I’ve now been running SpamBayes for a little over a week. I’ve trained it with about 6000 known good and about 600 known spam emails. According to SpamBayes it should be equal for best results, but I didn’t have 6000 known-spams lying around for training, but I ran it this morning against a single one of my email accounts (IDSS) for just the emails received between the time I went to be and the time I woke up. This is what I got:
Out of 46 total emails:
- 1 (2%) was caught by Symantec’s AV
- 26 (66%) were spam caught by SpamBayes
- 3 (7%) were spam missed by SpamBayes
- 16 (25%) were valid emails (high only due posts from a yahoo group)
So far it’s the best spam filter I’ve tried, and it’s still learning.
I remember playing Adventure on my Sears Atari-2600 clone. Today’s kids would probably get bored in the first 30-seconds, and I can only imagine what it’ll be like when my kids, both under 2 now, start playing.
Here’s a pretty entertaining “interview” with kids playing some of the classics:
Since I’m fairly prolific on the web with newsgroup posts, articles, forum posts and the like, I get an inordinate amount of Spam – on the order of 400-500 spams a day. I’ve been fighting it for a few years now, and until now it’s always seemed like the Spammers were winning.
I installed SpamBayes just two days ago and it is now correctly identifying about 80% of the crap. I’ve got it set to not do anything with suspected items, only definitive Spam, so it probably could have a much better rate, but I figure why have it mark possible Spam that I have to read in a separate folder? I mena if I have to read it anyway to verify if it’s Spam, just leave it in my Inbox.
Most importantly, it has not incorrectly marked or deleted any valid email, even when it had zero training. I figure with a month of training this thing may be over 90%, saving me potentially a few hours a week. And best of all, it’s free.
I’ve really got no free time to speak of, but when a couple DVDs arrived in my mailbox yesterday, I just had to run the install.
For the record, when using Virtual PC, 128MB or RAM is not a large enough allocation for Longhorn…. 🙂 It could only do 4-bit color at 640×480 and was slooooooooooowwwwwwwwwwww. With a 600MB allocation on a 3.8GHz machine it’s not bad (considering it’s a very early release and hosted on Server 03) .
Now to install Whidbey. Unreleased products running on unreleased products. That’s a recipe for a crash if I’ve ever seen one.
Ever want to have an application idle timer for you CF app? Maybe to pop up a login screen, power down a peripheral or log off a network? A recent newsgroup post prompted me to write one, and in VB of all things. The thing that took longest was figuring out that VB.NET doesn’t actually have an application entry point and call to Application.Run unless you modify the project settings. And this is supposed to be easier than C#?
At any rate, the code is in the OpenNETCF Wiki. Expect a C# version in the near term. In fact I’m going to work on a set of IMessageFilter implementation examples, so if you have any ideas or things you’d like to see, let me know.
Just in case you’ve missed the news, on Monday we released version 1.1 of the Smart Device Framework.
Well, we’ve gotten most of the details worked out and have hung out the shingle. OpenNETCF Consulting is open for business. If you’ve got a project or a need for a consultant/contractor with solid Compact Framework experience, contact us.
Not that I really think anyone questions our capabilities, but to dispell any doubt, just look at our company resume.