at least on the plus side, it would be nice to have a range (similar to the 5 star system).
I understand but IMO that just makes it easier to *game* the system (i.e. getting 10 +5 *fake* ratings is certainly going to be easier than getting 50 +1 *fake* ratings - I have the same problem with the "skill points" system I have implemented in CIYAM Open and am now thinking of changing it for just this reason).
Regarding the issue of scoring — my untested instinct is that flexibility in the scoring will be useful to users. There are plenty of types of highly structured or very simple trades that I can imagine where -1/0/+1 scoring would be totally appropriate. But for other types of trades that have multiple elements of trust (do I trust this person's ability to deliver on time? to deliver to spec? to communicate clearly? to be courteous in interactions?), more flexible scoring could allow for more nuanced interpretation.
If the protocol can support both, I imagine it will allow more widespread usage than if it rigidly supports a single scoring methodology.
This doesn't prevent any specific user or community — say CIYAM Open's community — from only considering +1/0/-1 scores for the purposes of its own, community-specific trust criteria.
Another untested instinct I have is that interpreting the trust data in the blockchain should be considered a separate process from getting that data into the blockchain (I put something to this effect in the FAQ section on my
pitch). For users concerned about specific patterns of scoring that "game the system" (like your example of getting 10 +5 fake ratings versus 50 +1 fake ratings), those users can define what types of scoring data carry weight for their personal decision-making. Scoring histories that are acceptably trustworthy to some users might not be trustworthy to others. I think users should be expected to consider what algorithm/criteria they use to interpret the trust data; it keeps them responsible, rather than encouraging blind trust in a single number.