Let derivative and raw buckets happily diverge - there's no reason for them not to.
This also confounded me at first, and I have to say, the divergence still bothers me a little. Mostly because I'm still shooting JPEG for now, so I can't do even simple exposure or color adjustments without creating a derivative. So I may be leaning more heavily on derivatives than some do. Although I don't feel the need to do even adjustments very often, and not (just) because I'm not a "serious photographer". It's worth noting that I'm also a semi-professional landscape painter, and I do virtually all my work in the field and only rarely revisit my paintings in the studio later. And actually, my real profession is music - specifically, I am a jazz pianist. Most of the music I create is improvised - this has a similar what-you-see-is-what-you-get, take-it-or-leave-it vibe. So my attitude regarding my photos is in keeping with my attitudes about my other passions. On the other hand, I'm quite convinced I'll be shooting mostly RAW at some point soon. But for various reasons, I'd prefer to get a really good JPEG workflow figured out to use for now and keep re-evaluating when it makes sense to switch as time goes on.
Since I ended up going with ACDSee Pro (despite some very valid concerns raised by Peter), I have no built-in version tracking, but I'm experimenting with ways of tracking derivatives myself. One idea I'm playing with is setting up nested virtual sets just for this purpose. As I have it currently, I have a top-level set called Derivatives, and underneath that are sets for Original and Master. Right now, that's the only kind of derivative I'm really interested in keeping - any others I use are usually created and deleted automatically for me (when, for example, I choose to send some as screen-sized email attachments). As I'm doing my initial work with images assigning metadata and rating, if I see any I feel need post-processing, I toss them in my "Derivatives/Original" virtual set. If I'm feeling industrious, at that time, I'll also make my adjustments and save them into a new file. Just as one would in the "usual" DAM Book workflow, the new file goes into the derivative bucket, which is completely decoupled from the originals bucket. But I'll also assign it to the "Derivatives/Master" virtual set. At that time I can also copy the metadata and assign it to all virtual sets the original was in, and perhaps *de-assign* the original or even remove keywords from it so only one copy of the file shows up in future searches.
So far I have just one big set for the originals and another for the masters, but of course, I could subdivide these into virtual "buckets", which could correspond to the actual originals buckets, or be content-based. Assuming I actually do follow through and make masters for all files in my Derivatives/Original set, then there will be a one-to-one correspondence between originals and masters, which I figure may help me keep track of them down the line. I like the way this makes it easy to view just derivatives, just originals, or both side-by-side (by viewing both sets and sorting by name). That way if I later upgrade my processing software - or, more to the point, if I upgrade my own processing skills :-) - I can easily browse these images and see which might be worth revisitng.
But I think the main idea here is that my master for most practical purposes replaces the original. In fact, I would generally be treating my originals more like RAW files that have been converted to DNG but are still being archived just in case.
Anyhow, I'd appreciate any feedback on these ideas. And feel free to remind me of what a mistake it is to be shooting JEPG or using ACDSee Pro; at some point, I'll write up a defense of these choices for myself. But one thing that impressed me on reading Peter's book is that the ideas in it are *not* just valuable for the professional, but many of the specific techniques described are going to seem so daunting to many amateurs that they won't go there at all, or will make only token efforts that miss the point (eg, the guy who said he wasn't sure he needed a database...). I'm interested in working out a "DAM for the rest of us" approach that is as true as possible to the ideals put forth by Peter. And given that "the rest of us" is going to look a whole lot like the set of folks for whom shooting RAW is just out of the question, and hence will not be able to take advantage of ACR adjustments and DNG in doing the sort of basic post processing that even we like to do, we're going to need a way to deal with dealing with the derivatives we create.