[RFC] "http-types" breakage / additions / rework

If this is so then it’s a strong argument for requiring tight upper bounds on Hackage packages.

It doesn’t matter. If I need to upgrade a dependency and cabal complains I add an allow-newerin cabal.project and hope it compiles, if it doesn’t I fork the library and fix it (pointing cabal.project to our fork) and move on to the next 30 broken packages. “preemptive” upper bounds don’t really help me.