new pipelets loop

pipelets_loop.fs

    let computeAndRoute data routes = async{data |> transform |> router <| routes}

    let mailbox = MailboxProcessor.Start(fun inbox ->
        let rec loop routes = async {
            let! msg = inbox.Receive()
            match msg with
            | Payload(data) ->
                ss.Release() |> ignore 
                Async.StartWithContinuations(computeAndRoute data routes, ignore, errors, ignore)   
                return! loop routes     
            | Attach(stage) -> return! loop (stage::routes)
            | Detach(stage) -> return! loop (List.filter (fun x -> x <> stage) routes)
        }
        loop [])

I prefer this one, personally. The other's use of RunSynchronously would seem to cause problems. Also, why are you passing ignore into the callback? Is the result expected to be unit?

yep, the other functions are not needed in this instance, the other is the one Don recommended. Its from the MSDN documentation...

Did he mention why he recommended it? Is it better within the agent to behave in a more synchronous manner? What about deadlocks? I know those are possible when calling .Result on a Task.