Re: Solution to churn problem and possibly to the Pitch Black attack

--===============1122352834==
Content-Type: multipart/signed;
  boundary="nextPart22537910.xYimplrUpO";
  protocol="application/pgp-signature";
  micalg=pgp-sha256
Content-Transfer-Encoding: 7bit

--nextPart22537910.xYimplrUpO
Content-Type: Text/Plain;
  charset="iso-8859-1"
Content-Transfer-Encoding: quoted-printable

On Sunday 15 May 2011 06:14:31 Pierre Abbat wrote:
> I read the Pitch Black paper and I think I've figured out a solution. Ins=
tead=20
> of trying to fit a circular keyspace in a graph that's spread over the=20
> surface of the earth with some long-distance links, I think it would be=20
> better to make the keyspace have more dimensions. Here's my proposal:
>=20
> Break a key or location into four equal parts. These are coordinates of t=
he=20
> point on a four-dimensional torus. Distance is measured in 8-space by=20
> embedding each torus dimension as a circle.
>=20
> Every 1024 times a node adds something to its store, it performs this=20
> calculation: First it adds all the keys in its store as vectors in 8-spac=
e,=20
> then reduces that to a 4-dimensional point on the torus. Then it adds all=
 its=20
> neighbors' locations and again reduces it to a point on the torus. The re=
sult=20
> is its new location. The location will be close to its neighbors; if the=
=20
> surface containing locations is curved, more of the store will be on the=
=20
> outside of the curve. Young nodes with few keys stored will be pulled rap=
idly=20
> toward their neighbors; old nodes will remain stably located near their d=
ata.=20
> Unlike the current method, where the product of distances rewards nodes f=
or=20
> having locations next to each other, in this system nodes next to each ot=
her=20
> will be pulled apart, spreading more evenly over the keyspace.
>=20
> The number of dimensions could be any divisor of 32 (the number of bytes =
in a=20
> hash). More than that and the notion of averaging in a dimension becomes=
=20
> problematic. I have a hunch that the optimal number of dimensions is 4, w=
here=20
> two represent the surface of the earth and the other two different social=
=20
> classes in a region. Could someone who has a Freenet simulator run some t=
ests=20
> and see how well this method performs for various numbers of dimensions?

You should simulate this. There are a few simulators but you might have to =
write your own.

The main difficulty I see is that greedy routing only works if most of our =
connections are close and a few are distant. If they are all distant, a req=
uest will get to roughly where it should be quickly but will not be able to=
 get to a specific location in a reasonable number of hops.

--nextPart22537910.xYimplrUpO
Content-Type: application/pgp-signature; name=signature.asc 
Content-Description: This is a digitally signed message part.

-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.10 (GNU/Linux)

iEYEABEIAAYFAk3bqucACgkQYUNbc3WUHYjUvACgt5Fgw6njLcNiUj3Af0fNltGK
wrwAnRZRGKuA1wyWXilvNC4nSgdxTodP
=SXdx
-----END PGP SIGNATURE-----

--nextPart22537910.xYimplrUpO--

--===============1122352834==
Content-Type: text/plain; charset="us-ascii"
MIME-Version: 1.0
Content-Transfer-Encoding: 7bit
Content-Disposition: inline

_______________________________________________
Tech mailing list
[email protected]
http://freenetproject.org/cgi-bin/mailman/listinfo/tech
--===============1122352834==--