On 1/27/2011 5:07 AM, Ian Eiloart wrote:
But let's not forget that defining "reputation" means two things. First,
the word "reputation" has a fairly tightly constrained meaning in the
English language. Those who argue that reputation can mean anything
probably aren't thinking of a new candidate entry for the Oxford English
Dictionary.
No, the existing dictionary definition may be "constrained" in one
sense, but not necessarily constrained well enough for communication via
a computer protocol. Computers don't find human language easy after all.
Some write as if it can be adequately be represented as a single "good",
"bad" or "neutral" indicator, or some scale between the three.
But can it? Even the dictionary indicates it can't.
Here's one dictionary definition:
reputation [ˌrɛpjʊˈteɪʃən]
n
1. the estimation in which a person or thing is generally held; opinion
2. a high opinion generally held about a person or thing; esteem
3. notoriety or fame, esp for some specified characteristic
have a reputation to be known or notorious, esp for promiscuity,
excessive drinking, or the like
Note the "specified characteristic". The reputation may not necessarily
be in the good-bad dimension. Or the good-bad may depend on the context
in which the question is asked. Eg: could you rely on that same
good-bad in the context of children versus adults?
Do we _really_ believe that there's a single reputation dimension in our
context, or at least, that we can "get by" with just one?
I don't think so.
But, second, the method of determining the reputation score that you're
going to publish for any given identifier. That method can vary widely.
Some operators may just viscerally hate the products or services that are
associated with a given domain (vi4gra.com or microsoft.com), whereas
others may have a carefully considered metric based on (say) feedback from
tens of thousands of sites around the world. Those arguing that
"reputation" can mean anything probably are referring to the methodology
used to arrive at a reputation score.
That speaks more to reputation of reputation instantiations in some
sense - which is probably out of scope. But it also does imply that a
reputation service may well be measuring different but still valid
things, and it would be best that the reputation service be able to tell
the questioner what it actually means.
We all know what a bad reputation *is*,
Kinda like porn, "we know it when we see it"? ;-)
Seriously, but do we? Entity A and entity B may both have validly "bad"
reputations, but for entirely different reasons, even within the same
reputation service, in such a way that they need to be interpreted
differently. A reputation service needs to be able to impart at least
some notion of context/reason so that a receiver can make an intelligent
choice to respect/disregard/amplify/deprecate or whatever.
I don't think a simple yes/no/dunno or even a scale is adequate.
Spamhaus Zen, for example, is multi-dimensional. Even its grossly
simplistic expressive power is greater than I see the road we're
treading here.
_______________________________________________
Asrg mailing list
Asrg(_at_)irtf(_dot_)org
http://www.irtf.org/mailman/listinfo/asrg