----- Forwarded message from Oskar Sandberg <oskar at freenetproject.org> -----
X-Spam-Checker-Version: SpamAssassin 3.1.4 (2006-07-26) on servalan
X-Spam-Level:
X-Spam-Status: No, score=0.0 required=5.0 tests=none autolearn=ham
version=3.1.4
Envelope-to: toad at amphibian.dyndns.org
Delivery-date: Mon, 25 Sep 2006 21:54:09 +0100
From: Oskar Sandberg <os...@freenetproject.org>
To: toad <toad at amphibian.dyndns.org>
Subject: Re: [toad at amphibian.dyndns.org: Re: [freenet-dev] Open net - initial
design ideas]
toad wrote:
> So we force the distribution? Hmmm. We don't have to force the
> distribution if we path fold on successful requests, but forcing it is
> better in that it will almost certainly work regardless of e.g. local
> traffic levels.
I don't get it? Why would you want to try to pick links explicitely
according to a given distribution, when we know that just choosing the
destination of incoming queries should work fine. Did I miss something?
> I'm not sure about the "probability proportional to 1/d" bit, now that I
> think about it; doesn't that mean that there will be infinite
> probability that a point infinitesimally different to the current
> location will be chosen?
That is correct. For such a thing to make sense, you have to select
between an minimum and maximum value. You would probably want:
double r = rand.nextDouble();
double d = Math.pow(n / 2.0,r-1.0);
which will select something between 1/n and 1/2 with the correct
distribution. Note that you have to know n (the number of nodes) to do
this - there is no way to get away from that. Also, if two nodes are
closer than 1/n, that is bad, so really one needs a centralized
authority giving out evenly spaced IDs. All this is the reason why
destination sampling is much, much, better.
// oskar
----- End forwarded message -----
-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 189 bytes
Desc: Digital signature
URL:
<https://emu.freenetproject.org/pipermail/devl/attachments/20060926/e9f99bd2/attachment.pgp>
