Keeping CALM: when distributed consistency is easy

Re:Keeping CALM: when distributed consistency is easy by elierotenberg on Wednesday March 06, 2019 @07:45AM (#19318442)

It seems like monotonicity (or a subset, stronger monotinicity, still covering many practical use cases) can be statically enforced in an algrebraic type system. With some generics magic you can encode monotonicity as a type assertion on (optionally asynchronous) function calls mapping an underlying transport/RPC mechanism (REST over HTTP, RPC over Websocket, custom protocol over TCP/domain socket...).

In TypeScript for example you can strongly type your communication protocol (web requests/reponses, websockets, parent/worker ipc, etc) and get some guarantees at compile time. For example, you can guarantee that your state updates over the wire are eventually consistent.

The same goes of course for other algebraic type systems by encoding monotonicity in the form of static type assertions (Flow, Caml, Haskell...).

[ Parent | on HN ]

Re:Keeping CALM: when distributed consistency is easy by marcosdumay on Wednesday March 06, 2019 @11:00AM (#19319874)

Well, since you just need to restrict the operations one can do with data, even a Java or C++ style type system is enough.

What is best done with another language (or a complex library on a very malleable language) is permitting non-monotonic operations with a clear boundary between them and some code style that makes them undesirable.

[ Parent | on HN ]

Re:Keeping CALM: when distributed consistency is easy by achou on Wednesday March 06, 2019 @12:53PM (#19320951)

Interesting idea. Can you elaborate on the TypeScript comment? How would you encode something like this in TS?

[ Parent | on HN ]