A separate data-registry for each dataset or combine them into one?

I wonder if it is better to use the same data-registry for two or more separate projects? Or would it better to use a new one for each dataset? For example, what would happen if I have two folders /data in two different projects, then check them add them to a single data repository and then try to run dvc checkout data ?? Or dvc pull.

Hi,

can elaborate on your usecase?

I guess a typical setup is to have all datasets in the same dvc repository and then use it in other projects using dvc import (docs: import)