GitLab

Many HTTP frameworks will automatically encode data streams using the aforementioned method. I propose this behavior for Icecast:

• check if the protocol is not HTTP/1.1 or there is no Transfer-Encoding header, continue as in the past (i.e. assume identity encoding)
• if the Transfer-Encoding header is present and it is identity, continue as in the past
• if the Transfer-Encoding header is present and it is chunked, accept the data and strip the encoding information both from the output stream and from the dumpfile
• if the Transfer-Encoding header is present and has a different value, terminate the request with 501 (Unimplemented) and provide an HTTP response body listing the supported encodings (in case developers need to debug this).

That behavior in compliant with RFC2616 (Section 3.6) and RFC7230 (Section 3.3.1).