Bad Temptations, Average Music, Great Software

[elfs]

This afternoon on the drive home I was so hungry and, despite my usual admonitions and reasonable self-control about food, I grabbed a bag of Doritos out of the gas station while filling the tank and downed the whole thing on the trip home. Urrgh. 700 empty calories on top of a bad day full of coffee.

The more reliable nutrition news outlets recently were full of articles about putting distance between yourself and temptation-- the human animal is driven to eat, and when food is immediately at hand fighting the unconscious will to grab it and go is hard. I've been so good about this recently, and it's one of the reasons I'm not suffering the spread some of my peers seem to be-- yet-- but if the blind speed with which I bought and consumed junk food yesterday starts to happen with some regularity, I'm going to be concerned. It's convention season, and the temptation to quick snacks is all around.

---

Godsmack should not be listened to on headphones. All of the power is missing. The new album, Faceless, is interesting, but unfortunately it's all the same. If you've heard the single Straight Out Of Line, you've more or less heard the whole album.

---

I've installed SpamAssassin on my account and it seems to be working reliably. It's filtered out about 90% of the spam I used to collect-- for a while there, it was really bad. But I've been more or less spam-free for a day now, which is wonderful. The only thing it hasn't filtered out yet is a "Microsoft Security Update" virus and a solicitation for a porn website that was so short it was hard to identify anything in it. I've turned on global tagging (even messages tagged as "non-spam" get their scores tallied aggressively) so I can analyze them by hand and adjust my local rule set to identify these two nuisances.

So far, there have been no false-positives, which is equally nice and much more important to me. I jacked up the score for a forged Microsoft Outlook header to almost guarantee that such a mail document looks spammish, since these usually contain viruses, and that might be a problem for some. I'll wait and see.

Comments

[technoshaman.livejournal.com]

I've been using Bogofilter (.sourceforge.net), a Bayesian filter written in C, to good effect. SpamAssassin takes less time to nab obvious spam up front, but a really well-trained filter (and you can get spam corpuses from various places on the net) will work better in the long run, and, besides, SpamAssassin beyond 2.5 has Bayesian stuff in it..... the downside is S/A is Perl, and between that and doing both the rulesets AND the Bayes stuff, it's piggy and slow. Bogofilter or SpamBayes (a Python implementation) cuts it down to JUST the db calls, and is just as good at weedout in the long run. talek and kendaer argue that Spambayes has a better algorithm; I on the other hand prefer the pure C implementation for speed. Either way we don't deal with a whole lotta spam around here. :)

(And there are utilities that let you tweak a token's score.... :)