I just wrote to Reas and Fry about this problem, on my side!
For the record, I saw the message "by
jack.kern on 27-Jul-2010 08:20 PM" indeed marked as spam, and I had a button (as moderator) to mark it as not spam, which I did. Was it hidden until I "free" it?
On my side, I had a message indicating the above message was marked as spam because it had the word "loss" in it. I had other messages marking messages because they had words like "width", "big" and "video" in them.
Although I can see how spam messages can have these words (loss of weight, width of manhood, sexy videos... - I hope I won't be marked as spam myself, but hey, I am moderator!
), the Zoho team chose the worst way to detect spam. Trigger on words? come on, this is so 90's, totally inefficient and with many false positives (width and video words are prone to come up regularly in this forum!).
At least, a software like SpamAssassin (on mails) uses several heuristics to detect such spam (all caps subjects, links to addresses different of the displayed one, etc.) with a distinct weight for each and only when the note is above a limit it reports the mail as spam. Zoho's method is too simplistic, relying too much on moderator's work.
See the
Obscenity Filters: Bad Idea, or Incredibly Intercoursing Bad Idea? article for more thoughts on this. It is about filtering dirty words, but it is the same issue.
The
An Initiate of the Bayesian Conspiracy article shows how to do spam fighting more efficiently (but it is better just to use a good existing filter!).