Here is a summary of the responses that I received to my queries about
chunked POST in HTTP/1.1.
1. All HTTP/1.1 applications that receive entities MUST accept the
“chunked” transfer-coding, thus allowing this mechanism to be used for
messages when the message length cannot be determined in advance.
2. However, an origin server MAY refuse to accept a request without a
defined Content-Length by responding with status code 411 (Length
Required).
3. The Content-Length header field MUST NOT be sent if a Transfer-Encoding
header field is present.
4. An origin server acting as a CGI 1.1 gateway for a request MUST
determine and set the CONTENT_LENGTH metavariable.
5. Origin servers MUST remove the Transfer-Encoding before passing a
request body to a plug-in, servlet, (Fast)CGI, or across any other gateway
boundary.
Origin servers supporting CGI 1.1 have two options when receiving a POST
request with "Transfer-Encoding: chunked" for a CGI 1.1 resource:
1. Reject the request with 411 (Length Required).
2. Filter and buffer the request to determine CONTENT_LENGTH before
passing the decoded request body to the CGI application. If the buffered
request grows too large, the server MAY reject the request with status code
413 (Request Entity Too Large) and the server MAY close the connection to
prevent the client from continuing the request.
Origin servers supporting the Servlet API 2.1 have three options when
receiving a POST request with "Transfer-Encoding: chunked" for a servlet
resource:
1. Reject the request with 411 (Length Required).
2. Filter and buffer the request to determine CONTENT_LENGTH before
passing the decoded request body to the servlet. If the buffered request
grows too large, the server MAY reject the request with status code 413
(Request Entity Too Large) and the server MAY close the connection to
prevent the client from continuing the request.
3. Pass a filtered input stream to the servlet and filter the request body
on-the-fly to remove the chunked transfer-coding. Indicate the end of the
request body with EOF (end of file) on the servlet input stream.
Implications for IPP:
1. The IPP layer doesn't have to deal with chunking. In the context of
CGI scripts, the HTTP layer either rejects a chunked POST request with 411
or removes any chunking information in the received data and supplies
CONTENT_LENGTH.
2. The HTTP/1.1 standard does not guarantee that an origin server will
accept chunked requests, regardless of the resource identified in the
request.
- Carl Kugler