Let me make note of my random thougts before I forget.

1. Reputation needs to have an identifier of somebody being scored.
2. The same for who is scoring.
3. For what criteria, this reputation score was made.
4. For the reputation to be aggregatable, it has to have a distribution that we know about the aggregated distribution (such as normal distribution).
5. The information about the distribution, including what distribution, mean, and standard diviation must be published together with the score.
6. Display score must be intuitive for an average person.
7. Date that score was made
8. Signature by the score maker

So, the reputation score file should contain:

In the above table, I am proposing to use cumulative distribution P(X<x) as the display score, so that the meaning of the score is clear for anybody. If the score is 95.5, the subject is among the top 5% of most trusted in that criteria.

Also, public key of the subject being rated is included as par OpenID TX proposal.

Using this, parties who are trying to talk to the subject can be sure that the party really is the party that has been rated by the above rating agency.

This data can be serialized in XML format, or JSON, or tag=value format etc.

OK. This is another input to forthcoming ORMS TC at OASIS Open.