About this document

This document is a set of quick notes representing thoughts while reading myidproject.net -- =Bill.Barnhill

"What is Reputation Service"

'parties' seems to imply live user agents, but they could be devices or semi-autonomous agents acting on behalf of an entity (which could itself be another device or agent)

I'd say trust and reputation are different

a reputation can be viewed a collection of n reputation metrics representing an aggregate over 2 or more opinions

trust is a threshold of values above which the party with the reputation is considered trusted.

Is it just 6 directed reputation edges? I'd say we also should consider the case where reputation is propagated. An example is a web of trust.

"Rating methods"

I would add a third, hybrid, that involves both modes, for example a restaurant reputation service based on a combo of Zagat's and diner feedback

"Contents of Reputation"

As previously stated I don't think every reputation can be expressed as a single numeric score

"Score Calculation Method of the Reputation Score"

How is this different than the two rating methods discussed previously in the paper?

"Other Requirements"

"Reputation needs to have an identifier of somebody being scored."

"The same for who is scoring."

"For what criteria, this reputation score was made."

"For the reputation to be aggregatable, it has to have a statistical distribution that we know about the aggregated distribution (such as normal distribution)."

"The information about the distribution, including what distribution, mean, and standard diviation must be published together with the score."

"Display score must be intuitive for an average person."

"Date that score was made"

"Signature by the score maker"

The reputation response...

Criteria

I'd see this more as an ordered set of identifiers, each identifying one criteria

Display score, Raw score, distribution, mean, st deviation

These all are based on a single numeric value reputation, which in my opinion is a small subset of the use cases an open reputation model needs to support

Subject public key

This seems to assume a specific algorithm for digital signatures. You could also have the public key resolved via the subject's i-name, and other algorithms are possible. It also does not specify the algorithm used to generate and verify the signature (see for example www.w3.org—xmldsig-core)

Published Date

Published date is good, but I'd also add a last changed date

"Reputation Request"

It's important to note that XRI resolution would be resolved to get the service URI, including following service refs