Dear Evgeny,

My primary question is around the fact that libmicrohttpd implements the http 
102 code as MHD_HTTP_PROCESSING.  The only meaningful use of this code is to 
transmit it to the client and then send a follow up response when the 
processing is complete.  There is no other defined behaviour for this code.  
So, what I'm trying to deduce is: Does libmicrohttpd actually implement this 
functionality?  If so, how would one go about doing this?  If not, why is the 
code implemented?  Should it be removed from the library?

Thanks again for all your comments.  I'm sorry if I wasn't clear before but I 
am not working from obsoleted specifications, in RFC 4918 it mentions that the 
for HTTP 102 continue to refer to RFC 2518, not that the code is no longer 
supported.  The RFC was obsoleted, not the the code.  It is still registered 
with IANA and is still supported with all major browsers.  I appreciate your 
attempts to educate me, but http1.1 supports what I am trying to do, I'm just 
trying to figure out if libmicrohttpd does, as the definition of the status 
code would seem to imply.

Regards,
Daniel


________________________________
From: libmicrohttpd <libmicrohttpd-bounces+daniel.tweed=mail.mcgill...@gnu.org> 
on behalf of Evgeny Grin <k...@yandex.ru>
Sent: Thursday, December 1, 2016 9:18:17 AM
To: libmicrohttpd@gnu.org
Subject: Re: [libmicrohttpd] Out of band authentication with libmicrohttpd

To make it clear: MHD follow HTTP specifications and do not allow to use
two responses for single request.

--
Best Wishes,
Evgeny Grin

On 01.12.2016 17:07, Evgeny Grin wrote:
> On 01.12.2016 16:30, Daniel Tweed, Mr wrote:
>> On 01.12.2016 3:04, Daniel Tweed, Mr wrote:
>>>> I'm really new to libmicrohttpd and the examples/tutorial have been
>>>> really helpful, but I can't figure out how to implement one specific
>>>> thing I need.
>>>>
>>>> What I am trying to achieved is:
>>>>
>>>>  1.
>>>>     Send http 102 to client who requested anydoc.html which requires
>>>>     authentication
>>> HTTP standards doesn't define response code 102.
>>> See https://tools.ietf.org/html/rfc7231#section-6
>>> and https://tools.ietf.org/html/rfc2616#section-6.1.1
>>
>> It was defined in RFC 2518, which was updated/obsoleted by RFC 4918 for
>> WebDAV.  Granted they removed http 102 from the specification (due to a
>> lack of implementation) and advise that its IANA registration should
>> continue to refer to RDF2518.
>> At a minimum, Firefox and IE respond correctly to this status code, and
>> it is implemented in libmicrohttpd as MHD_HTTP_PROCESSING
>
> I highly recommend you not to use obsoleted specifications - support in
> existing application can be dropped in any moment and new application
> most probably will not work with this code.
> So you will need to re-do your work one more time when clients
> unexpectedly start to refuse your server responses.
>
>>>>  2.
>>>>     Some processing on the serverside, including out of band
>> authentication
>>>>
>>>>  3.
>>>>     If authenticated, send response built from somedoc.html, otherwise
>>>>     generic not auth message
>>
>>> HTTP use request-response logic. If you already responded (your code
>>> 102) to some request then you can't add another response later.
>>
>> The purpose of the http102 message is so the client will not timeout
>> when waiting for a response if the request will take a long time.  Using
>> the MHD_HTTP_PROCESSING code requires some mechanism to first send a
>> response with this code and then send a final response.  I just figure
>> out how or if this functionality has ever been implemented in the library.
>
> These is not grantee that client will wait more after 102.
> Moreover, chances of successful response even lower, if you take into
> account obsoleted nature of this status.
>
>>>> I'm having a failure of understanding somewhere, in that I cannot seem
>>>> to figure out how to send the 102 and save the connection details so
>>>> that I can forward the response in step 3.   I had thought I could queue
>>>> a response then either enter a wait loop or suspend the connection, but
>>>> as far as I can tell I have to return from the
>>>> |MHD_AcceptPolicyCallback| for the response to be sent.   Then I cannot
>>>> figure out how to get back to the connection as I cannot suspend it and
>>>> save the pointer. I have looked at the request completed call back but
>>>> this still results in the 102 not being sent.
>>>>
>>>> I really can't figure out a process to achieve these steps from the
>>>> examples or the manual.  I'm sure I'm either missing something about
>>>> http processing or about libmicrohttpd any help or advice would be
>>>> appreciated.  I'm happy to share anything that would make my question
>>>> clearer, including more details on the overall application or specific
>>>> code I am working with.
>>
>>> MHD_AcceptPolicyCallback could be used to choose whether to process
>>> connection or does not process connection at all.
>>
>>> You should ether call MHD_queue_response() from your callback
>>> MHD_AccessHandlerCallback specified in MHD_start_daemon() parameter or,
>>> if your application need some time to generate response - call
>>> MHD_suspend_connection(). When application is ready to generate response
>>> - call MHD_resume_connection(), then MHD will call again your
>>> MHD_AccessHandlerCallback where you can call MHD_queue_response() to
>>> provide response to client.
>>
>>> If you need some kind of authorization, you can use MHD built-in functions.
>>> See examples:
>>> src/examples/authorization_example.c
>>> src/examples/digest_auth_example.c
>>> and
>>> doc/examples/tlsauthentication.c
>>
>> Thanks, I took a look at these but I'm specifically looking at
>> out-of-and authentication.  Since posting this, I found a discussion on
>> a similar topic on the IRC logs to use either keep-alive or
>> MHD_create_response_from_callback ().  My concern is that I cannot know
>> how long the out-of-band authentication will take and I need to advise
>> the client not to time out.  I need the http 102 message for other
>> reasons, but I could work around them, but it seems that this function
>> is implementable somehow, given that the code is defined, so I'm really
>> hoping not to have to rework my other applications around it.
>
> You can use MHD_create_response_from_callback() with MHD_SIZE_UNKNOWN
> and generate response by chunks. This will prevent clients from
> detecting of timeout.
>

Reply via email to