OTOH, "Types are not sets" from *1973* has a better discussion of types than 2025 Tiger Style.

This paper is so old it was written on a type writer, and it's still a banger.

dl.acm.org/doi/pdf/10.1...

Code Complete was great for its time, and a lot of the advice is still applicable, but we've also made a lot of progress since then.

You're probably one of the saner ones. I think "technology for the self" is really under-rated. Journaling isn't just for angsty teenagers, meditation not just for woo types, etc.

Lots of industry efforts have tried to make cross system type systems (this is what GraphQL, OpenAPI, etc. are about) but they are usually awful: inexpressive and imprecise semantics.

There's opportunity here for interesting, but hard, research.

However, the constraint is shared between Rust code and a ClickHouse query. Hence it's outside the purview of any one system's static analysis.

This is often the case in Internet software. Choreographic programming is not sufficient, IIUC, as it assumes systems are all in the same language.

That episode could have really gone in a different, more philosophically inclined, direction when Sonny asks "it was real, wasn't it?"

What I'm doing is trivial with data, but I want to work with interfaces---i.e. codata---not data. So is modular programming (in the SML functor sense) the dual of standard value composition?

It feels that way, but the machinery is baroque. For example, modules (via inheritance) share a namespace.🤔

Both sides are political. Those who object to restrictive licenses should ask themselves why their believe they have the right to force their politics on the code authors.

(The answer, in the case of the posters I recognized, is that they're right-wingers who want to enforce their hierarchies.)

Nothing necessarily wrong with it, but if there is an easier to use API I would prefer that