Post #1823389
2026-04-25 16:48 UTC
@InfiniteNutshell Usually there's an implementation that's trivially safe or trivially live, so either by itself is not that useful. You want it to be both, if possible (though in distributed systems this is often impossible). It's desirable to exclude trivial cases because if you say "we showed that our implementation is safe and live", you want that to really mean something, you know?
Replies (1)
-
@InfiniteNutshell@mastodon.social 2026-04-26 05:57
@lindsey hmm maybe it's that I don't know enough about the context, but my intuition is that "safe" doesn't inherently entail "useful". it seems like the "non-triviality" you want should fall under its own separate property, maybe something we might call correctness or completeness. also, if you're discussing a class of algorithms that captures information about an arbitrary system...well what about the case where the underlying system is trivial? do we want definitions that allow for that?