single-process nonblocking operation

This type of custom server handles a request to completion, or to such a state where it is waiting for an asynchronous event and is free for some other activity. This means that requests that can take a comparatively long time to return do not introduce performance limiting delays when using multiple channels simultaneously.

In this model, the main loop normally uses CA_Poll() in place of CA_Receive_DT_Msg(). The custom server will need to wait not only for requests from channel processes, but also for other events, such as voice data becoming available for a CA_Record_Voice_Stream(), or interprocess communication from other processes or programs. CA_Receive_DT_Msg() is called only when CA_Poll() has returned the information that requests from a channel process are waiting, and is normally called with CA_NOWAIT set.

In this model, the custom server is not handling a single request to completion, but is swapping between a number of requests, so it will normally need to maintain some data for each open link.

Advantages

The advantages of single-process nonblocking operation are that:

It is a relatively simple design with single process.
It can handle requests from multiple channel processes without blocking.
It can be used by custom servers that access voice channels directly.

Disadvantages

The disadvantages of single-process nonblocking operation are that:

It cannot use a system-generated main function.
Although the custom server does not block, it has only a single process, so access to it is still serialized. Intensive processing for a request from one channel process can deny access to the custom server by another channel process for an unacceptable time.

Single-process nonblocking operation

Advantages

Disadvantages