Fabio: The resulting data is the same. 10,000 POCO objects with a sub list of 100,000 other POCOs. Obviously ADO.Net is faster, but that was never the point. I only ran that test because you asked for it.
Stephen: Raw DB access performance is NOT an overriding criteria. I've tried in this discussion to point out that it is. We just need to know what happens if we try to do it. NH performs well enough on large sets that we don't have to implement two ways of loading data, which is the result we were hoping for.
