The Haskell community could be justifiably accused of navel-gazing, that is, spending inordinate amounts of time designing beautiful programming models and APIs, while spending relatively less time building applications that solve real problems. In this talk I want to take the opposite approach: start from an application that we want to build---a simple concurrent chat server---and use that to inform some library design. Along the way we’ll learn useful techniques for building concurrent network applications, and show how the power of Software Transactional Memory can be brought to bear. Yet this application demonstrates some holes in the abstractions provided by Haskell as it stands, so I will show a simple layer called “async” that fills the gap and should be widely useful for concurrent programming.

No outline is available