I have some serious issues with this article on CSS-based table layouts. Put simply, my issue is that they're still table layouts. The author goes into how to create a table layout without an element acting as a row or a containing table, which is interesting, but doesn't seem very useful for anything beyond three-column layouts. Circa 1998 I did not use table layouts to create nice, even columns - I used them to slice up complex designs (ok, not that complex, but they still required colspan).
I was coming around to her way of thinking until the point after the discussion of minimal markup when the author shows an example of four boxes in a grid. It falls apart for me here, because to get padding between the boxes, it seems that a faux table and two faux rows are required. I may as well use an actual table at the point, as I'll have the same amount of superfluous markup.
I can do the same thing with just four divs. Yes, I will probably use overflow:auto to keep the height consistent. Naturally, the divs will be floated. But unlike the author of the article, I guess I don't see any inherent problems with floating. Floating works well for me. I am a fan of it. You know what else? I think calling faux columns a trick or a hack is missing the point. Faux columns are a design solution to a design problem. There's no trick there, they create visual separation between different areas of a page with an image that's not part of the markup.
Maybe I just don't get it. I came around to CSS early, with relief and no reservations. I don't miss table layouts. They're a bitch to maintain and they're terrible for dynamic content or large websites. CSS has made my job incalculably easier. I can't imagine why anyone would want to approximate those bad old days, after we've come so far.