Hi

With this new authentication support (Digest, GSS-Negotiate and NTLM) we've
introduced a new problem to libcurl which I think we need to address somehow:

They all deliver one or more pages with contents that isn't what the user of
libcurl wants but is delivered anyway... libcurl currently has no concept of
"hiding" data or silently ignoring it. What to do?

NTLM is a fine example:

Currently it works like this (one round-trip could probably be moved out):

 1) GET

 2) 401 returned (with page contents)

 3) GET

 4) 401 returned (with page contents)

 5) GET

 6) 200 returned, with the contents we care about

So, in step 2 and 4 (or if we cut out 1 we won't get number 2, only 4) we get
contents we don't need, like or want. We need to read it since we want to
re-use the same connection.

Should I simply introduced another state to the debug callback and provide
all that data to it and smile, or is there anything more to it than I'm
thinking of right now? This would of course only hide the data if everything
is working OK, if anything goes wrong in the authentication it would produce
the data "as usual" when a bad HTTP password is used.

Thoughts? Other approaches?

-- 
 Daniel Stenberg -- curl: been grokking URLs since 1998
-------------------------------------------------------
This SF.NET email is sponsored by: eBay
Great deals on office technology -- on eBay now! Click here:
http://adfarm.mediaplex.com/ad/ck/711-11697-6916-5