Perhaps keyword tags are an artifact of the application domains involved, and aren’t so intrinsic to the idea of tagging. In Flickr, the items being tagged have no text in them, and so to get keyword search to work, you have to explicitly add keywords. In del.icio.us, you’re tagging URLs, which may have plenty of text but aren’t directly searchable; that is, you can’t ask Google to search only within pages you’ve bookmarked.

By contrast, if you’re tagging your own local textual data, and have full-text search available, you don’t need to tag with keywords. A simple search will do what you need, so there’s no reason to bother tagging, except for folder-style classification like “toread”.

Of course, now and again you may want to label something with a keyword that its text doesn’t contain, and there will still be other sorts of things (like pictures) that need keyword tags. To quote my own comment on Tom’s post: “The beauty of tagging is that folder tags and keyword tags can coexist in the same system. I think that’s a big improvement over having folders in one place (the GUI) and keywords in another (the search box).”

Computers suck. They’re a hassle to use (especially for non-geeks). The advent of the GUI has helped, but we still have a long way to go. I think that replacing folders with tags will be a usability win, and I want to test that hypothesis by building Trifle, while adding other goodies like versioning and search. I also believe that users should be given the tools to build their own UIs, extending the power of the spreadsheet to something more like a database, but without the relational-model overhead.

And yes, it’s not just about doing something noble for the users. I freely admit that I’m scratching my own itch: I want to build Trifle because I want to use it.

There is no shelf. We can organize the virtual world any way we please. Most metaphors we have now, though, are based on physical models. It’s time we got more creative, and started building interfaces that work in ways that physical reality cannot.

The files-and-folders (desktop) interface was based on a physical model to make it easier to learn. Will nonphysical models be harder? Not necessarily. Take the spreadsheet, the all-time success story of a new conceptual model. At a glance, it has the reassuring familiarity of a piece of lined paper, but type in one formula and you enter a new world. The real conceptual model of a spreadsheet has no physical analog at all — yet millions can and do use it.

My first thought was to turn folder names into tag names, but that quickly breaks down. For instance, “Other” can be a reasonable folder name, but it makes no sense as a tag name. And what about nested folders? Should a path be concatenated into a single tag, so “/foo/bar/baz” becomes “foo-bar-baz”? If you do that, each item gets only one tag, and it’s no different than having them in folders. But the alternative has its own problems—some subfolder names make no sense in a global context.

There’s no getting around it: I think of different terms when I think “What should I name this folder?” versus “What tags should I put on this?”. But with five-hundred-odd bookmarks to migrate, tagging them all by hand isn’t exactly appealing.

One thing I could do is take advantage of the aggregate smarts of del.icio.us. Presumably a sizable majority of my bookmarks have already been tagged in del.icio.us—so why not write a little script to tag all of my bookmarks with the three most popular del.icio.us tags for each? It’s a fun idea, but I doubt I’ll do it. The whole point of having a personal tagbase is to use tags that have meaning for me; of course, that means that I probably will have to tag those bookmarks manually, sooner or later.

David Weinberger posted the first section of a long piece on tags he wrote for the latest issue of Release 1.0. At $80 for a single copy, I won’t be rushing to buy it, but the section stands on its own, giving a historical and philosophical context to tags.

Matt at Peer Pressure considers the meta-issue of how to represent information about tags themselves. He winds up proposing metadata about tags (tagged tags?), then nails the main issue: “The big question is how to present all these new options to the user without making the whole thing hopelessly more complicated.”

That’s the very reason I keep waffling about whether to include metatags in Trifle. Still, if tags have their own URLs, and anything with a URL can be tagged, metatags are inevitable, aren’t they?

But spatial metaphors have been unable to scale as the quantity of information has exploded. The spatial-folders model of the standard GUI is overtaxed already, and would be even more so if information was stored at a more granular level. Zooming interfaces are touted as an improvement, which they probably are, but they still don’t solve the essential problem: each object lives in only one place. (Symlinks/aliases/shortcuts don’t solve that problem, either: they just ameliorate it.)

