uwMike » Archive » Tagging Beyond

I'm in Waterloo at the moment, and next available to work in September 2008.

Tagging Beyond

June 27th, 2005

Update: This article is missing its diagrams. I’m looking into recovering them.

Tagging. From del.icio.us it has spread like a worm. “Don’t bother with folders or albums or categories, just tag everything.” It’s chaos.

And Zeldman doesn’t like it.

But wait a second– tagging works because it’s simple enough that people actually do it. It’s straightforward. On the class website I administer, the photo gallery is an absolute disaster area because no one knows how to file anything or even where it should go. So in a system that’s supposed to enforce organization, there’s no organization at all.

Chaos

My clothing was a mess until I decided that rolling was as good as folding. Now my shirts and shorts are on the shelves, rolled, instead of on the floor, not folded.

Every time I have to hunt for a document on my computer, I grimace at the files scattered all over the place. There’s a “downloads” folder, there’s the desktop, there’s “Mike’s Documents”, there’s a web-dev folder, plus countless other nooks and crannies where something may or may not be hiding out.

A Tagged Filesystem

It occurred to me that it would be extremely cool if an operation system could be devised where the hierarchy of folders is completely removed in favour of a filesystem based on tagging.

What is the power of tagging, after all? It’s the ability of a particular entity to have as many or as few tags as desired. For example, consider the following ‘tagged’ system:

In the above, I’ve constrained the wallpaper and photo tags so that applying either of them applies the picture tag implicitly. After all, can a file be a wallpaper or a photo without also being a picture?

But that’s folder-like behaviour– something completely constrained inside of something else. The real power of a system like this comes with its ability to express so many different ’states’ with so few axioms. With only five tags, I’ve represented what it takes ten or fifteen folders to sort to my satisfaction.

Besides the simplicity; imagine opening up an Explorer and instead of navigating a folder tree, you just see a tag-cloud. If you know precisely what you’re looking for, you can type queries like:

mike & funny & picture & !photo (funny pictures I’ve made that aren’t photographs)
(wallpaper | photo) & mike (photos and wallpapers by me)

Does this make sense? Isn’t it extraordinary? Instead of having a scrapyard folder for ‘old work’ or ‘archive’, you can just slap the archive tag on a document and it disappears out of the views that exclude that tag.

What about the messy-desktop conundrum? Instead of having a load of shortcuts to executables, how about just a desktop tag that causes the executable itself to show up in the desktop view. The program file co-exists with everything else having the program tag and the tag of its publisher, and also shows up on the desktop. When its time to banish it, it simply goes continues to exist wherever else it did before.

Security

I’m not sure what the implications would be on security. Most operating systems have quite sophisticated file-access rights. How this would translate into a tagging system is unknown. Have tags that are ‘owned’ by particular users? Have special hidden and system tags that prevent certain types of operations?

Implementation

I’m sure this is nothing more than a curiousity. But as more and more of our modern computing experience moves away from the desktop, webapps offer a fresh start. While the FreeBSD machine running the program may be on ext3, it could be servering up online briefcases that look like vast collections of intricately tagged and inter-related files.

It’s sure worked for Flickr.

Mike

← Earlier Later → Posted at 8:52 pm to Internet.

Discussion

thats exactly the things that came to my mind, when i first saw the tag thing

tagging suits more the way a human organizes information (inside the brain..) well at least i think so

Posted at 2:48 am on June 28th by Michael.
Dude, I am honestly jealous of how much you think about things. I get to a certain point in my trains of thought where I inevitably get T-Boned by some other train, such as “My feet are tired” or “Hey, boobs!” Good on ya’. Hope you had a good Canada Day Weekend, too, dude.

Posted at 2:45 am on July 4th by Alex Spence.
Well… these are the things I think about when I should be thinking about, you know, Calculus. It’s all part of Structured Procrastination.

Anyhow, yeah, it was a great weekend, we need to hook up at the end of the summer.

Posted at 3:38 am on July 4th by Mike Purvis.
If you basically associate the tags with the directories then you can extend the paradigm to old applications. Here:

Move: would just add/remove tags associated with the file.

your current working directory is just the (ordered) list of tags that you need for things like ls, file creation and relative paths to work. It needs to be ordered so that you can use ../.. type constructs.

ls (or dir) would return you the files that match the current list of tags exactly.

creating a file would give the current list of tags.

This paradigm has the advantage of having docs/uni/calculus/project being the same as docs/project/uni/calculus

Posted at 5:45 am on July 4th by Antonio.
Also, take a look at
http://chneukirchen.org/blog/archive/2005/04/a-tagged-filesystem.html

Posted at 5:46 am on July 4th by Antonio.
Interesting, thanks for the link! I deliberately avoided discussing implementation, but I’d been tossing around the notion of multiple symlinks/hardlinks.

Perhaps such a system would work well if the ‘explorer’ or tag-browsing application maintained some kind of index such that it could keep track of the more complicated relationships. The system is useless, of course, if you have to know the exact tags on a document in order to find it.

And ultimately, it would simply turn into another piece of data put to use by tools like Spotlight.

Posted at 6:04 pm on July 4th by Mike Purvis.
Interesting thoughts, Michael.

