[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: [freehaven-dev] Re: [Freenet-chat] MojoNation

On Wed, Aug 09, 2000 at 09:45:33PM -0500, Brett Wooldridge wrote:
> > But I'm not convinced that Freenet needs it. It has a rudimentary defense
> > system against spamming already, both against publishing (unpopular data
> > dies out) and against retrieving (it caches where you retrieve it, so
> > it's harder to attack many servers at once).
> I think there is still a problem.  Regardless of the popularity of the data, it is
> my understanding that publishing something "new" to a node can and will, if
> the node is at capacity, push "old" data out -- least popular first obviously.
> But, if enough "new" material is published -- especially with very similar keys,
> the effect would be to push all valid data out of a given node.  Someone
> correct me if this is not the case.  One solution would be to treat "new"
> data as (infinitely) unpopular, and therefore reject it if the node is at capacity.
> But this could have the effect is rejecting legitimate publication.  Is there a
> mechanism in Freenet for a server to "split" its keyspace off to another node?
> Thus making room for more data?  Even still this wouldn't prevent excessive
> publication/splitting of keyspaces by spammers.
> brett
you're trying to micromanage the system. you should build a set of robust
primitive behaviors and hope complex behaviors emerge. don't try to add
special-case rules like 'if you get near full, then contact a neighbor
and negotiate with him' -- it will make your system fragile, difficult to
maintain, and very difficult to reason about.

confer http://freehaven.net/archives/freehaven/dev/Feb-2000/msg00001.html
(and my response) for another example of trying to micromanage a solution
to a similar issue.

freenet doesn't need key splitting notions -- besides, the keys are found
to be close (in hashspace) compared to IP, so a 'nearby' neighbor would
be somebody whose IP hashed to a similar hash..that doesn't seem very
intuitive to me.

freenet defends itself from this sort of flooding attack by saying that
sufficiently popular data will be elsewhere too, and so you won't be able
to kill it; insufficiently popular data deserves to die. 'insufficiently'
is defined as a function of the degree of attack on the document.

in my opinion, this makes freenet less useful as an eternity-like service.
it is designed as a broad publication/availability service that gets
the most popular documents out there and unkillable. it really does seem
optimized for distributing mp3's and porn. [please don't respond to this
paragraph on freehaven-dev. i'd rather keep this technical.]

which has its uses. but this is why there are other projects out
there too.

freenet's (current) level of anonymity is also optimal for grey
areas like mass copyright infringement, as opposed to more specific
activities (where catching a single person is very profitable) like