[Top][All Lists]
[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
Fwd: Re: [GNUnet-developers] Re: amortizable hashcash paper
From: |
Christian Grothoff |
Subject: |
Fwd: Re: [GNUnet-developers] Re: amortizable hashcash paper |
Date: |
Tue, 13 May 2003 11:45:41 -0500 |
User-agent: |
KMail/1.4.3 |
As usual, mail.gnu.org eats anonymous mail, so here's January's message.
C
---------- Forwarded Message ----------
Subject: Re: [GNUnet-developers] Re: amortizable hashcash paper
Date: Tue, 13 May 2003 08:59:57 -0700
From: "January" <address@hidden>
To: address@hidden
Cc: address@hidden
I've becoming convinced that the hash-cash global popularity model
isn't the best, as it's too easy to subvert by someone who just
cranks up a machine to vote like made. I think a trust based
system more like GNUnet's general economic trust model would be better.
Here is, in broad strokes, a trust based moderation system for
rating content. Clients (humans) rate content on an arbitrary scale,
sending messages with a unique pseudo id for the client, the id of the
content they are rating, and an arbitrary numerical rating.
Nodes maintain lists of client pseudo ids, ranked by how similarly to
the
Node's local clients they rate content. The votes of top pseudo id's
in
this ranking are used to calculate a rating for content that has not
been rated by the node's local client.
There's holes in this big enough to herd cows through, but I am
hoping that folks can figure out elegant ways to close 'em.
Broad Details:
Client-to-Node messages:
Client rates content.
Rating is stored as the priority of the content.
Rating is sent to Log(known Nodes).
Client Search with a rating.
Regular search executed.
If the content has a priority of 0 (unknown rating),
Top 50% of rated pseodo ids votes are used to calculate
a rating for content, which is stored as priority of the
content.
Contents of priority below desired rating are not
returned.
Client Gestalt Search with rating.
Node-to-node search executed.
Results rated and filtered as per
Client Search with a rating.
Node-to-Node messages:
Rating.
Client pseudo-id, content id, and rating.
Message is processed, ignored, or relayed as per
normal economics and load.
If the Rating is from a known Pseudo-id, store it.
If the Rating is from an unknown Pseudo-id, replace
one of the lowest ranked Pseodo-ids with this one.
Gestalt Search with rating.
Search with a limit on rating of content.
Message is processed, ignored, or relayed as per
normal economics and load.
Only return results for local content rated greater than
or equal to requested rating.
Stored Local Ranking of Pseudo-id's
A fixed number of pseudo-id's are tracked.
A fixed maximum number of votes are retained for each tracked pseudo-
id.
Tracked pseudo-id's are ranked by reliablity.
The top 50% of pseudo-ids' votes are used to calculate
content rating.
The rating of pseudo-ids' reliability is modified as follows:
If the difference D between a client vote for some content
is greater than a margin M, the reliability is decreased
by D - M.
If the difference D between a client vote for some content
is less than the margin M, the reliability is increased by
M - D.
If a client search with a rating or gestault client search
with a rating is executed and the pseudo-id has no recorded
vote about that content, the reliability is decreased by
a small amount S where S < M.
When the priority of some content is calculated, the votes used
to calculate it are discarded.
Ps: Christian, if you'd be so kind as to forward etc. etc. etc.
Concerned about your privacy? Follow this link to get
FREE encrypted email: https://www.hushmail.com/?l=2
Free, ultra-private instant messaging with Hush Messenger
https://www.hushmail.com/services.php?subloc=messenger&l=434
Big $$$ to be made with the HushMail Affiliate Program:
https://www.hushmail.com/about.php?subloc=affiliate&l=427
-------------------------------------------------------
- Re: [GNUnet-developers] Re: amortizable hashcash paper, (continued)
Fwd: Re: [GNUnet-developers] Re: amortizable hashcash paper, Christian Grothoff, 2003/05/07
Fwd: Re: [GNUnet-developers] Re: amortizable hashcash paper, Christian Grothoff, 2003/05/09
Fwd: Re: [GNUnet-developers] Re: amortizable hashcash paper,
Christian Grothoff <=