openstack team mailing list archive

[Soren Hansen <soren@xxxxxxxxxxx>]

2011/7/11 Ed Leafe <ed.leafe@xxxxxxxxxxxxx>:
> On Jul 11, 2011, at 2:04 PM, Eric Day wrote:
> It's a shame that the ipv6 proposal was never more fully considered. That would handle the uniqueness, with the added benefit of providing simple zone routing via DNS, with the exact same 128-bit/32 char size.

I can think of a number of reasons why using ipv6 addresses are a bad idea.

1. They only naturally map to VM's. Using them for any other object
types seems artifical (why would a snapshot need an ipv6 address?)
2. One of the reasons UUID's were chosen was because it gave us a
gigantic key space that made collisions unlikely enough that we
needn't worry about them. If you use ipv6 addressses as your ID's,
you're limited to the part of the ipv6 address space you've been
assigned (rather than the full 128 bit key space). I realise /64's are
pretty commonplace, but collisions are ~10 orders of magnitudes more
likely in a /64 keyspace than they are in a /128.
3. You'd never be able to recycle IP's. UUID's are (and EC2 ID's are
said to be) perpetually unique. They do not get recycled. If an
instance's ipv6 address is its instance ID, you'll have to either live
with your pool of ipv6 addresses slowly depleting (even though their
corresponding instances have been deleted) or recycle the instance
ID's (and hence IP's). Neither of those options seem very cool.


-- 
Soren Hansen        | http://linux2go.dk/
Ubuntu Developer    | http://www.ubuntu.com/
OpenStack Developer | http://www.openstack.org/