Let's discuss the performance cost

[paluh]

• To make it better we have to measure it from the start I think so we should provide easy to use performance testing utils as a part of this framework or as a contrib.

• Because we are working on top of the purescript-run I think that we can consider interpretation to different HTTP backends (purescript-warp, purescript-httpureor even to Node.js) directly.

• Thanks to the run we can consider parsing pieces of the request on demand and providing them as an effect. We can even mutate the request object underneath as we won't expose it directly.

• It would be interesting to check how much speedup can we gain by interpreting State (or session API) as a ReaderT { ref :: Effect.Ref | r } Aff.

• Is it possible / feasible to drop Reader all together and provide the context per request by using something like continuation-local-storage mentioned here: https://stackoverflow.com/questions/25151952/global-data-per-http-session-request

• How fast is routing duplex? How much we can tune it?

[paluh]

autocannon seems like a really nice tool for the job. Thanks @Woody88 for suggestion!