What about a hybrid blocking/timeout approach? At the sending end of the network bridge you have the "taker" and at the receiving end the "putter". The "putter", when it sees an object on the wire, puts it on a local channel (blocking put with timeout). If it succeeds it sends an ack up the wire. If the timeout elapses it sends a heartbeat up the wire instead, then tries again. The "taker", meanwhile, takes from a channel and sends the object on the wire, then waits for an ack or a heartbeat with a longer timeout. If neither arrives within this timeout, it uses another channel to signal the application that the network connection to the "putter" was lost. If a heartbeat arrives it waits again, with the clock restarted. If an ack arrives, it takes another object from a channel and sends it on the wire.
This should cause backpressure to go over the network to the sending end and block the sender, but with the caveat that if the network connection is actually lost, the sender's business logic will discover that shortly, and can distinguish it from the receiver just being slow to process each item. In particular, it can listen for a network drop event on a control channel, and separately use a timeout when putting onto the "taker" channel to react if the receiver is too slow but still connected. On Sat, Jan 25, 2014 at 9:11 AM, Patrick Logan <patrickdlo...@gmail.com>wrote: > In CSP you might have a limited size buffer, but then block on the next > Put. That's not something you want to casually attempt over a distance. It > seems you want an interface like Channels that deal in fully formed > objects, but you don't want CSP blocking semantics. > > -- > -- > You received this message because you are subscribed to the Google > Groups "Clojure" group. > To post to this group, send email to clojure@googlegroups.com > Note that posts from new members are moderated - please be patient with > your first post. > To unsubscribe from this group, send email to > clojure+unsubscr...@googlegroups.com > For more options, visit this group at > http://groups.google.com/group/clojure?hl=en > --- > You received this message because you are subscribed to the Google Groups > "Clojure" group. > To unsubscribe from this group and stop receiving emails from it, send an > email to clojure+unsubscr...@googlegroups.com. > For more options, visit https://groups.google.com/groups/opt_out. > -- -- You received this message because you are subscribed to the Google Groups "Clojure" group. To post to this group, send email to clojure@googlegroups.com Note that posts from new members are moderated - please be patient with your first post. To unsubscribe from this group, send email to clojure+unsubscr...@googlegroups.com For more options, visit this group at http://groups.google.com/group/clojure?hl=en --- You received this message because you are subscribed to the Google Groups "Clojure" group. To unsubscribe from this group and stop receiving emails from it, send an email to clojure+unsubscr...@googlegroups.com. For more options, visit https://groups.google.com/groups/opt_out.
