Having gone through the BYU CS program in the mid-90s, I actually thought there wasn't enough math and statistics. Zipfian, Gaussian, or otherwise. I suppose one of the problems is that it is hard to know what a future graduate might need. For someone who would be getting a job in device driver software coding, I suppose EE 220 was a good course to include in the CS program. But for someone like myself, who went more into information theory and research, I didn't really need to know much about that layer. I would have rather spent my time learning about more advanced statistics or about how to do good, efficient approximation of NP-hard problems.

And whether learning Gaussian vs. Zipfian is the wrong type of distribution to learn.. I suppose it all depends, again, on what you are going into. If you are going to be a marketer (Amazon/Google/eBay) or social networking, then yes, learn the power law family. But the sort of work I have been doing is with content-based music and image retrieval -- e.g. imagine presenting the system with a "cha cha" and having it automatically scan through your collection and find, from the raw audio signal of the music itself, other songs that fit the cha cha profile. Or humming a tune into the computer, and having it automatically find the song you are looking for. Or imagine pulling up a picture from your digital collection, and drawing a box around the face of your Aunt Frannie.. and telling the system to find all the other pictures in your collection with this face.

For that sort of CS work, you'd better know the Gaussian, too. Power laws don't quite cut it.

I suppose it is all about having the right tool for the right job. And at the undergrad level, it has got to be really hard to anticipate all possible future tools that people will need, would you say?