From: Hal <hfinney@shell.portal.com>
Date: Sat, 10 Sep 94 13:37:53 PDT
To: cypherpunks@toad.com
Subject: Re: reputation credit 3/3
In-Reply-To: <199409101958.PAA01656@bwh.harvard.edu>
Message-ID: <199409102037.NAA19749@jobe.shell.portal.com>
MIME-Version: 1.0
Content-Type: text/plain


Adam Shostack writes a very interesting set of articles on a concrete
proposal for reputation credentials.

A couple of suggestions: maybe you should distinguish between respecting
someone as a writer and respecting them as a reviewer.  In the real world,
we have editors, publishers, and others whose main job is to discover and
facilitate the good writers.  Just because you write well doesn't mean
you will be good at recommending other writers, and vice versa.  Adam
brings this up himself when he talks about a good writer who intentionally
makes bad recommendations.  Creating these two different kinds of credentials
would help solve this.

A related point is that doing this helps remove some of the normative or
reward/punishment aspects of this system.  Saying that you like someone's
recommendations is similar to saying that you have similar tastes to
theirs.  There is not so much stigma or insult associated with refusal to
give a credential saying that you like someone as a reviewer.  It just
means your tastes differ.

OTOH refusing to endorse someone as a writer is a stickier business.  It
may offend others and it could bring retribution upon yourself.  It could
be a way to create enemies.  Especially if you went with numerical
rankings so you said "I like John Doe's writing 5% of the time", this
could be insulting.  If you don't have these "negative" credentials it is
not so bad but it still may be noticable if someone endorses a lot of
people with a few notable exceptions.

The problem, then, is that people may be reluctant to be honest with
their opinions.  They may find it safer to follow the crowd and add their
own endorsements to those already popular than to take a chance with
honest praise of some pariah.  There was some discussion about this in
the development of PGP.  Should there be a way for people to say how much
they trust another person as a signer?  If you had this (in a public way)
then you could have transitive trust to some extent and it would expand
the web of trust considerably.  But again the concern was that people
would not want to expose what they truly thought of the signing policies
of their friends.  I suppose you could get around this by having one set
of opinions for public consumption and another set used for personal
message rankings, but that seems a bit extreme.

Still, I think it would be a worthwhile thing to try.  It would be nice
if we could do some more interesting cryptographic stuff than just simple
signatures, though.

Hal