Hi,
Thank you very much for so detailed explanation
J
The order is significant in my case due to the regression testing I do on my project. My tests are very strict and include expectations even in the order of HTTP headers. Thus, this change in the
order of the headers involves a huge number of changes (I have thousands of such tests).
However I understand the changes in libmicrohttpd are for good, so it’s something that I should fix on my test regression sooner or later.
Thanks!
Best regards,
------
Fermín
De: Evgeny Grin <k2k@yandex.ru>
Enviado el: miércoles, 11 de enero de 2023 7:53
Para: libmicrohttpd development and user mailinglist <libmicrohttpd@gnu.org>
CC: FERMIN GALAN MARQUEZ <fermin.galanmarquez@telefonica.com>
Asunto: Re: [libmicrohttpd] Response headers order change between 0.9.70 and 0.9.75
Hi Fermín,
MHD internally has two types of response headers: user headers and automatic (mandatory) headers.
For automatic headers MHD now follow RFC recommendation: "...it is good practice to send header fields that contain additional control data first, such as Host on
requests and Date on
responses" (See
RFC9112, 5.3-5). Probably it would make sense to move "Content-Length" and "Transfer-Encoding" headers to the top as well.
For user headers MHD recently fixed the wrong output order. Before the fix, user headers were sent in as reversed sequence relative to the order of headers addition (LIFO). Currently MHD sends user headers exactly in the same order as user added them (FIFO),
with documented exceptions, like special handling of the "Connection" headers. The Changelog entry is
here.
However, these changes are mostly cosmetics, as RFC requires "The order in which field lines with differing field names are received in a section is not significant."
(here) and "...a sender MUST
NOT generate multiple field lines with the same name in a message..." (here). Previous RFC versions have similar requirements.
Anyway, for user defined headers you have full control of the header order: the first added header will be sent as the first user header.
Could you share more details about the situation where the order of the headers is important?
--
Wishes,
Evgeny
10.01.2023 20:17, FERMIN GALAN MARQUEZ wrote:
Hi,
I’ve recently upgrade from libmicrohttpd 0.9.70 to 0.9.75 in my project (for reference, this one:
https://github.com/telefonicaid/fiware-orion).
I have observed a curious change of behavior in response headers. In 0.9.70 the server responses like this:
HTTP/1.1 400 Bad Request
Content-Length: 129
Content-Type: application/json
Fiware-Correlator: 86753370-9108-11ed-91ea-000c29583ca5
Date: Tue, 10 Jan 2023 17:02:11 GMT
But in 0.9.75 response is like this:
HTTP/1.1 400 Bad Request
Date: Tue, 10 Jan 2023 17:02:11 GMT
Fiware-Correlator: 86753370-9108-11ed-91ea-000c29583ca5
Content-Type: application/json
Content-Length: 129
That is, same headers but in different order.
At this point, I have a couple of questions, please:
1.
Why this has changed between 0.9.70 and 0.9.75? I have tried to find the reason in the changelog at
https://github.com/Karlson2k/libmicrohttpd/blob/master/ChangeLog but it is a bit confusing…
2.
Is there any way of keeping the older behavior?
Any feedback is welcome. Thanks in advance for that!