Tony Morris

Doesn’t it depend on context? Aren’t there instances where semigroups can also eliminate edge cases? E.g.: Validation (NonEmpty String) a

I don't have a good answer for the case of Validation

It's the Applicative (Validation e) instance, originally in the paper gives Monoid e => which is problematic, and amended to Semigroup e =>

Sorry I figured @GabrielG439 was familiar. It makes me wonder if there is a dual preference for Semigroup for where he prefers Monoid.

Seems to me that preference shouldn't enter into it. Either that extra state should be representable, or not.

^ This person agrees. Why can't I use do-notation on (Writer (NonEmptyList x)) in Haskell ? I can even with Scala, and that's Scala!

This is analogous to arguing "Why can't I use goto?" Sometimes forbidding functionality improves the code

I'm not using that argument. There is no code that improves by the non-existence of Semigroup (or Monoid or any other appropriate thing).