Re: [domainrep] Updated documents prior to IETF LC

["Murray S. Kucherawy" <[e-mail]>]

On Sat, Jul 13, 2013 at 11:42 AM, John Levine <[e-mail]> wrote:

> In article <[e-mail]> you write:
> >> Think of it as the likelihood that a bad rating from a given user is a
> >> local incident versus a trending problem.
>
> How about this:
>
> well-behaved: An estimate by the reputation service provider of the
> long term behavior of the rated identity, expressed as a
> floating-point number between 0.0 and 1.0 inclusive.  If the entity's
> rating is significantly different from the well-behaved score, it is
> likely an aberration.  For example, in an e-mail rating system, a high
> well-behaved score and a low rating might indicate a responsible
> network leaking spam due to security problems, and a low well-behaved
> score and a high rating might indicate a spammer who is between
> hosting providers.
>
>
>
Close.  The second sentence isn't quite what's intended; if the
"well-behaved" score is high, then a client can expect a unusual rating
(high or low) to be short-lived.  I don't think there's a need to tie the
two values together as your text has done.

Does that make sense?

-MSK