Hi Chris,

I was awaiting the response to this question as I was myself interested in knowing whether there is anything different with respect to IPv6 address allocation and migration. Please note that while you have provided links to documents how the IPv6 addresses are allocated and requested (for USA region), this does not answer the OP question - whether the management of IPv6 addresses is the same or different to IPv4 addresses when changing ISPs.

Personally I have not heard of any supplementary mechanism for this in IPv6 and my personal opinion is that the situation with IPv6 address migration when changing ISPs is just the same as with IPv4 addresses but I don't want to assert myself here and I would welcome anybody with better experience to comment on this.

Best regards,

Peter

Peter,

My take on this is if an organization gets a netblock assigned from ARIN the service provider is irrelevant.  The difference between IPv4 and IPv6 becomes insignificant as the two protocols can run in parallel over the same media.

I agree with you that when changing ISP's the user experience is much the same in regard to IPv4 or IPv6. If an end user can get an IP address assignment from ARIN versus an allocation from the ISP it makes the firing of ISP's much easier; regardless of the version.  Lets face it; ISP's need to be put on notice for the mediocre 'once size fits all' solutions they attempt to impose on end users.

Chris

Do we know what is the percentage of IP prefixes assigned by ISP's vs ARIN?

I hear so  many people complaining about this that I have the impression the typical and majority of assignments are coming from ISPs.

It's logical for most of the IPv4 assignments to come from ISPs, since we are running out of addresses: http://www.potaroo.net/tools/ipv4/index.html

ISPs won't let you get away with an IPv4 block that has been assigned to them for practical reasons and not because they want to force you to stay as their client. If the IPv4 blocks assigned to ISPs get punched, multiple prefix announcements need to be made for the block that has the hole in it (summarization/aggregation is broken). This means even larger Internet routing table, which means more memory requirements for all routers with full routing on the Internet, which means more cost for devices/upgrades, and guess what would happen to the cost of the service any end client would get.

The biggest motivation towards IPv6 is the larger address space. I agree with Christopher that since users can get their own IPv6 blocks, renumbering might become less of an issue. If, however, for any reason, you need to renumber an IPv6 network, IPv6 has features that make things easier. Manual changes are still needed (e.g. ACLs, DNS, etc). Some procedures are described here: http://tools.ietf.org/html/rfc4192

Some features that make things easier are: link-local addresses, autoconfiguration, and support for multiple addresses on the same interface (which means you can have both prefixes configured in parallel). Stateless autoconfiguration can deprecate a prefix at a certain day/time and advertise a new prefix from that moment onward (interface command : ipv6 nd prefix).