Defining identity well is something that causes philosophers to tear their hair out. Pragmatically, I think it's easy enough to say that a node acquires an identity when it's created and retains it for the duration of a query, and even that it retains identity through updates; the problem comes when the node leaves the comfortable closed world described by the query spec and acquires its own life in the outside world. The XQuery spec doesn't know anything about that world, so it can't say anything useful about it. But the only way of testing whether XQuery Update has preserved node identity is to conduct the tests in that outside world, using interfaces over which the XQuery spec has no control. The WG, incidentally, decided to keep the clause that says updates preserve identity. What isn't clear to me now is whether that is supposed to impose a requirement on the environment/API to provide a way of determining that identity has been preserved. In some environments, for example client/server models, that's pretty tough.