In the earlier days of the Web, it was easier to find what you were looking for by poking through Yahoo’s hierarchy than by using a search engine. Now it’s easier to just Google for a few words and get what you’re looking for. And so the primary interface of the Web has moved from spatial-hierarchy to search. I’m convinced that this same transition will happen at the level of individual-user data, but don’t take my word for it: look at the investments Google, Apple, and Microsoft are making in that direction.

Search-based interfaces eliminate the need for each object to live in only one place, and thus solve the fundamental spatial problem. But this also gives up one of the better features of the spatial model: the sense of security that comes from knowing “I’ve put this thing here, and that’s where it’s staying.” By contrast, a search-based system is more like “I’m dropping this in the ocean, and hoping that I can fish it out later when I need it.”

And this is where tags come in (meaning personal tags, not social tags/folksonomies). Tags can give you back that sense of security where simple search cannot. You don’t have to worry that something relevant might fall through the cracks because it didn’t have the right word form to match your search. Of course, the success of your search depends on having things tagged correctly, but that’s no different than looking for something in nested folders. Besides, you still have full-text search to fall back on.

Search is already replacing spatial hierarchy, from the Web on down; tags will only accelerate that transition, and, with any luck, lead us toward interfaces that are more usable–and more useful.

I could just replicate the hierarchy in a single tag, so that /Seattle/Libraries is just “Seattle/Libraries”. But then tags would have no real advantage over the original folders. On the other hand, won’t making separate tags for “Seattle” and “Libraries” make it harder to craft meaningful queries?

For instance, suppose I want to query for items tagged “Seattle”. I’d like to group together the “Libraries” items in the results, but how would the system know that “Libraries” is a useful grouping within “Seattle”?

Perhaps I could encode that information in the query, specifying grouping tags rather than leaving it implicit in the tag structure. Queries would be more verbose, but also more flexible.

The worry is pointless, IMO, because Clay Shirky is right: “It doesn’t matter whether we ‘accept’ folksonomies, because we’re not going to be given that choice. [emphasis his]” Tags and folksonomies have all the hallmarks of a Worse Is Better solution, and by the time a working Right Thing comes along, tags will be too entrenched to displace. I personally think that’s a good thing in this case, but whether it’s good or not, we’re stuck with it.

Folksonomies do have scaling limits (and just wait till the spammers start arriving in earnest!), but there’s a simple way out: folksonomy by subscription. What if, just like you subscribe to someone’s RSS feed, you could subscribe to their tagbase? A tagbase aggregator could collect information from the people you’ve selected and integrate it into your personal tag set, either invisibly or with a marker (e.g. tags from Joe get a “Joe_” prefix). You could even use RSS as the mechanism–it wouldn’t be all that different from a linkblog.

Cocoa I don’t particularly care about yet, since I’m not a Mac user. But I suppose it’s possible I could fall head over heels in love with it, once I get a Mac. (Buying a Mac has turned from “if” to “when” for me; the question at the moment is whether to wait for Tiger.)

Connectivity includes having hooks into whatever apps and services you use, so you can organize your information in the midst of whatever you’re doing. It also includes sharing data by syncing with a server, and by generating RSS feeds. These, I think, are all important. The more ways you can access and process your data, the better.

Tags cover not only simple labeling, but things like nested collections and automatic retagging. To me, tags are the heart of what we need. They’re the means of escaping the location-based metaphors that are holding us back.

Trifle will be focused on tags and connectivity, and in that (implementation) order. It may not ever become Merlin’s ideal productivity app, but I hope that it becomes mine. Now if I only spent as much time building it as theorizing about it…

… in the period between the invention of the printing press and the invention of the symlink, we were forced to optimize for the storage and retrieval of objects, not ideas. Now, though, we can scrap the stupid hack of modeling our worldview on the dictates of shelf space.

That, in a nutshell, is what’s good about tags: they’re simple to use and understand, yet far more flexible than schemes based on a location metaphor (like the file system).

Most of these posts talk about tags on a group level, as used by Flickr, del.icio.us, and Technorati. Often overlooked, though, is the value of tags on a personal level. (To be fair, Many2Many is about social software, so you would hardly expect them to be talking about the individual case.) I’m convinced that applying tags to private personal stuff is just as important as applying them to shared stuff. That’s why I’m building Trifle: I want to organize my stuff.