What is sioscgi?

sioscgi is an implementation of the Simple Common Gateway Interface (SCGI) protocol under the Sans-I/O philosophy.

What is SCGI?

SCGI is a protocol used for communication between HTTP servers and Web applications. Compared to CGI, SCGI is more efficient because it does not fork and execute a separate instance of the application for every request; instead, the application is launched ahead of time and receives multiple requests (either sequentially or concurrently) via socket connections. Compared to FastCGI, SCGI is a much simpler protocol as it uses a separate socket connection for each request, rather than including framing within a single connection to multiplex requests (a feature which is rarely used in FastCGI anyway due to the lack of per-request flow control).

See the Wikipedia and Python SCGI pages for more information.

What is Sans-I/O?

Sans-I/O is a philosophy for developing protocol processing libraries in which the library does not do any I/O. Instead, a user of the library is responsible for transferring blocks of bytes between the socket or pipe and the protocol library, and for receiving application-level protocol items from and sending them to the library. This obviously makes a sans-I/O library a little more difficult to use, but comes with the advantage that the same library can be used with any I/O and concurrency mechanism: the same library should be usable in a single-request-at-a-time server, a process-per-request or thread-per-request blocking server, a server using select/poll and continuations, or a server using asyncio, Twisted, or any other asynchronous framework.

See SansIO for more information.

How do I install it?

sioscgi’s releases are published on PyPI for installation through pip. You can run pip install sioscgi.

For development, the source is available at GitLab and GitHub.

How do I use it?

In general terms, as follows:

1. Accept an SCGI connection from the HTTP server (or other SCGI client).
2. Construct an SCGIConnection object.
3. Receive the request from the SCGI client by repeating the following until a RequestEnd event occurs:
   1. Read some bytes from the connection and pass them to SCGIConnection.receive_data.
   2. Call SCGIConnection.next_event to receive high-level events corresponding to the received data (one RequestHeaders, zero or more RequestBody events, and one RequestEnd).
4. Send the response to the SCGI client using SCGIConnection.send, sending first a ResponseHeaders event, then zero or more ResponseBody events, then a ResponseEnd event, and sending the returned bytes over the connection.

This being a sans-I/O library, how exactly you implement each step will depend on what I/O and application framework you’re working under. For example, for a thread-per-request or process-per-request server, you would likely do a blocking receive from a normal function in step 3.1; in an asyncio-based server you would instead await new data from a coroutine.

For detailed information about the classes and methods available, see the module documentation provided in the docstrings by running import sioscgi followed by help(sioscgi).