Talk: (English)
Purely functional distributed programming for collaborative applications
Wouldn’t it be wonderful if we could program distributed applications without having to send messages around?
In this talk I’ll introduce a purely functional programming model which I’m calling relativistic functional reactive programming (RFRP) which enables precisely that.
Where the original functional reactive programming model semantically described programs as functions from time to output value, RFRP takes locality into account by defining programs as functions of points in spacetime to output value.
In this ongoing work I’m applying RFRP to eventually consistent distributed programs.
Therefore I’ll first introduce the basics of eventual consistency by defining useful RFRP primitives, then explain how composition of these primitives can be used to precisely and succinctly describe the semantics of replicated data types (which make programming eventually consistent programs easier), and finally I’ll show how RFRP can be used to define complex collaborative applications without any concurrency-related surprises. Everything will be illustrated using code samples in my prototype RFRP library written in Haskell.
Adriaan Leijnse
I’m in love with finding ways to express correct programs as succinctly and clearly as possible. I also dream of a future in which people own their data and can collaborate without corporate intermediaries. These two goals are being combined in my PhD studies for which I’m developing a purely functional programming language for peer-to-peer collaborative applications.
My PhD studies are at the Universidade NOVA de Lisboa supervised by professors Carla Ferreira and Annette Bieniusa. My work is sponsored by Protocol Labs Inc.
- Slides
- adriaan-leijnse.pdf