Closed (outdated)
Project:
Drupal core
Version:
6.x-dev
Component:
openid.module
Priority:
Normal
Category:
Bug report
Assigned:
Unassigned
Reporter:
Created:
2 Mar 2010 at 18:00 UTC
Updated:
2 Mar 2016 at 22:18 UTC
Jump to comment: Most recent, Most recent file
Comments
Comment #1
damien tournoud commentedHm. The way I'm reading that is that when the request
openid.assoc_handleis expired, the OP has to "proceed as if no association handle was specified" (ie. as if that field was not in the request), and has to put the expired association handle in the response asopenid.invalidate_handle.As a consequence, the response should not contain any
openid.assoc_handle, and openid_verify_assertion() will proceed with direct verification.So, except if the OP doesn't comply with the spec, nothing wrong should happen. The only issue is that we will continue to send invalid association handles to this OP, that will have to be verified in the (less performant) direct verification mode.
Comment #2
damien tournoud commentedAnyway, this should do it.
Comment #3
damien tournoud commentedJust to clarify #1, my analysis is that 1) we do not violate the spec here, because the spec only specify the behavior of the OP, not of the RP, 2) nothing wrong should happen as long as the RP comply with the spec.
Comment #4
heine commentedAlas,
If we receive an invalid_handle response, we must verify directly with the OP (we do now, assuming the OP is well-behaved and sends an empty assoc_handle).
Only then, when the OP confirms that the handle is in valid should we remove it. see section 11.4.2.2:
TODO - I still need to look into 11.4.2.1. about shared keys.
TODO - we must do ns checking - maybe in a separate issue?
Comment #5
heine commentedGah, wrong patch.
Comment #7
heine commented#5: do-730462-invalidate_handle.patch queued for re-testing.
Comment #8
damien tournoud commentedRerolled with a some additional comments and a smallish style issue.
Comment #9
dries commentedCommitted to CVS HEAD. Thanks!
Comment #11
c960657 commentedD6 backport.
Comment #12
pwolanin commentedstill an open core bug