem with this, and perhaps it shaved a little time in infrastructure support for managing routing tables for assignments of new IPv6 blocks. It certainly gets IPv6 traffic right to the Linode, and any subsequent blocks assigned can just use this address as the routing destination.

Even my other nodes with he.net tunnels work similarly where the default routed /64 is distinct from the block from which the IPv6 address of the tunnel endpoints are taken.

-- David

No offering to the routing gods from a client perspective; merely the default router advertisement packets. Yes, a small routing table at the linode end, purely automated build processes.


Yes, but at least you've have the option of using the addresses you want to.

_________________
Rgds

Stephen

(Linux user since kernel version 0.11)

If you're only seeing the fe80 prefix address then you're not enabled for IP6 (or you haven't rebooted since it was enabled). You should see two addresses; the fe80 link local and the routable IP6 address

_________________
Rgds

Stephen

(Linux user since kernel version 0.11)

Um, not sure how you interpreted the above quote to imply I only saw the fe80 prefix? Note the "and" in between the global scope and link-local addresses (which are clearly two distinct addresses). I certainly didn't mean to imply I was only getting a link-local address so my apologies if that was confusing.

-- David

You're right; I misread what you wrote. Sorry

_________________
Rgds

Stephen

(Linux user since kernel version 0.11)

http://blog.linode.com/2011/05/03/linod ... 6-support/

Any news on Dallas being included?

_________________
Why can I not enter my IRC info in the forum profile?

"soon"

_________________
Matt Nordhoff (aka Peng on IRC)

Works great.

Thanks for adding this, and thanks for switching it on rather than waiting for all the datacenters.