On Wed, 27 Apr 2005, Augustus Saunders wrote:

>> Hm. What if you just set it to 0 (zero) instead? To me, that seems like the
>> most proper thing to do there to reset it back to original state.
>
> See, the thing is, I want it to keep whatever authentication method it's
> already determined. See, I reuse the curl handle for a bunch of files, so I
> don't want it to redo the authentication dance.

I think you're missing the point of that particular code. It is called on
connection disconnect and those assignments are only made if NTLM has been
used.

NTLM is connection based and thus it must be re-negotiated at next connect no
matter what you or I like!

> FYI, 7.13.1 did not do this, this is a regression.

It didn't do what? The assignment has been there for a year, that is not new.

But sure, bugs still creap in... Can you perhaps tell us how you get to see
this problem so that I can write up a test case for it?

> [regarding SSPI]
>> You need to set CURLOPT_USERPWD to "" (a zero length
>> name/password), and enable NTLM auth.
>
> If I set CURLOPT_USERPWD to "" and use AUTH_ANY, will everything work once
> it figures out that NTLM is required?

Yes. But using AUTH_ANY with a blank user is not very good if something else
than NTLM was picked!

-- 
  Commercial curl and libcurl Technical Support: http://haxx.se/curl.html