When I mean ordering, I don't mean the term order, but rather the order of the generators by which you execute the multivariate division algorithm. BTW, I've looked at Cox-Little-O'Shea and I think the answer is that one can use different orders (in the latter sense), but just want to confirm.