On Wed, 2011-01-26 at 13:38 +0100, Alessandro Vesely wrote:
No, it shouldn't. Although delays in fixing problems may affect it,
reputation measures the willingness and the ability to deal with them,
rather than the mere frequency at which problems occur.
I (and no doubt several others) beg to differ.
You may define "reputation" in that way, but that definition cannot be
automated.
At $workplace I use Exim to measure "reputation" - or "badness", or
possibly even "goodness" - as defined by a function of the rate of
invalid recipients, messages classified as spam, messages classified as
malware, unverifiable sender domains (per DNS lookups, not callbacks),
relay attempts, and a number of other indicators.
If that function returns a value under a certain threshold as compared
to the volume of "good" and "bad" mail arriving, that message is
deferred. Currently we don't reject - but we could change the threshold,
and do that instead.
There's no way for me to factor in human values such as "willingness" or
"ability" into that function; nor should I need to. If the data stream
from a given source is rubbish, I don't want it - and yes, that can hurt
legitimate messages (and has done so) but that's a chance we've been
willing to take.
So in my system, "reputation" is automated according to rates of hits on
certain criteria. This is just as valid a system as a manual one, and
should not be dismissed because it doesn't use human value judgements.
Graeme
_______________________________________________
Asrg mailing list
Asrg(_at_)irtf(_dot_)org
http://www.irtf.org/mailman/listinfo/asrg