PATCH: kio_http

[George Staikos]

On Monday November 11 2002 09:47, Dirk Mueller wrote:
> On Mon, 11 Nov 2002, staikos George Staikos wrote:
> > I can't test it at the moment, but is this not the way it works? 
> > Shouldn't we not send the referrer in that case?  Or should we?
>
> (questions with double negation are hard to read)

  Sorry.  :)

> we should not send a referrer in that case. As I said, this is terrible to
> implement with the current design.

   Which khtmlpart actually sends the KIO request in this case?  Is it the 
part for the frame, or the part for the main page?  If it's the part for the 
frame, then it is broken.  If it is the part for the main page (the toplevel 
khtmlpart), then it is fine because ssl_was_in_use is going to be "true". 

Then we have to consider this case:

Main frame no SSL
    Child frame using SSL

User clicks on link in the main frame.  I think we still send a referrer here.   
Clicking in the child frame should not send a referrer.  This means that the 
only case we send a referrer is if (!ssl_was_in_use && 
!ssl_in_use_in_any_parent).  So I guess we need to propagate a flag down the 
embedded frames, or walk the tree back up at url click/redirect time, to 
determine ssl_in_use_in_any_parent.

Then of course there are all the Java, JS, and Plugin related areas.  Are they 
still affected?


-- 

George Staikos