What prefix will IPv6 ISPs assign?

[Kizaki]

Hi there,

I have two questions. I just started using, and reading about IPv6 yesterday.

IPv6 so far has been easy to set-up and use. I have three /64 prefixes up so far using OSPF (Quagga) on my /48 provided by HE over their tunnel. IPv6 doesn't seem to have much of a learning curve over IPv4.  What's keeping people from moving over to IPv6, since its not that hard to set-up?   Personally I find IPv6 easier to configure and manage than IPv4.

The second question has to do with prefixes that ISPs might assign to their customers.

I am finding a lot of contradicting information about the prefixes that ISPs will assign to their customers once they start allocating IPv6 addresses. Are they going to assign /48 addresses to end users, like they are supposed to?  That would be a better option in my opinion, as /64 are not supposed to be split into subnets. 

[broquea]

Probably the cost of replacing/upgrading any equipment not ready to handle both stacks. Some places have buying cycles, like every 3-5 years. Also I suppose demand. If no one asks for it, they don't feel inclined to bother.

As for allocations, I've no idea what other providers will do, however we reserved a /48 in our /32 so we could allocate /64s out of to colo/transit customers that want to take those first baby steps and learn. Otherwise we either set up a point-to-point allocation with a static route through it (/64 with a /48 routed through their side of the /64), or we run BGP and either allocate them some space to re-announce (but is part of our /32 so we don't have to advertise the individual /48) to us, or get their own allocation.

I saw one person's IRC comments on a provider who only assigns the customer a /48, sticks x:y:z::1 on the provider's router, and tells the customer to figure out the rest.

[Kizaki]

One more question. 

What is the (bandwidth limit/cap/reasonable usage) or whatever you want to call it for the IPv6 tunnels provided by HE.  I doubt I will be using much any time soon.  I just want to know for future reference.

[broquea]

One more question. 

What is the (bandwidth limit/cap/reasonable usage) or whatever you want to call it for the IPv6 tunnels provided by HE.  I doubt I will be using much any time soon.  I just want to know for future reference.

Whatever you can manage to push out, every tunnel-server has a GIGE uplink with an option for more. Also IPv4 TCP windowing will be a factor since you still have to traverse that. Obviously native IPv6 wouldn't encounter that particular issue.

[Kizaki]

I am using a Linux server with 1&1 as my IPv6 router.  I also have two Windows 2003 servers with them connected to my Linux server as well, for the IPv6 access. The Linux server that is running the IPv6 tunnel has an 100Mbit connection.  Also, my ping time to the Chicago server is about 11ms.  So I can probably push out quite a bit from the servers, let alone my home and business connection. 

That's the reason I asked about the tunnel traffic limit.  The tunnel is free, so I don't want to abuse your connection if I actually end up using it a lot one point or another.

1&1 is supposed to be supporting IPv6 one point or another.  They told me that their network is ready for it, but they can't until all their peers have IPv6 support.  If your AS 6939, then I presume you are one of their peers.  I don't know why they need support from all their peers rather than just a few. 

I got that AS from this website
http://www.fixedorbit.com/AS/8/AS8560.htm
It has all their peers listed.  I presume HURRICANE is you

[broquea]

Yup, we are AS6939

Waiting for "all" peers might take...well who knows how long, that includes whoever they pay for transit probably.

[Kizaki]

I saw one person's IRC comments on a provider who only assigns the customer a /48, sticks x:y:z::1 on the provider's router, and tells the customer to figure out the rest.

Don't you need to set-up a point-to-point connection first, and then route the /48 to the customer? Thats one friggin huge subnet.

[broquea]

I saw one person's IRC comments on a provider who only assigns the customer a /48, sticks x:y:z::1 on the provider's router, and tells the customer to figure out the rest.

Don't you need to set-up a point-to-point connection first, and then route the /48 to the customer? Thats one friggin huge subnet.

Sadly, they had requested that exact setup and their provider refused, stating that how they were doing it was policy.

[Kizaki]

Sadly, they had requested that exact setup and their provider refused, stating that how they were doing it was policy.

That's what I was afraid of when I asked about prefix assignments on my first post.  If common ISPs (or any ISP) do things like that, people are going to resort to using NAT.  That's just going to make the transition to IPv6 chaotic, and kill end-to-end connectivity as well. The big advantage to IPv6 is to have a global routable prefix, and get away from NAT entirely.

Also, one of the network administrators at my university who is familiar with IPv6 told me that ARIN requires ISPs to provide either /48 or /56 to their customers.  It's supposed to be part of their agreement with ARIN when they obtain their IPv6 prefixes.  I can't find anything on the net that says this.  Do you know if this is true or not?

Thanks

[broquea]

http://www.arin.net/policy/nrpm.html#six

See section 6.5.4

From what I've seen for our customers requesting IPv6 allocations, ARIN stated: "To qualify for an initial allocation, you need to show a projection to make 200 end-site assignments within five years." I'm pretty sure this was them requesting a /32.

[testmonster]

Also, one of the network administrators at my university who is familiar with IPv6 told me that ARIN requires ISPs to provide either /48 or /56 to their customers.  It's supposed to be part of their agreement with ARIN when they obtain their IPv6 prefixes.  I can't find anything on the net that says this.  Do you know if this is true or not?

RFC3177 (http://www.rfc-editor.org/rfc/rfc3177.txt) does say this:

This document provides recommendations to the addressing registries
   (APNIC, ARIN and RIPE-NCC) on policies for assigning IPv6 address
   blocks to end sites.  In particular, it recommends the assignment of
   /48 in the general case, /64 when it is known that one and only one
   subnet is needed and /128 when it is absolutely known that one and
   only one device is connecting.

It would be useful to match that up with ARIN if somebody wants to dig up the specific URL.