I really enjoy seeing this debate because I hope more anti-ORMers unite and we see less use of ORMs in future projects. There are other disadvantages of ORMs th

I really enjoy seeing this debate because I hope more anti-ORMers unite and we see less use of ORMs in future projects. There are other disadvantages of ORMs that I haven't seen mentioned much: 1) low-level performance. Even if, and it's big if, you manage to get your ORM to generate somewhere near the optimal query, ORMs in my experience are always significantly slower than hand-written sql. When I last benchmarked, I couldn't get hibernate to be any better than 4x as slow as JDBC, and keep in mind that's pure CPU overhead. Think your service is I/O bound? It's probably not, and it's probably your ORM to blame. This may be less of an issue for a dynamic language like Python, but I see it as a much bigger issue for Java/C# and friends. 2) Debugging/understandability. Did you know that hibernate maintains a cache of every object you load in a session until you flush it? I didn't, until we had an outage because our service OOM'ed while loading too much data without flushing. Do you know how exactly your ORM is loading and saving data and when? Depending on your use of the various lazy-loading and storing features of your ORM, it can be very difficult to reason about when and how your ORM is talking to your database. Do you know how your ORM is integrating with your cache, which is likely memcached? Why is your ORM integrating into itself the concept of a cache in the first place? In my experience, hibernate gets caching wrong, and that's not entirely its fault. It's difficult to get caching right in the general case. But I would rather be forced to think about caching up front and get it right for my use case rather than try to understand how Hibernate is doing it and working around its mistakes and limitations. The common theme is that the use of an ORM makes it incredibly more difficult to understand, reason about, and debug your application rather than using a simpler library. In my experience, this alone makes an ORM not pull its weight.