After some severe data loss I'm getting a little more concerned about the issue.
I do backups, but not regularly, and without much of a system. I know several ways to do it, but I'm trying to get input from some of you people in person.
I'm not so sure how to exclude some directories from being... for example packed into a tar file by a cron job (fist thing that comes to mind)...
Also I don't want them to be overwritten by the next backup right away. I need a weekly backup of about 30 gig - obviously only changes should be included), And I'm willing to sacrifice a 250 gig drive which I currently use for my static backups... ~100 gig free.

I use dar with a custom Bash script to do all my backups. Dar is capable of doing what I think you need. It will do incrementals and differentials based on previous backups as well as the ability to slice the backup into smaller files. You can also exclude directories and best of all it will also include file deletions in your differential backups based on a previous full backup. For instance, say you do a full backup. Then you end up deleting certain files. When you do a differential backup based on that full backup, dar will record that you deleted those files in the differential backup. In that way, you can restore your system to its exact state by first restoring your full backup, then applying the differential. I've restored my system countless times using dar and so far, never a problem. The bash script is set to run in cron in the wee hours of the morning when the system is up. Also, I have it change to runlevel 2 to reduce any system chatter while the backup is running. Make sure you trap any system signals in your Bash script just in case the system wants to shutdown in the middle of a backup. That way you can clean up before it goes down and not have a corrupted backup file. I could send you my bash script if you would like to look at it. Hope this helps.

Windows: 'ntbackup'
GNU/Linux: tar and copy to cd-r's or dvd-r's

For Linux, a bash script that is run nightly by a cron job. The script tars the necessary directories then copies them to a network share on another machine, which is backed up by tape.

I run an hourly cron job that calls a script. The script uses rsync over ssh to do an incremental backup to a partition (on a remote machine) dedicated to this purpose. The script can be written to rename the last backup and create a new one, if you want. But this will consume more bandwidth. If you wanted to do it with "rolling" backups, you'd only want to run it perhaps once daily, at night when the bandwidth won't be missed.

I've been using rsync to keep three complete copies of all my data, I have never lost anything. In fact I haven't lost any data since I went 100% Linux some 8 years ago.

One of the copies (all automated mind you) goes onto a USB removable hard drive, so if (God forbid) there be a fire or other emergency, I can grab the removable and run.

I have enjoy serious data security peace of mind, all thanks to rsync and ssh. It's elegantly simple to do.

I, too, use dar run from a bash script. My script is setup to read a parameter to determine just what I want to backup at the moment - either the whole system or just my home directory.

I had problems using dd and tar in the past because if they ran into a problem reading a disk sector they would stop. Dar will skip the file, print an error on the screen and keep going.