ASGIHandler fully buffers request before handling it

[Klaas van Schelven]

The ASGIHandler ​reads the full request body before proceeding with actually "doing the Django thing" (processing the request with a view).

In this aspect the ASGIHandler has less support for streaming than the WSGIHandler.

When running Django as a WSGI application, I am able to start reading from a request (using request.read(chunk_size) as it becomes available to my web server. This enables me to process large requests in a streaming way (lower latency, support for abort-early). When running Django as an ASGI application (at least as per the provided asgi.py) I am unable to do this.

This is contrary to my general expectation that things starting with an 'A' should have equal or better support for streaming than things starting with a 'W'. The following line from the ASGI spec gives credit to that expectation "ASGI attempts to preserve a simple application interface, while providing an abstraction that allows for data to be sent and received at any time"

[Carlton Gibson]

Duplicate of #33699. This behaviour is mandated by the ASGI spec. See comment:6 for details.