Re: Auth framework questions (2.5)

Offhand
        that looks fine. auth user requests are refcounted, so nothing bad
should happen in any case.

Rob
On Mon, 2003-05-19 at 04:13, Henrik Nordstrom wrote:
> Hi Robert,
>
> I am looking into fixing some of the rough edges of Digest
> authentication (see Bug #630 for details) and as part of this I had
> to make a small modification in the auth framework to be able to
> indicate stale digest nonce.
>
> My question to you is if there is any ill effects from doing this:
>
> in authenticateAuthenticated()
>
> switch (authenticateDirection(*auth_user_request)) {
> case 1:
> + if (!request->auth_user_request) {
> + /* lock the user for the request structure link */
> + authenticateAuthUserRequestLock(*auth_user_request);
> + request->auth_user_request = *auth_user_request;
> + }
> + /* fallthrough to -2 */
> case -2:
> /* this ACL check is finished. Unlock. */
> authenticateAuthUserRequestUnlock(*auth_user_request);
>
>
> The purpose of this is to have the auth_user_request sent to the
> header fixup when the authentication scheme indicated a new challenge
> has to be sent.
>
> >From what I can tell this should not affect NTLM authentication as it
> should en up as the same as conn->auth_user_request and probably is
> what is expected, but the exact details of this relation is a bit
> unclear to me and without a test environment available I cannot
> verify.
>
> Basic auth does not use the auth_user_request in header fixup and
> should not notice at all.

-- 
GPG key available at: <http://users.bigpond.net.au/robertc/keys.txt>.