[OpenID] Nonces generated by the server?

Allen Tom atom at yahoo-inc.com
Wed Apr 1 01:21:38 UTC 2009


OPs that are globally distributed would have issues with validating the 
uniqueness of the nonce, as it implies that the nonce needs to be 
replicated everywhere. OPs which use IP-based load balancing could have 
a lot of issues.

For instance, in the worst case where the user's browser and the RP are 
on opposite sides of the world, the nonce would be issued by the OP in a 
datacenter closest to the user's browser, and then immediately submitted 
for verfication by the RP in a datacenter on the opposite side of the 
world. In theory, the OP serving the verification request could forcibly 
try to sync the data if its not already present in its local cache, but 
there are often networking issues between distant datacenters, and OPs 
may want to have their systems be resistant to networking issues between 
distant geographic locations, at the expense of having slightly stale data.

I would recommend that RPs that want to protect against replay attacks 
use HTTPS for all OpenID requests.

Allen

Breno de Medeiros wrote:
>
> I believe that the spec should make it clear that the OP is 
> responsible for validating the uniqueness of the nonce in stateless mode.
>  



More information about the general mailing list