Ludovic Courtès writes: > At first sight, to me, it’s not entirely clear how to determine whether > something is to be upgraded in the presence of propagated inputs. I think the linked discussion suggested we checked upgrades recursively. Wouldn't that work? > And more importantly, I think the “imperative” model is necessarily not > as expressive and “clean” as ‘--manifest’. There are tradeoffs to make. Sorry, I don't understand what you mean. How do we upgrade declaratively? I could be missing something big here! :) -- Pierre Neidhardt https://ambrevar.xyz/