HTTP client sends "Tranfer-Encoding: chunked" header with empty requests

[bclozel]

Sending any requests with the HTTP client, without a "Content-Length" header set, results in sending a "Transfer-Encoding: chunked" header - even if the client is sending a GET request with no body at all.

This was previously addressed properly (see #119) but that part has been moved to AbstractHttpConnectionBridge.

This header should only appear if the "Content-Length" is not set and the request has a body.

[NiteshKant]

@bclozel does it violate the HTTP spec if the transfer-encoding is chunked but there is no content?

[bclozel]

I'm not 100% sure about this, since the spec sometimes refers to messages with "empty body", but I've found this that at least suggests that a body should be present:

The presence of a message body in a request is signaled by a
Content-Length or Transfer-Encoding header field. Request message
framing is independent of method semantics, even if the method does
not define any use for a message body.

The next section does not strictly say that Transfer-Encoding: chunked requires a body, but there's still:

If a Transfer-Encoding header field is present and the chunked
transfer coding (Section 4.1) is the final encoding, the message
body length is determined by reading and decoding the chunked
data until the transfer coding indicates the data is complete.

and

If this is a request message and none of the above are true, then
the message body length is zero (no message body is present).

Which both suggest that for an empty request, we should have either nothing or a Content-Length: 0 header.

[NiteshKant]

As with any other RFC, I think two people interpret it differently :)

In order to make sure that we do not put Transfer-Encoding: chunked for empty bodies, we would have to delay writing headers until the content stream emits an item or terminates. Although possible, I think it does not achieve much benefit. A server that handles chunked encoding, must be able to handle such requests correctly if the request is correctly having the last chunk.

I would be interested in seeing if there are any servers that restricts such requests. I haven't seen any.

[bclozel]

I ran into this issue while testing the client with a MockWebServer instance.

[NiteshKant]

@bclozel see #520
We now have a case where sending empty chunk is an issue :)

[NiteshKant]

Fixed via #522