There are a lot of fuzzy edges. One big one is managing different aspects of "sameness" for a file in different contexts that makes sense. You mentioned the pro

There are a lot of fuzzy edges. One big one is managing different aspects of "sameness" for a file in different contexts that makes sense. You mentioned the problems with binary hashes; there are others, like comments in scripts and config files. If distro-X just spams boilerplate at the top of the config file, is that the "same" file? What if they add or rewrite informational comments in a config file? I think the answer is that sometimes they are; it depends on the question. I also noticed that it is really easy to end up wanting to pull data from source repos and bug trackers, but down that path lies madness. Anyway, I don't have any plans to start building this. But it is an interesting project to think through.