> In 2.0, you can insert "filters", which may parse the content, and
> optionally modify it.
> It works great with http->http, http->https, and https->http.
> However, it doesn't work with https->https, because there is no point
> of time, during the whole process, that the content is readable.
> I'm sure it's possible to implement it, because all the needed
> functionality is already implemented either in http->https or
> https->http, and the new https->https will have to only combine these
> (en/de)cryptions.

The fundamental problem is that both the proxy server and the target server
would have to know the SSL encryption and decryption keys.  These are
different for each session, so the relationship between the proxy and the
target would be hard to maintain.  At the start of each session, one of them
would negotiate the session keys with the client, then tell the other what
they are.  One proxy could be handling many targets, of course (and vice
versa), which compounds the problem.

The problem is further compounded by a bug in the IE browser.  A browser
should send keepalive messages when things are quiet to keep the SSL session
open, but IE doesn't, so the session is highly likely to time out before the
user sends the next page request.  The upshot of that is that just about
every page request from an IE browser involves creating a new session.  This
issue doesn't make the problem any harder to solve, but it means that a busy
proxy and target will have to share new keys many times a second.

Simon

Reply via email to