The code example in getting_started/mix_otp ch 8 about elixir-lang.github.com

[josevalim]

@tallakt Thanks for the feedback. Can you please detail why it is confusing? The final solution needs to have two recursive loops and that's how an actual web server / tcp acceptor would work. The acceptor needs to loop for new connections and the multiple request/responses can be sent in the same connection. Maybe we need more detailed comments?

[tallakt]

I just looked at the code and could not make out why you had two loops, when the inner loop could not ever exit. I guess what I would have expected is that you added the outer loop in the later examples where it became necessary. But i guess it's no big deal.

Another note on the same thing; i think learn you some erlang spawned a new process for the next accept, handling the requests in the original 'accept' process. The reasoning being that you could respond without a context switch. If that code is indeed better, could be an idea to do it like that here as well.. Also no big deal..

[josevalim]

@tallakt I see your point now. I will check which approach is better, to add the outer loop later or at the beginning. Adding it at the beginning means we need to do fewer changes later on, but as you described, it may cause confusion. I think I mention early on the fact we get stuck with just one request is a bug that we aim to fix.

Thanks!