A note on depth-first derivations