On Monday 05 January 2004 19:13 CET Dan Kohn wrote:
My view is that nothing is more obviously confusing and in-your-face
wrong than the text "X-Spam-Status: No, hits=5.0 required=5.0"

I agree. The test as whether we are doing the right thing should be that if X-Spam-Status says "No" then "hits" should display a number smaller than "required". If X-Spam-Status says "Yes" then "hits" should display a number equal to or greater than "required".


Any rounding that achieves that is correct, and any that doesn't is broken.

The flaw in the reasoning that led to the current implementation is in the following line from the wiki:

"Scores in decimal are meant to be more of an informational thing rather than a filtering criteria, so there is no need to round down as with stars."

What's wrong is the unstated assumption that rounding to the nearest half is "correct" independent of the use of the header as "an informational thing".

That's wrong because rounding up results in a header that provides confusing information and rounding down results in a header that provides not confusing information. If the point is to provide information, I am better off being told that a message that actually scored 4.96 is a "4.9" than I am begin told that it is a "5.0".

 -- sidney



Reply via email to