I think this is absolutely worth the trouble to do. The benefits of XHTML/CSS will be realized for a long time to come. Combined with OACS I18N it would be an extremely powerful package.
In my own experience doing limited XHTML/CSS with OACS, the easiest way to do it is to start at the modules. One by one as the modules become XHTML/CSS, the entire toolkit can be migrated. It's a long road to get there, but there's no reason we can't start now.
The Zeldman book is a good start to understanding the "why" of XHTML/CSS, and Eric Meyer on CSS (New Riders) is a good "how" book, which shows practical, page-by-page examples.
Now there's just a matter of "how" to do it in OACS. I spent a bit of time working on exactly this with Dirk Gomez on Calendar at the Berlin bug bash in February, and Bruce has suggested this issue for the Copenhagen bug bash in March. With enough discussion, I'm sure we can come up with a system that works OACS- (and dotLRN-)wide.
