The first thing I’d say is that this project is one of compromise; you won’t get a seamless workflow, if only because not all of those programs support Applescript. Sente and Bookends are a problem there, and I use Bibdesk as a result.
Also, as I work in the humanities, the ability to search for papers on the Internet through my bibliographic tool is pretty limited. The bulk of my research is scanned from books and journals via xeroxes. The other parts are online databases (JSTOR, etc.) and hand-typed notes.
The other problem is metadata. You could store it in a DTPO sheet, and perform certain operations on it via Applescript, but quick sorting is an issue. OmniOutliner is pretty wonderful, but it’s lack of aliases is a real minus for me. After Tinderbox finally moved to Unicode, I was right back with it. In spite of its flaws, there really isn’t anything like it. I find I don’t have much need to search my notes in DTPO once they’re in Tinderbox (and structured via metadata), but I do go back to DTPO (and the “raw” data) often to see if I’ve missed something.
My components now are: DTPO, Skim, Bibdesk, Tinderbox, TextMate, MultiMarkdown, and a LaTeX system that includes: XeLaTeX, Memoir class, Biber and chicago-biblatex-df style, a complete UTF-8 typesetting path.
I also think that Mark Bernstein’s idea of “gardening” is the right approach. Organize and invent stuff when you need it. I can’t tell you how much scripting I’ve thrown out as unnecessary once I got down to really using it. Right now I’m scripting in Python, only because I was doing other work in that language, and didn’t want to confuse Ruby with it. Moving away from Applescript is a profound improvement, and I’m surprised that people won’t switch over in spite of the torture of that language.
Anyway, I put a rat’s bag of stuff here:
(EDIT. Maybe you haven’t seen these:)
which you may find of interest.
My post’s a bit disjointed, but I’d be happy to continue the conversation.