[Spambayes] 'Bayes' patent...

[Tim Peters]

[Michael C. Neel]
> Looks like the patent office has done it again:

But they do this so often it's not worth worrying about.

...

> I'm by no means able to understand patent documents, but this one looks
> as if they have a patent on the combo of several anit-spam techniques and
> is not as broad as made out to be.  Still, given the usefulness of sb, a
> dev dropping the eff an email to put them on their radar may not be a bad
> idea.  One area of concern I saw was the use of 'paragraph hashing' which
> sounded somewhat like some of the advanced options in sb.

The patent specifically mentions using an MD5 hash for paragraph
fingerprinting, as one example of the infinite number of techniques they now
have the exclusive right to combine with an infinite number of other
techniques.  SB doesn't do paragraph fingerprinting, so can't be affected by
that clause.  And despite its name, SB doesn't even use Bayes Rule (although
it uses a different Bayesian technique).

> I'd hate to see sb unable to explore better methods because of this
> patent.

It is hard to concentrate when you're laughing too hard to breathe <wink>.