On Thursday 06 May 2010, Dawit A wrote:
> On Thursday, May 06, 2010 04:31:30 David Faure wrote:
> > > #3. When internal metadata is received, update ALL ioslaves, not only
> > > the idles ones, that match the criteria specified through the keyword.
> > 
> > I guess there's no other solution, but this sounds a bit like a race.
> > A running slave might or might not get this new metadata in time for
> > whatever it's currently doing. Then again, we're talking about
> > authentication, so obviously we can't send auth data before it's typed in
> > by the user :-) OK, ignore this.
> 
> Well, even for circumstances other than authentication, there is nothing we
> can really do about an ioslave that is already in the middle of processing
> a request. We just want to ensure that it gets updated for the next
> request, that includes the ioslave that originally sent the internal
> meta-data.

Right.

> For authentication this is actually not an issue because not only do you
> have to wait for user input as you stated, but also subsequent requests
> into the protected site are completely dependent about the first request.
> IOW, if you do not get the first page, well you cannot parse it and make
> other requests.

Yes -- well, one can also imagine a page where some iframes or images etc. 
need auth, but anyway, we already handle this in kpasswdserver, so no problem.

> Updated patch attached...

Looks good to me.

-- 
David Faure, faure@kde.org, http://www.davidfaure.fr
Sponsored by Nokia to work on KDE, incl. Konqueror (http://www.konqueror.org).