Thanks for the feedback John. Re (2) you seem to be suggesting that the PUL should actually be exposed to the calling application as part of the API. I wasn't actually intending to do that, I was thinking in terms of the lowest-level method call being "run-and-apply-updates" that (conceptually) runs the query to obtain a PUL and then applies the updates. But either way, it's still not clear what happens to "insert into ". After all, the upd:applyUpdates() primitive doesn't actually return the nodes that were updated. My current thinking is that my run-and-apply-updates should return a set containing the root nodes of all trees that have been updated, including temporary trees. That leaves the user the problem of what to do with them. Re (5) supporting HTTP PUT seems reasonable in principle but it probably needs quite a bit of API complexity (collecting credentials etc) to make it work in practice, wouldn't you think? Might be better to let the user application take care of this.