Edd Dumbill writes:
The amount of information we store on our computers is increasing rapidly. It’s not just the volume that’s increasing, either. Over the last few years we’ve added a significant number of new types of data we rely on computers for. Photographs, music, grocery shopping, banking, video. Add to this our increased reliance on data sources on the web. With a few exceptions, most of our interactions with companies, and increasingly government, can be carried out online.
With this increased reliance on our computers come associated dangers and difficulties. Leaving aside for the moment the problems of archival and obsolescence, two of the major challenges lie in findability and integration. Let’s look at findability first. In the real world we have many cues as to where to find things. We use special files or boxes, we leave things in particular places, we recognise the colors and branding on forms, we sort into piles. In effect we are attaching metadata to our information, or taking advantage of metadata already there. Metadata is data about data. It’s not the numbers on our bank statement, but everything else about it. The color of the ink, the typeface used, and so on.
Such real world metadata is very expressive. On our computers, the world is more restricted. We are mainly reduced to filing things inside a tree-structured filesystem. Imagine ordering everything in your house that way. Not only would it take a long time to do, it doesn’t really make a huge improvement to our lives. Pure formal organisation is no match for the blend of intuitive, cognitive and formal organisation we use in real life. Nevertheless, until now that’s the best computers have to offer, in general.
For findability, having extra metadata allows you flexibility of presentation and organisation of data, and extra parameterisation for search. Rather than making you do the filing, the computer can use the metadata to show you views according to your current needs. Though we can’t and shouldn’t try and ape the world, we can introduce different locality-based methods to make working with data on the computer easier. This can be done dynamically in many ways, as opposed to establishing a static filesystem hierarchy.
The integration story is even more exciting than improving findability. With sufficient metadata applications can deal intelligently with data they’ve not encountered before. For a really basic example of this, consider the MIME type of a document downloaded from the internet. The browser is most often capable of taking an appropriate action, rather than making the user do the work. Integration works best when there are agreed metadata standards, such as MIME types, filesystem attributes or EXIF data. The history of personal computing software has unfortunately been largely centered around creating isolated packages, with interoperability strictly controlled and used as a commercial lever. We have a long way to go.
My premise then is that more metadata is required to create a usable desktop for users and manage the increasing volume of information stored in our homes. That’s a conclusion other people are agreeing with, too. Microsoft’s next-generation operating systems will ultimately include WinFS, a file system supporting the attachment of arbitrary metadata to files. Mac OS X is acquiring similar functionality. ReiserFS has been trying to do it for ages. Closer to GNOME, there are projects like Dashboard, Storage and iFolder.