[Spambayes] Wild one

Tue Aug 12 23:00:29 EDT 2003

[G. Armour Van Horn]
> This has *got* to be targetted at us, given the number of random
> strings they are including in hidden text.

In general, random strings have no effect on spambayes scores:  e.g., it's
generally the first time we've seen the token "callpapakhi", and so that
word gets a spamprob of 0.5, and is therefore entirely ignored.

> Also note that there is no <body> tag, possibly to confuse HTML parsing?

Possibly, but that has no effect on us either.  We don't do "real" HTML
parsing, and ill-formed HTML doesn't bother us.

> Van
>
> Subject: literami UMW Make yourself Rich
> ...
> X-Spambayes-Classification: ham
> X-Spambayes-Spam-Probability: 0.0565997915183

It would be interesting to see a breakdown of why the score came out that
way for you.  The quoted message scored as solid spam for me; it simply
didn't contain any tricks that cause any trouble for spambayes.  Unless
maybe you're using a (by now) very old version of the tokenizer?