Your Biggest Regrests Setting Up Databases?

I’ve been using DEVONthink Pro for about two weeks now. It’s certainly bringing out the digital pack rat in me as I dump tons of papers, reports, and websites into my database. A lot of advice on these forums encourages this behavior, suggesting that DEVONthink works best when it has a lot of information to work with. I’m starting to get worried though that this hording will backfire. It’s happened before in other contexts.

Maybe DEVONthink will be that magical software that rewards the digital pack rat, or maybe I’ll have to overhaul the database in a few weeks when I realize the mistakes I’ve made. Either way, the software has been amazing, and I’m sure it will all work out in the end.

I’m curious. What are your biggest regrets with how you first set up your DEVONthink databases and how did you end up fixing things?

No regrets. I frequently rearrange, split & merge databases, groups and documents. Let your structure emerge (with the AI’s assistance) depending on the data you’re collecting. If you get to the point of wanting to do a major reorg - be sure to backup the database first and/or clone it and reorg the copy.

I haven’t yet discovered the ideal system for internal organization. I’ve tried some things based on the Library of Congress system, the Dewey Decimal system, and others, but it’s all ultimately arbitrary. You can make it as vague as you want or as specific as you want, but ultimately, it’s going to be a compromise. There’s not much you can do about it.

Tagging (and eventually key-value metadata, hopefully) and smart groups are, IMHO, the best way to minimize the artificial and arbitrary folder hierarchy metaphor. If you want to work in a particular area, summon the relevant documents with a smart group, and go from there. Of course, this depends on a certain behavior, namely that you don’t just dump things into DTP.

There are different usage scenarios, and it’s important to realize that. Bill DeVille represents one particular philosophy about DEVONthink. From what I can tell, it’s by far the prevailing philosophy. And Bill DeVille is extremely intelligent, educated, knows what he’s doing, and does it well. I can’t criticize what he does or how he does it, and his advice is impeccable. All I can say is that his philosophy doesn’t work for me… but, fortunately, DTP doesn’t force one to work that way.

I don’t use DEVONthink as a general document storage system. That’s what the filesystem is for, as far as I’m concerned. I tried shoving all of my documents into DTP databases and, well, it was absolutely ridiculous. I have thousands of books and hundreds of thousands of other documents in PDF format, and if I need to find a particular one I use Spotlight. Even one database for each discipline (History, or Philosophy, or Psychology, or Religion) would be untenable, and the benefits of having DTP manage the information would disappear, even as well as DTP scales.

Essentially, where Bill and I differ (and he may have different thoughts on this – this is just my impression) is that he views DTP as a warehouse, and he relies upon DTP to do warehousing work. He stores documents and pieces of documents, takes inventory, and expects to be able to find the information quickly when he needs it. He also enjoys the connections DTP makes between different documents. The warehouse gets more full over time, but that just makes Bill’s setup function better. As far as I can tell, this is what DTP was intended to accomplish.

I view DTP as a refinery. I bring in raw material (books, articles, other documents), I process it, creating new documents from the distilled and purified essence of the raw material, and then I throw that raw material out. I know every word of my database, because I wrote every word of it.

So my databases are almost always far smaller than Bill DeVille’s. The content is almost always RTF files, not PDFs. I rely much more heavily on properly naming, tagging, categorizing, and describing documents. I rarely (if ever) use “See also” or certain other of the tools provided by DEVONthink’s artificial intelligence – because I have always prided myself on my ability to see patterns and draw connections between disparate phenomena, and at least with my databases I haven’t seen compelling genius from DTP’s implementations.

Those are two different ways of using DTP, and they’re quite different. The extreme ends of one spectrum. There are probably other spectra, and Bill and I definitely do share some strategies in common… but anyway, since you’re a new user, I wanted to express my opinion clearly: There’s no “right” way to use DTP, and it’s a very versatile and powerful program, and you should just concentrate on what works for you.

