February 9, 2009

Alex Payne recently wrote The Case Against Everything Buckets, which earned a rebuttal from Buzz Andersen.

Alex Payne’s post is ranty and prescriptivist, but there’s a nub of a good point buried in there: “Computers work best with structured data…With an Everything Bucket, you … miss out on opportunities to do interesting things with data”

What Alex Payne means by an “everything bucket” is a notebook-style application that you dump all your random notes, clippings, web links, pictures, etc into. There are a lot of independent software developers making interesting apps that fall in this general category. I don’t use one myself, mostly because I don’t need to manage big piles of notes.

I’ve always gravitated towards structured data—I put my contacts’ info in my address book, my links on delicious, and so on. And this can pay dividends—on a Mac, if you use the Address Book, other apps know where to look for your contact info and can do “interesting things” with it—like sync it to your phone, or check whether incoming e-mail is from someone you know. That’s what Alex Payne means by “interesting things.”

Here’s what’s funny, though: the distinction between the everything bucket and structured data may be a false dichotomy. That is to say, there’s still a difference in how you would get to the endpoint, but you’re still getting to the same endpoint of being able to do interesting things with your data. Those two paths are what Mark Pilgrim referred to as million-dollar markup vs milllion-dollar search.

Macs today (and also about ten years ago, right before the switch to OS X) come with “data detectors,” which will notice when a chunk of unstructured text contains something that looks like, say, a date, and will offer to create an iCal entry based on it.

Long before that, Simson Garfinkel wrote an app called SBook that looks like an everything bucket, but also attempts to do interesting things with your data. This is pretty much limited to contacts and related notes, but the idea is there.

Google searches can recognize mathematical formulas to give the results, personal names to give their contact details, musical groups to give their discographies, and so on.

If the software is smart enough—perhaps with a little coaxing from a person—to recognize the structure into which a chunk of data might fit, it shouldn’t really matter whether everything gets tossed into an everything bucket or meticulously sorted into multifaceted, hierarchical, schematized structures. The tools aren’t quite there yet, but there’s no technical reason it wouldn’t work.

Right now, though, it doesn’t work, and the benefits of those interesting things outweigh whatever cognitive load is associated with context-switching between different containers for different kinds of data.