Comment 6 for bug 738188

There were some more race conditions that I hadn't handled, and the solution I had was getting more and more unwieldy. I'm going to have to rethink this. It may mean doing away with "references" in marshalling, and always passing an ID/opmask/args tuple, so the receiver can create a new proxy if the old one is deleted in the mean time.

One thing I'm still not sure about is how to notify the originator of an object that it can release the proxied object. We need it to keep a reference for as long as the client may refer to it, so the originator can use the original object, not a proxy of a proxy. I'll have to think about this some more.