openstack team mailing list archive

[Brian Schott <brian.schott@xxxxxxxxxxxxxxxxxx>]

I was just re-reading RFC4122 on UUIDs and the different variants:
  http://tools.ietf.org/html/rfc4122.html  
 and what Python gives you in the UUID class:
  http://docs.python.org/library/uuid.html

We need to decide which variant is right for us.  With the namespace variant uuid5(), it is possible to regenerate the UUID from a namespace and a name.  The namespace could be the DNS of the API service and the name could be a concatenation of the username and a generated instance_id, reservation_id, volume_id, etc. string.  That way, as long as you are talking to the same API endpoint, you can use EC2 style ids and convert easily to globally unique UUIDs internally.  Backwards compatible 1.0/1.1 OS API servers can do the same.

Brian


Brian Schott, CTO
Nimbis Services, Inc.
brian.schott@xxxxxxxxxxxxxxxxxx





On May 27, 2011, at 3:01 PM, Ed Leafe wrote:

> On May 27, 2011, at 1:15 PM, Erik Carlin wrote:
> 
>> With the proliferation of new openstack services being built, is there any reason not to use UUID as the standard resource ID format?
> 
> 	The consensus at the last summit was to move to UUIDs for instance IDs. The biggest concerns were that a) it breaks the current API (which can be updated), and it breaks the EC2 API (which can't be updated). I know that there were some ideas for working around the ec2 issues, but I don't remember them.
> 
> 	I think moving ahead, UUIDs scale way better than locally-generated sequential integers.
> 
> 
> 
> -- Ed Leafe
> 
> 
> 
> 
> _______________________________________________
> Mailing list: https://launchpad.net/~openstack
> Post to     : openstack@xxxxxxxxxxxxxxxxxxx
> Unsubscribe : https://launchpad.net/~openstack
> More help   : https://help.launchpad.net/ListHelp

Attachment: smime.p7s
Description: S/MIME cryptographic signature