1995-09-17 - Re: Anonymous WWW proxies

Header Data

From: Christian Wettergren <cwe@Csli.Stanford.EDU>
To: Laurent Demailly <dl@hplyot.obspm.fr>
Message Hash: 8dc071376daa02785c9983b45580d177ce2fdd8c35fe3792f187e5169ae2476b
Message ID: <199509170517.WAA16625@Csli.Stanford.EDU>
Reply To: <9509162347.AA09904@hplyot.obspm.fr>
UTC Datetime: 1995-09-17 05:18:02 UTC
Raw Date: Sat, 16 Sep 95 22:18:02 PDT

Raw message

From: Christian Wettergren <cwe@Csli.Stanford.EDU>
Date: Sat, 16 Sep 95 22:18:02 PDT
To: Laurent Demailly <dl@hplyot.obspm.fr>
Subject: Re: Anonymous WWW proxies
In-Reply-To: <9509162347.AA09904@hplyot.obspm.fr>
Message-ID: <199509170517.WAA16625@Csli.Stanford.EDU>
MIME-Version: 1.0
Content-Type: text/plain



|  + Chaining would be a imo good idea (ie cli <-> anonproxy1 <->
|    anonproxy2 <-> ... <-> server) but how would you manage to tell
|    your favorite web browser to add in its header something like
|    Http-Proxy-List: anonproxy2, ...
|    An alternative would be to have a database of avaibale (running)
|    proxies and that the proxy itself randomly choose a next route ?

Doesn't most of the browsers support a "firewall-proxy-mode", where
all queries are sent of to a special daemon, that forwards the query
on. This would probably be the place to add the header-munging.

How do you plan to get the reverse-path working? Having a
encrypted/chained return path in the request?

|  + A way to solve previous pb and to add in encryption (but would it
|    be fast enough for web browsing ?) would be that each user runs a
|    local proxy (that could be optionnaly used by other folks) that
|    would do pgp encryption/decryption, 'routing' selection,etc...
| 
| Would ppl with mail remailers and/or W3 experience comment ?
| (or tell me the pointer toward the already solved, already implemented
| beast that would do the above)

Encryption speed isn't all that an issue always. I'm planning to do an
Mbone encryption gateway, (RSN). I will precompute a cryptographic
mask during idle cycles, that can be XORed together with the
clear-text packet when it arrives. I expect it to reduce the latency
quite a lot. (This might not work, since it assumes the key distr
problem is already solved in good time before the packet arrives, to
be able to amass "precomputational power".)

/Christian





Thread