The fact that we are still having this discussion in 2004 is testimony to how tenaciously humans will hold onto the first idea that comes along that vaguely works.
Here's something I wrote for Joe Gillespie's WPDFD in July 2001. It speaks about why HTML tables are a poor metaphor and an even poorer technological solution for creating web page layouts. At the time it probably was a pretty brave and foolish stance to take because IE5 for Windows really was the browser you had to target. Nowadays famous people, albeit jokingly, talk about jettisoning it from their list of browsers to lose sleep over. Quite aside from that though there are lots and lots of layouts that work just fine in it anyway.
What interests me though is why this idea of using tables for layout has been such a successful meme. The obvious answer is that initially it just worked, and nothing else did. So it had the first play advantage. I think this accounts for about 70% of its longevity. The other 30% is the fact that it sat fairly well on top of a technique that graphic designers had been using for generations to lay out pages for print based media: the grid. The vast majority of professional web designers these days still come from a traditional graphics background. In one way this is a good thing because it means we get to see loads of beautiful pleasing color schemes and graphics that design dilletantes like myself can only stand before in awe. But it's a bad thing in that many of these designers are yet to learn that it's time to let go of the old dispensations.
Just in case you haven't read it already, I'll leave you with this piece of enlightened thinking, just as relevant today as it was in April 2000. And apart from that, just bring on the sherbet :-)