There are no direct flights from Amsterdam to Denver, so I was forced onto the Layover Roulette playing field. It hardly ever snows here in Amsterdam, but on the morning of my flight to Detroit of course it did. With the snow plows and the deicing equipment overstretched, my flight was delayed to the extent that I missed my connecting flight to Denver. I was rerouted via Minneapolis in the end, and my luggage even managed to keep up with me. I think was lucky to make it. In any case, on the final flight from Minneapolis to Denver I sat next to a woman working on her laptop. I know it's rude to peek, but with the screen always within my peripheral vision I could not help but notice what she was doing.

She spent quite a bit of time working in Excel. Now, in my job, I spend a ridiculous amount of time working in Excel, and I have broken the program many times. What I saw on the flight was probably how the vast majority of people use Excel, and I was astounded. It was completely contrary to anything I have ever done with a spreadsheet. She was hardly using any formulas other than SUM() to add numbers together. Even more astounding was that she was creating new views of the same data, but kept on re-entering the numbers from cells on other sheets. Sometimes she copied the numbers over, but often she'd just type them in again. Then she actually whipped out a huge calculator, did some multiplications and divisions on the numbers shown on the screen, and entered the results in yet another cell. Absolutely amazing.

My spreadsheets tend to get out of hand. I parametrize everything, because I know that I will want to play around with various scenarios later (and I don't mean the built-in Scenario option, it's too limited). I am lazy. Very lazy. I don't want to solve problems more than once. So I prefer to solve a class of problems rather than an instance of a class. This means more work initially in building a spreadsheet to deal with all aspects of the class of problems I am solving, but it pays off enormously later on when a different instance of the same class of problems comes along. It also means I break Excel a lot. It has a tendency to disappear. Not crash, but disappear. It usually happens in running VB code, and I have not been able to isolate the conditions yet... you're happily running your code and suddenly Excel is gone. It does not even go through its Dr. Watson-post-crash thing. Excel also tends to become unstable if you have large amounts of data in the sheets. Filling a few sheets with say 20 columns of data that's 50,000 rows deep will usually do it. It'll crash sooner the more formulas you have in the cells.

In any case, this brief glimpse I got into the world of Excel usage supports the thesis that computers are already fast enough for most people. Reducing the complexity of computing is likely to bring far greater benefits to these ordinary users than faster chips. For my work though, I could always use a faster computer, although here I have run into a curious problem as well. I have a dual P4 running at 833 MHz at work, while one of my colleagues has a newer machine with a dual 1.7 GHz P4, DDR RAM versus my SDRAM, etc. In running some of our heaving spreadsheets there's hardly any speed difference between the two machines. I am absolutely stumped as to why this should be. Raw CPU speed is almost double and memory bandwidth is much greater. The spreadsheets are not I/O-bound. Yet the speed improvement of running them on the newer machine is marginal. There's something very strange going on. But that's nothing really new...