From: Hal <hfinney@shell.portal.com>
Date: Sun, 3 Sep 95 21:52:26 PDT
To: cypherpunks@toad.com
Subject: Re:  pseudonyms & list health
Message-ID: <199509040451.VAA23173@jobe.shell.portal.com>
MIME-Version: 1.0
Content-Type: text/plain


From: "Vladimir Z. Nuri" <vznuri@netcom.com>
> hence, what I am suggesting is that one
> of the "stable attractors"  that TM is always talking about
> may be a reality in which people pool most of their accomplishments
> under a single nym. furthermore, they will wish to avoid conflict
> with other nyms in respect to attribution. 
> isn't this, functionally, the equivalent of the supposed
> anti-cypherpunk "true names"?

An interesting point.  I have long wished that there would be a form of
"credential certificates" which people could give as special signatures
on other people's public keys.  Then using Chaumian credential technology
it would be possible to anonymously transfer these credentials from one
pseudonym to another.

This is not a perfect solution, of course.  Much reputation is informal
and simply resides in the opinions held in people's minds.  But perhaps
if a more structured solution like this became widespread it would help
to prevent the "concentration of reputation" which Vlad describes.

Along with the usual flames, I occasionally get messages saying nice
things about postings I have made, and I sometimes save these in a file
called "praise".  Here are some excerpts:

> Again, thanks for posting some useful information that
> actually has *direct relevance to crypto*.

> I really enjoyed reading this.  It was well written and comprehensive. 
> Thanks for sharing it. 

> Nice post!

I certainly appreciate these kinds of comments, but it would be even
more useful if such messages were expressed as the kinds of
certificates I am describing.  I wonder whether people would be willing
to use a program which would let them issue such "reputation
signatures" of various kinds, and display the signatures which were
present on keys.

Discussion of such schemes has often bogged down in considering the
various categories or types of credentials people might want to give.
This is somewhat analogous to the "rate-the-net" schemes we have talked
about where a similar issue arises if we try to mark pages with a whole
range of characteristics so people can judge whether they should let
their kids read them.  Perhaps the solution needs to be found in
simplicity.  SurfWatch (as I understand it) gives a simple "thumbs down"
to selected web pages.  Maybe a simple "endorsement" would be useful as a
reputation credential without trying to identify exactly what it is about
the person you are endorsing.

I could see such a system initially being piggybacked on PGP keys (the
signatures would not be understandable by PGP though), although for
Chaumian credential transfers the keys have to be specially structured
and that would require a new approach.

Who would be willing and/or interested enough to use such a system if it
existed?

Hal