On Fri, 31 Jan 2014, Maciej Puzio wrote:

>> Now I'm back on this train and yeah, the question isn't exactly easy. I'll
>> try to discuss this matter with some guys over the FOSDEM weekend to see if
>> I can get a more firm opinion.
>
> For me personally, any way of fixing this issue would be a step forward.
> Even if you decide to overrule my objections.

Okey, we discussed the matter more over the weekend (mostly me and Dan F) and
I've given it more thoughts.

I'm landing on:

- The current behavior is by design and should be clarified in the
   documentation to reduce the risk of anyone getting surprised.

- CURLOPT_PROXYHEADER should be added to allow applications to set headers
   that are sent to the proxy only

- We need a way make the CURLOPT_HTTPHEADER set only get sent to the host and
   not to the proxy. That option also needs to be reversible. The current
   generation of the patch has this done in a less-than-ideal way and I think
   we should make it less strange and possibly even opt for a separate new
   option. (CURLOPT_HOSTHEADERS_ONLY ? or CURLOPT_HEADEROPT set to
   CURLHEADER_SENDHOSTONLY or something like that)

- This functionality should be possible to get merged really soon, considering
   we have three new tests for it and most documentation already updated.

-- 
  / daniel.haxx.se
-------------------------------------------------------------------
List admin: http://cool.haxx.se/list/listinfo/curl-library
Etiquette:  http://curl.haxx.se/mail/etiquette.html