As great as EXR was (compared to as mentioned the crappy alternatives) EXR was still handicapped by the same huge flaw that’s going to be Alembic’s undoing: the standard writer’s pathological fear or having an opinion. EXR should have had a “best practices” policy where ancient, well established concepts were codified. For instance EXR has no “official” name for its depth channel. As a result Maya, Max, VRay, Brazil, Nuke and Combustion all decided they needed to come up with their own name. We have depth, Zdepth, Z_depth, Z-Depth etc etc. ILM should have listened to the developers saying that failing to standardize would result in a hodgepodge of incompatible implementations. They also should have defined standards such as 1/Z or true Z etc. I can understand wanting to take a step back and be flexible so that you don’t lock someone into one way of doing things, but EXR does lock people in. It generally implies that you’re saving a linear non-gamma encoded image. And simple draconian decision made it far superior to DPX which 50% of the time is sRGB and 50% of the time is Cineon.
I feel like Alembic is falling into the exact same trap. They’re refusing to define anything so everybody is defining it differently, even when there is no advantage to that “diversity”. XMesh I think will do better as an interchange format since it’s actually putting a flag in the sand and saying “This is for Normals defined as…”