There seems to be an underlying assumption that SQL performance is crucial. Of course an ORM is going to produce less efficient SQL, that's the tradeoff. Howeve

There seems to be an underlying assumption that SQL performance is crucial. Of course an ORM is going to produce less efficient SQL, that's the tradeoff. However, object caching may eliminate this concern. For example, I've worked with a system where each field is a table and data retrieval can require huge inefficient joins just as you'd expect. Due to general revulsion at the idea when it was introduced, there was an initiative to create materialized views. But this completely collapsed when it was discovered that the existing object caching benchmarked just as well whilst removing system complexity. Also worth thinking beyond selects - another very big advantage of table per field is in altering schema on large datasets, which was a major undertaking on more read-efficient schema. I think this is an example of worse being better.