[Mnet-devel] RE: [p2p-hackers] Freenet, Mnet, GNUnet, Circle

[Greg Bildson]

Well, I expect that the queries that Gnutella experienced were somewhat
unique in that our changes could not be applied to your situation.  Our
searches were a mixture of keyword searches and SHA1 hash searches.  The
major problem we had was caused by clients that would automatically requery
for more sources of a file.  Few developers could see a problem from
requerying but in practice we found that any more than one requery per hour
total per client would cause massive network traffic.

Our initial approach to reducing the problem for LimeWire users was to group
LimeWire clients together in the network and thus somewhat shield ourselves.
This worked to some extent but still the majority of traffic appeared to be
requeries.

Given that requerying is not the primary function of the network, we had
great freedom in taking draconian action.  We observed that requerying was
either done using SHA1 searches or by almost the entire file name of a file.
Gnutella's primary interest was in allowing user typed queries to travel
through the network, so we eventually just started blocking all queries that
we did not feel a user would type.

I'll leave the details to your imagination but for a long time this was an
extremely successful approach.  The problem was reduced further with our
introduction of dynamic querying which basically did away with unconstrained
(other than TTL) broadcast queries.  Only after obliterating the bad actors
did we allow SHA1 searches back on the network.

We continue to try and enforce a policy of no automatic requerying on the
network including in our own client.

Thanks
-greg

-----Original Message-----
From: p2p-hackers-bounces at zgp.org [mailto:p2p-hackers-bounces at zgp.org]On
Behalf Of Ian Clarke
Sent: Thursday, July 15, 2004 10:58 AM
To: Peer-to-peer development.
Cc: development of Mnet
Subject: Re: [p2p-hackers] Freenet, Mnet, GNUnet, Circle


Greg Bildson wrote:
> We spent a year dealing with those "spammy client" issues on Gnutella.
Most
> of the offenders ended  up dying an early death after we devised both
> specific and general mechanisms for dealing with offenders.  Having a
> rapidly evolving protocol helps to some extent.

I am curious as to what the general mechanisms were.  Clearly, it would
have been easy enough for us to hardcode some limits in a user's freenet
node to prevent them from spamming the network, but since they have
access to the source code, they could easily remove these.  We wanted a
network-wide solution that would remove the incentive to spam the
network but that didn't rely on a user being policed by code that they
are free to modify.

Ian.
_______________________________________________
p2p-hackers mailing list
p2p-hackers at zgp.org
http://zgp.org/mailman/listinfo/p2p-hackers
_______________________________________________
Here is a web page listing P2P Conferences:
http://www.neurogrid.net/twiki/bin/view/Main/PeerToPeerConferences