My problem is processing of a lot of files, each file is processed independently. If the process crashed or data updated/added I’d like to re-produce only missing data-products, not everything. Makefile makes such tasks easy, because everything you need is the only one rule for some template, i.e. dir/*.dat. In DVC I don’t know how to do it without N identical “runs” which makes dvc.json non-human-readable and can cause errors if developer forgot to add new runs when data is added