Right on, kalisphoenix, I’ll defend to the death your right to ‘think different’. I’m probably a lot lazier than you, hence very different approaches. :slight_smile:

Your advice is on target. Choose an approach that works for each user. Experimentation (otherwise known as ‘playing’) is probably the best way to work out a system that is satisfying. Within the community of DEVONthink users there are wide differences in the ways users approach issues such as organization and tagging.

Some users tightly organize and/or tag every new item as it’s entered into a database, and for some kinds of content and projects that’s probably a good idea.

My main database contains about 25,000 references and thousands of my own notes about that content. I’m very sloppy about organizing material as it’s added to the database and I never bother to tag items as they are entered. When I start a new project that requires me to analyze the information in that database, I’ll probably do a bit of work refining organization of interesting material, and I may assign tags to important kinds of references and notes for the purpose of the project. When I’m finished with that project, though, I’ll probably remove some or all of the tags, so that I’ll start with a fresh look at the information for the next research/writing project (perhaps a personal eccentricity, not intended to be a general recommendation).

My notes tend to provide ‘extended metadata’ about some of the content, along with such DEVONthink tools as searches, smart groups, See Also, etc. For what I do, that free-form, ‘start each project afresh’ works for me.

But I hasten to add that I don’t treat some of my other databases, especially the financial one, in such a free-form fashiion. That’s because initial tight organization and/or tagging pays off in my financial database, e.g., at tax reporting time.

I don’t disagree with the suggestions made by Sophie and kalisphoenix; in fact, some of those suggestions would save me the extra time and effort that I put into some of my kludges to accomplish my objectives. But Christian tells me that some of them, especially a good UI to go along with user-created metadata fields and associations of notes to their targets, would probably require an extensive redesign of the database model. Perhaps in DEVONthink 3?

Hey, kalisphoenix, it’s good to have you back making suggestions. :slight_smile:

Not likely, especially with work on version 3.0 coming after the release of DEVONagent 2.5. (Tinderbox is looking better all the time!)

One great possible UI for this already exists. If I open a document, and click “See Also & Classify”, a drawer slides out with two pre-defined associations to other documents/items: See-Also and Classify. I think almost the same panel could show a document with its annotations / notes, and allow to open the notes, remove them, and add them.

I think general user-defined meta-data would be very powerful (for a relatively small number of users), but is not necessary for solving the “notes-for-this-document” issue. It “just” needs 1 or two (for inverses) DT-defined attributes on most elements.

If ever there were an example of “Worst UI ever on the planet. And I mean EVER,” I would humbly submit Tinderbox.

It looks like it would have been bad even in the days of System 6, when it was apparently designed.

Another vote from me. Tinderbox looks like a brilliant idea, executed in the style of somebody vomiting random UI pieces from the 1990s into a Mac app. Blech and awful, doesn’t begin to do it justice.

Yikes! A coupla guys with their berets on. Better run!

(11,100 hits for “tinderbox awful ui” on Google, and with these guys: 11,102)

korm, kalisphoenix, and Bill_DeVille - Thanks for the replies. I found kalisphoenix’s explanation of the differences between the two main philosophies for database management very helpful. This is all stuff I got an impression of from reading the forums. Now different working styles, and how they manifest in DEVONthink database philosophies, makes a lot more sense.

I think I’m more like Bill in that I will end up with a big, sprawling database. For specific projects though, I’ll split stuff off, as korm suggests, and refer to kaliphoenix’s method for maintaining order. I have a lot easier time maintaining order when the project has an outcome defined. Until things get to that point, gathering materials just has to be more ad hoc. The finder might be fine for this, but one thing I love about DEVONthink is using it to hold stuff I am currently processing and have yet to process. I can see how this kind of use would make the database more difficult to use for focused projects, but hopefully splitting off new databases will help with that.

And it’s nice to know that as limited the UI of DEVONthink is, I’m at least getting something better than I would with Tinderbox!