Fragmentation
August 3, 2007 – 7:15 pmCatherine Devlin’s birthday wish list expresses a desire for a tag-based file access system, for (re)organizing her folder structure and tracking files to hang onto for migration and other purposes. Sounds good to me! I’d love to see a TortoiseCVS-style application I could run on top of Windows Explorer (on my work laptop), and a similar sort of thing for my Linux box at home. Tag things as I work on them, and then the ability to pull up items by tags/ tag intersections.
I see this as different from something like Google Desktop searching — sometimes I want to add my own (personalized) metadata to files, and tags work pretty well for that.
5 Responses to “Fragmentation”
I’m still amazed that this sort of extremely obvious feature has yet to be implemented in a truly usable manner yet. I suppose part of the problem is the multitude of different file systems, but still you’d think at some point an agreed upon relatively standardized method for adding simple tag metadata to files would have emerged by now. Certainly that *has* happened for a few of the prominent image file formats (like JPEG)…that there’s no corresponding generic metadata system for all files is almost unbelievable. We have a bajillion DRM schemes developed to keep us from accessing our data, but still no cross-platform metadata system for helping us to find and organize our data.
By Brian Carnell on Aug 3, 2007
Brian, I’m not convinced that we need metadata *attached* to the files, as long as there’s a way to point unambiguously to particular files with ones’ metadata. E.g., in some version management systems, each file under revision control has metadata attached (but the system just gives you the content you checked in when you do another checkout, without the metadata, but in others, the metadata is managed separately altogether. Or, for another example, with tag sharing as in Connotea, the tags I have are maintained by Connotea. Nobody needed to change the standards to allow me to have Connotea tags on their web pages and blog entries. So, I was picturing a personalizable tag-based file system that knows about, but does not need to modify, different operating systems’ file systems.
By configures on Aug 7, 2007
“Brian, I’m not convinced that we need metadata *attached* to the files, as long as there’s a way to point unambiguously to particular files with ones’ metadata. E.g., in some version management systems, each file under revision control has metadata attached (but the system just gives you the content you checked in when you do another checkout, without the metadata, but in others, the metadata is managed separately altogether.”
For my local file system, I’m extremely wary of metadata systems that do not embed the metadata within the file itself, as you then become dependent on that system for the metadata. For example, the first version of Photoshop Album (which is now a subset of Photoshop Elements) had a very nice photo tagging system built into it. But, like a lot of photo tagging software, it stored the metadata entirely within a database rather than embedding the metadata in the files. That’s fine as long as you’re committed to using that particular app for all your photo needs forever; when you decided to switch to a different product, however, too often you end up having to start over. Another common area this causes problems is song rating metadata in iTunes where people invest a lot of time rating songs only to find that a corrupted database or switching to another app renders that metadata useless.
For example, I think it is neat that people can use things like connotea to tag web pages, but imagine if rather than allowing people to tag or add categories to posts on their own weblog, that Wordpress simply said “well, go use connotea or del.icio.us if you really want to tag the content on your blog.” Yuck.
Clearly both MP3 and JPEG show that it is possible to get everyone to generally agree on how to implement data cross-platform into the file format itself. A standard for a simply adding, say, a single text field to all other files would greatly improve our ability to tag and organize files. Its kind of silly that I can tag a picture of an action figure as “Superman” and that will just work pretty much across OSes, but I can’t do the same thing for a text or other file.
By Brian Carnell on Aug 8, 2007
I think WinFS (http://en.wikipedia.org/wiki/WinFS) was supposed to solve this, but they dropped it, sadly. I’ve always been interested in advanced file systems, but I’m very biased: I think the OS’s file system should built entirely on the link model – files connect (via tagged links) to other files. Actually, if we didn’t have application islands, we’d have files linked to *all* data in my world: People, events, things, etc. A kind of user-level OODB or Wiki. Naturally the links would work across networks. Maybe it’s just too hard to do.
By Matthew Cornell on Sep 4, 2007