Stardate 20040318.1856

(On Screen): As some have noted, I quite often link to AP or Reuters articles located on the ABC News site. The main reason is that ABC carries both feeds and has convenient summary pages for them, and they keep older articles online for a long time. AND they don't require registration or login to view either current or older articles.

But like nearly any commercial site now, they also do a lot of obnoxious stuff with advertising. I don't see any of that, though, because I've got it all blocked. I primarily use custom rules I've set up with The Proxomitron version N45J, a filtration proxy for Windows which can be downloaded for free and used with any browser.

So I thought I'd create a special article where I would put my most recent custom rules for The Proxomitron, so that anyone looking for some relief could use them. You can find that page here. It's also linked from the top sidebar, and when I update it I'll alter the date so you can tell.

A relatively small number of custom filtration rules and killfile entries eliminates all the crap at the ABC News site. I've also got some other entries which take care of obnoxious crap at other sites. Use them in good health.

And while I'm mentioning cool and useful tools, I'll go ahead and update my experience with email filtration. A couple of days after first downloading and setting up POPFile, I ended up switching to K9. There were two main reasons for that. First, POPFile runs interpretively whereas K9 is compiled, and I was told that when the database gets large that POPFile starts to bog down. Second, the user interface for K9 is much more friendly. In terms of reliability and filtration, they are both pretty much comparable.

I've been using K9 since the beginning of November. It is very reliable and has never crashed on me. After the initial learning period of about three weeks, I reset the statistics gathering. Since November 24, it has processed 10,064 email messages. Of those, 3619 were good and 6445 were spams. (RoadRunner implements its own spam filtration; these spams were the ones that got by that initial filtration.)

There were 201 emails which K9 initially classified as good which I reclassified as spam, an average of less than two per day. And there has not been a single good email which it incorrectly classified as spam.

To put it a different way, K9 correctly identified 98% of the spam I received, and 100% of the legitimate email.

Nigerian 419 scam letters are the ones it has had the most trouble with. You know the kind -- "My father was the assassinated dictator of Lumburia and we need your help to smuggle $63 million out of the country..." It's not actually all that surprising that those are so hard to identify.

K9 is free, and can be downloaded from here. Both K9 and The Proxomitron get the USS Clueless seal of approval.

Update 20040319: Ian Tyger comments.

By the way, I've been informed that the latest version of POPFile for Windows now uses compiled code.