On Wed, Dec 27, 2000 at 11:02:42AM -0800, Mr.Bad wrote:
> >>>>> "OS" == Oskar Sandberg <[EMAIL PROTECTED]> writes:
>
> OS> However, in regard to practical viability of the suggestion in
> OS> question, I side completely with Ian and Scott's
> OS> objections. [...] I have not seen one hint of an actual
> OS> refutation of this in the arguments given, and I'm afraid that
> OS> the burden of proof falls on Brandon and Mr. Bad here - there
> OS> are several bases for simulators out there (Serapis, that
> OS> which Theo wrote for his analysis, and little birds whisper
> OS> tails of ones greater still), so there is nothing stopping you
> OS> from proving that artificial separation of nodes does not
> OS> disturb the (at least somewhat) natural order of the network.
>
> Well, I don't think anyone is suggesting that implementing clusters
> and/or "Don't Talk to Strangers" for nodes is going to leave NO trace
> on the network. 100% fully-connected networks (every node is aware of
> every other node) are going to be the fastest, and we just can't match
> that with clusters or protected nodes, ever.
>
> I guess I'd want to know how much we're willing to trade off in
> performance and flexibility of the network for some modicum of
> security -- however small -- for the node operator. Is node operator
> security such a non-goal, and performance so overarchingly important,
> that NO COMPROMISE WHATSOEVER will be allowed? It doesn't seem worth
> doing a simulation if that's the case.
>
> However, what if we can show that protecting nodes and providing
> clustering will give (say) 40-60% of the response speed of an
> unclustered, path-folding network, without loss of data? Would that be
> sufficient to implement? What's the magic number?
As with many things in this Freenet, it is not as simple as implementing
something and taking a calculated performance hit. All the simulations
that have been done has showed us that even small changes to how the
network expands and self organizes can have a dramatic effect on it. We
are not talking about a 40-60% drop in performance here - we are talking
about the difference between a working Freenet and one not significantly
more efficient (or possibly even much worse) then simply sending requests
around randomly.
If I believed that Freenet could work in a manner where nodes constrained
their communications I would jump at it, even if it meant a considerable
cost to performance. Without putting words in there mouth I think that so
would Ian, Scott, and Adam. But, having spent the last year of my life
considering issues with Freenet's routing, and having myself been
considering "shy" nodes since pretty much the beginning (check the
archives, we discussed this in February last year), I am quite convinced
that this will put you in a situation where you are corrupting the very
process that makes Freenet scalable and dynamic.
> OS> With 0.4 we had already planned to close nodes off to anyone
> OS> who does not know the public key, which will make it
> OS> considerably more difficult to scan targeted individuals and
> OS> computers and attempt to entrap them.
>
> Is the public key given out automatically? Or can I only give out
> public keys to node operators I trust?
It's a _public_ key. It is given to every node that needs to communicate
with you.
> I guess I slightly mistrust defenses based on "it will be too
> expensive for Them to dot dot dot." Like, "It will be too expensive
> for Them to find everyone's IP address and then find everyone's
> pubkey." If this can be automated, it doesn't seem like it's that much
> of a challenge.
I agree, but that does not change the fact that as of now we do not have a
viable option. Note that I said "targeted individuals and computers" - it
will still be easy enough to find some Freenet node to attack, but with
enough nodes much harder to get into a particular one.
>
> ~Mr. Bad
>
> --
> ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
> /\____/\ Mr. Bad <[EMAIL PROTECTED]>
> \ / Pigdog Journal | http://pigdog.org/ | *Stay*Real*Bad*
> | (X \x)
> ( ((**) "If it's not bad, don't do it.
> \ <vvv> If it's not crazy, don't say it." - Ben Franklin
> ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
>
> _______________________________________________
> Freenet-dev mailing list
> [EMAIL PROTECTED]
> http://lists.sourceforge.net/mailman/listinfo/freenet-dev
--
'DeCSS would be fine. Where is it?'
'Here,' Montag touched his head.
'Ah,' Granger smiled and nodded.
Oskar Sandberg
[EMAIL PROTECTED]
_______________________________________________
Freenet-dev mailing list
[EMAIL PROTECTED]
http://lists.sourceforge.net/mailman/listinfo/freenet-dev
