It's definitely a cool idea. I think it makes sense to make it configurable, since some workloads might write already compresses bytes and then it's not possible to compress them any further.