But it can be done.
Speaking of sort.Interface, people say they want generics, but I think there is an implicit ask for operator overloading, too. Or the code they are asking to not write is probably still going to be there.
1. So you have a map of sort.Interface, and that's just as absolutely-not-convenient-or-type-safe as a map of interface{}
2. TFA does not want a sorted map, they want an ordered map, so sort.Interface has no relevance whatsoever: the order of iteration of an ordered map is the order of insertion, not the sort order of the elements.
> Speaking of sort.Interface, people say they want generics, but I think there is an implicit ask for operator overloading, too. Or the code they are asking to not write is probably still going to be there.
Not really. Java does OK without operator overloading. Having to use methods instead of operators can be annoying in some cases (mostly when working with non-built-in numerics e.g. decimals or infinite-precision integrals) but it doesn't limit you much if it does at all.
This is the opposite of DRY.
I often read criticism of Go which is based on a missing "feature". In my experience, implementing these features is surprisingly easy in Go, as the building blocks are all there. In this respect, Go has some resemblance on Lego :).
My background contains a lot of Scheme/Lisp, so abstracting things with functions comes natural to me. It turns out that Go has basically the equivalent function concept (first class functions, closures, ..) so enables all the powerful abstractions coming from them.