But your premise seems to be, that because tags work well for collections of pictures, they should also work well with collections of files.

Well, I dispute your premise.

Like Zeldman, I don\’t like tags, either.

And OK, I\’ll admit at the outset that I\’ve never really tried using tags (I\’ve just briefly tried them with Picasa), so maybe I don\’t know what I\’m talking about. But here are some of my concerns.

But wait a second– tagging works because it\’s simple enough that people actually do it. It\’s straightforward. On the class website I administer, the photo gallery is an absolute disaster area because no one knows how to file anything or even where it should go

On my hard drive, the opposite is true.

For me, putting my photos into folders works (well, it sort of works) because it\’s simple enough that I actually do it. OTOH, tagging pictures is (or seems to me to be) so weird and unnatural, that I could never imagine doing it on a regular basis.

My photo uploading software forces me to put my pictures into a folder, each time I upload another batch of pictures, and my present system of hierarchical enforces a rudimentary kind of organization on the process.

Yes, folders have their drawbacks, particularly as you say because some pictures seem to belong in several places at once.

But if you abandon folders, in favour of using a tagging system, it seems to me that the cure is worse, much worse, than the disease.

With my folders, I won\’t always remember where I put a picture, and if I\’m away from my computer I will not even remember what my folder hierarchy is, but for me the big advantage is that I can * browse until I find what I\’m looking for*. Just a few clicks with Windows Explorer and I can see all of my photo folders, and if I\’m looking for a special picture I usually I can narrow it down to 2 or 3 folders where it\’s most likely to be, and in most cases, in under a minute, I\’ll have found it.

But with a tagging system, I presume that you an ONLY do searches, you give up that ability to browse, don\’t you?

So I would have to start constructing a search string:
```
Joel + (tree fort or island or camping or holiday) not 2003 not 2004
```
and then see what gets dredged up by the search. But if I had tagged my lovely picture \”tree house\” instead of \”tree fort\”, and \”Kids\” not \”Joel\” (because Virginia was also in the picture), then I might waste a lot of time fiddling with my search parameters and still I might never find the picture I\’m looking for.

With a folder system, you can be confident that (as long as you haven\’t deleted the file) the picture is somewhere on your hard drive. If you keep looking, you\’ll eventually find it. But with a tagging system, the picture might truly be \”nowhere\” on your computer, if you were in a hurry the day that you uploaded your camera so no tags got put on that batch of photos. Hence, it seems to me that the system only works if every single picture gets tagged, with meaningful tags, and if you maintain an rigourous system to enforce consistency (let\’s see, is she \”Nana\” or is she \”Grandma\”, is it \”Church\” or \”Rez\”, etc). Then I suppose you have to periodically retag your old photos if the tag system gets updated (eg, I might have 100\’s of pictures tagged as \”Michael\”, but now he goes by \”Mike\” so I need to change the old tags, etc) To me, it sounds like a nightmare system that requires a huge an ongoing investment of time.
Posted at 1:01 pm on July 12th by Dad.
I guess my presumption would be that all information already available through a folder approach would continue to be available through tags. So each file would have created/modified stamps, and then unloading software would automatically stamp a tag for the ‘Unload Date’ on it. And then, for even more auto-tagging, you could have information such as ISO and f-stop turn into tags also.

Ultimately, producing queries is not a very intuitive interface. I would envision the ‘Explorer’ as being basically like Picasa is already, but with a lot of visual functionality such as:
- A lot of tag-application could happen through ‘matching.’ Put a couple files in the Lightbox, and then make all of them take on each others’ tags. (for example)
- The browsing ‘default view’ would be ’show all.’ So you’d have to have a way to, within a few clicks, get to the specific view you’re interested in. For example, click on a file and say “show me only ones files with at least these tags” or “show me files that have at least one of these tags”
And then the query could be built up visually in a ‘location bar’, where you can easily just double-click a particular ‘rule’ to banish it or whatever.

I think, for me, it’s basically a philosophy of “anything’s better than nothing.” So having zero organization, where my only browsing is by date is the current mechanism. Even if I could just tag ‘family’ pictures, that’s something I might do that would be helpful. But to move them out of their current by-date folders would already be violating the (simplistic) organization that’s been put in place.

Yeah.

It was funny that after I wrote all this conjecture, I started reading about WinFS, which may well include this sort of thing as a possibility.
Posted at 4:27 pm on July 13th by Mike Purvis.
Mike, you beat me in mentioning WinFS :)

Microsoft’s new idea for a file system is that the entire operating system will store all its data in a huge relational database (probably powered by SQL Server underneath).

This all sounds great. Unfortunately, it was too big and complex to finish in time, and it won’t be in Longhorn. Hopefully it’ll get done soon, the current file system mess is something that’s been driving me pretty nuts, too.

Posted at 11:55 am on July 14th by Alex.
Do we have to abandon the tree-like file system? Why not build a tagging system on top of it? Why change the operating system? I think it could be done with some kind of an application on top of the usual hierarchy and the security & access policy stuff will remain as it is. hmmmm…

Posted at 10:36 am on June 7th by Baris Ozdil.

You can use Markdown for style. I love hearing from readers, but please don’t hijack the discussion, use offensive language, or try to sell anything.