Re: [libmicrohttpd] How to specify content-length in a HEAD request

[Evgeny Grin]

Hi Rohit,

On 05.11.2023 19:15, Rohit Kota wrote:
> Hello,
> We have a use case where content-length needs to be set in the response 
> headers for a HEAD request. See for example:- 
> https://docs.aws.amazon.com/AmazonS3/latest/API/API_HeadObject.html#API_HeadObject_Examples <https://docs.aws.amazon.com/AmazonS3/latest/API/API_HeadObject.html#API_HeadObject_Examples>
>
> According to rfc-9110 section 8.6.6 
> <https://www.rfc-editor.org/rfc/rfc9110.html#section-8.6-6> this is 
> valid behaviour
> A server MAY send a Content-Length header field in a response to a HEAD 
> request (Section 9.3.2 
> <https://www.rfc-editor.org/rfc/rfc9110.html#HEAD>); a server MUST 
> NOT send Content-Length in such a response unless its field value equals 
> the decimal number of octets that would have been sent in the content of 
> a response if the same request had used the GET method.

Your quote from RFC is absolutely correct and MHD fully follow these 
recommendations. See below for more details.

> But from what I gathered here 
> <https://lists.nongnu.org/archive/html/libmicrohttpd/2019-10/msg00037.html>, I can set MHD_RF_INSANITY_HEADER_CONTENT_LENGTH to allow this behaviour.  

This is incorrect interpretation of the email discussion. The name of 
the option MHD_RF_INSANITY_HEADER_CONTENT_LENGTH suggests incorrect 
behaviour.

> For HEAD requests, I suspect libmicrohttpd is always setting content-length to 0. 

This is not what is written in the documentation.
For *any* request libmicrohttpd *always* set correct Content-Length that 
matches application-provided response, if RFCs allow the Content-Length 
header. Exceptions: 1xx and 204 response codes, where no reply body nor 
Content-Length are used.

> So for example when i set content-length to 24 manually I am getting this.
> < HTTP/1.1 200 OK
> < Date: Sun, 05 Nov 2023 16:10:44 GMT
> < Connection: close
> < Last-Modified: Thu, 26 Oct 2023 04:42:56 GMT
> < Content-Length: 24
> < Content-Length: 0
>
> I suspect the extra content-length is being set by mhd. 

No. The extra "Content-Length" header with size 24 as added by 
application. The application allowed to add this header because of 
MHD_RF_INSANITY_HEADER_CONTENT_LENGTH flag.
See 
https://git.gnunet.org/libmicrohttpd.git/tree/src/include/microhttpd.h?h=v0.9.77#n3882 
for manually adding wrong "Content-Length" header.
The "Context-Length" with number zero is used because application 
provided response with zero length.

The next MHD release will include clearer documentation and more options.
See 
https://git.gnunet.org/libmicrohttpd.git/tree/src/include/microhttpd.h?id=cdf9b66c61a444c09ba7c9f4cee385cbd2b3bcdc#n3575
https://git.gnunet.org/libmicrohttpd.git/tree/src/include/microhttpd.h?id=cdf9b66c61a444c09ba7c9f4cee385cbd2b3bcdc#n3728

> But as per rfc I should be able to set content-length manually for a head 
> request.

It is very easy: just provide the correct response object with correct 
context (response body). For HEAD requests MHD will set the 
"Context-Length" to the right number, but *will not* send the response 
body. If the same response object is used for GET requests, MHD will set 
the "Context-Length" *and* *will* send the response body.

The idea is that response objects are unified, can be created one time 
and re-used for both GET and HEAD requests freely.

The next MHD version will provide more flexibility and special responses 
that can be used only for HEAD requests. This functionality is already 
in git branches.

--
Evgeny

OpenPGP_0x460A317C3326D2AE.asc
Description: OpenPGP public key

OpenPGP_signature.asc
Description: OpenPGP digital signature