Bowie Bailey wrote: > Matt Kettler wrote: > >> Bowie Bailey wrote: >> >>> I was just looking through the default scores and I noticed that >>> ALL_TRUSTED is only scored at -1.8. I thought it had a much lower >>> score than that. Am I completely off track, or did something >>> change in one of the recent versions? >>> >> It got hacked back because trust-path misconfiguration is VERY >> common. >> > > So we reduce the effectiveness of a properly configured system in > order to prevent problems on a misconfigured system? Sounds a bit > backwards to me. >
You can always increase it. > Also, since the trust path issues that cause ALL_TRUSTED to FP also > cause other problems for various network tests, does it really make > sense to hide the main symptom of the problem? > No, it would not make sense to hide it, but here it is not being hidden. Besides, this score is NOT manually chosen. ALL_TRUSTED is still assigned by the Perceptron. I'm simply giving a reason why the perceptron assigned a lower score. Nobody manually went and hacked back the score. The test data demanded it. quotes from STATISTICS-SET3.TXT: 3.0.0: 1.453 0.0261 2.8051 0.009 0.74 -3.30 ALL_TRUSTED 3.1.0: 0.152 0.0081 0.4860 0.016 0.41 -1.80 ALL_TRUSTED Note that the S/O is nearly twice as high in the 3.1.x mass-checks (0.016) as it was in the 3.0.x mass-checks (0.009).
