Comment 67 for bug 1314129

The issue came up in a patch that I had in flight here: https://review.openstack.org/#/c/112178. I wrote a unit test to check my assumption and it failed on python2.6 in Jenkins but not locally in my own python2.6 virtualenv. I was perplexed until I updated oslo.messaging and then installed simplejson to match the Jenkins run. Without my new unit test, I never would have caught the issue because there is not enough testing with python2.6 in the gate to catch this.

What I'm reading is that it was assumed that simplejson would be a suitable drop-in replacement for for json without really looking very deep and you'd rather wait for issues to arise. Given the luck involved in catching this issue and reproducing it locally, I'm not sure that is wise. At this point, I'm not going to hold up my work to wait for a fix in oslo messaging and an update to Neutron to consume the updated dependency. I'm going to have to just change direction and find another way.

It should be noted that namedtuples should not be used in any RPC message with the current oslo.messaging as the serialized JSON representation will be platform and environment dependent.