<i>all_object_party_privilege_map</i>
<p>
Yes, joining with this as part of your query that selects classes may well be faster. It's worth a shot and since you're in measuring mode already, Andrew, you're a good candidate to take that shot :)
<p>
If you have time and want to report back it would add to our knowledge of how to work around permissions performance issues, regardless of the experiment being a failure or success.
<p>
As far as reordering goes ... you have some good, solid data and it sounds like OpenForce is getting some, too. Jon Griffin and I both played around with ordering many moons ago and settled on what you see now. I find it interesting that OpenForce appears to be discovering that the optimal order may differ between PG and Oracle.
<p>I don't have an answer for your question about the join with the users table in the public query. It looks very strange, doesn't it?
