I think its a bit early to start speaking 10.5. I suspect there are still many things we don't know about that will make it into 10.4. Specifically I suspect a lot of UI elements are rather fluid. I also think given Apple's comments that there will be a much longer wait between 10.4 and 10.5.
My bet is that by the time it comes out CPU power may be significantly higher. We'll probably have quad systems standard. (Dual die CPUs standard) So its hard to say what'll happen by then - say 2006 or 2007.
My guess is that broadband will have reached the point that on demand delivery of movies will be standard and both Apple and MS will target that market more. Right now Steve talks crap about that market, but MS has been targeting it even if its products are thus far underwelming. So see more entertainment systems features standard in 10.5.
I also suspect that OSX will make the jump to DPI independent interfaces. It's only a matter of time. Right now high res screens are hard to read with OSX. Windows has limited features in this regard, although because icons/graphics aren't scaled it is of limited utility. (Although still much better than OSX) Moving to a vector model isn't as trivial as many think - either in terms of compatibility or computing power. But perhaps it'll be more possible by 2007.
To me the real big things will depend upon how Apple's hardware business goes. Right now they've been losing market share (still) and haven't updated the low end in any significant way. Will they drop the low end and try to re-invent the workstation market? Hard given Linux/Windows. It'll be interesting to see. I suspect what they go after in 10.5 will have as much to do with hardware as anything. Will IBM keep the PowerPC competitive? Will dropping marketshare force them to open things up more? Hard to say.