| |
  |
| Why The File System Is Dead |
|
 |
|
posted by Editor on Friday October 05, @07:58AM
|
|
|
|
|
David Gelernter co-developed Linda, a coordination language for parallel processing which became an important tool for high performance computing. He is now developing a technology called Lifestreams, a data representation model that organizes content by time rather than space, so that users deal with their files through a kind of diary, rather than a desktop. Here is a transcript of Gelernter's keynote speech (PDF) at the 2001 PC Expo (you can also hear it in a Windows Media Player audio recording ). In it, he makes a compelling case that the desktop metaphor has outlived its usefulness in the web era, and provides details about the products he is developing through ScopeWare (formerly Mirror Worlds), the company he formed to commercialize Lifestreams. NOTE 10/12: FIXED BROKEN LINK TO PDF.
|
|
|
|
< Wiring Your Body To The Web | The Limits Of Command Line Interfaces > | |
|
|
|
Don't have an account yet? Go Create One. A user account will let you customize the site's content according to your preferences. It will also allow you to moderate the comments of other users.
|
|
|
|
|
|
This discussion has been archived.
No new comments can be posted.
|
|
|
|
|
|
|
|
|
Does this system work at all for non-textual data? It seems like you're fresh out of luck when it comes to organizing images or sound files, or even text data in non-standard formats.
|
|
|
|
|
|
|
|
|
|
|
|
|
I'm sure it's partly because the desktop metaphor has been deeply ingrained in my way of interacting with computers ever since I used my first Macintosh back in the 80's, but I'm having trouble grasping exactly how this gets represented on the monitor.
While the Lifestreams concept has some pretty good ideas (Sure, I'm always searching for files and URLs by date instead of directory name), it doesn't strike me as something that could completely replace the desktop/folders metaphor in just the same way that the desktop metaphor isn't a perfect replacement for DOS or *nix style command lines.
Fr'instance, in Unix/Linux all you need to to do create a new directory named 'foo' is type "mkdir foo" at the command line. In Windows you have to move the mouse up to the 'new folder' icon, click on it, and if you accidentally click the mouse before you type the new folder name you have to click twice on the icon of the new folder (once to select it, once again to indicate that you want to rename it) to set its name.
So what happens when I know exactly which file I want to use, but files are no longer organized in a directory tree but along some weird timeline? Am I forced to sift through files chronologically until the one I want goes by, wheras before I could just hit Win+R and then type 'C:\path\to\file\i\want' ?
This is my home page.
|
|
|
|
|
|
|
|
|
|
|
|
|
|
I saw a profile of the developer of this concept on "Big Thinkers" (regular show on TechTV).
Some here mentioned they had trouble visualizing what this would look like. What I saw during the show was basically a long chain of small windows - imagine if you opened up 30 thumbnail images in photoshop, and they were overlaid from one corner of the screen to the next. Each window was (I think) an "item" of the stream... they only presented it as an aside, and I wasn't paying close enough attention at the time to derive more details.
Mind you, that interface could be one of many or some sort of odd prototype, so I wouldn't judge the basic idea too harshly from that.
While I find the idea interesting, one problem I have with it is that I don't have a great "temporal" sense. I can probably think accuratley back about six months, past that things start to get fuzzy on the temporal timeline and even re-ordered. When I think back to events in childhood I almost always imagine them happening when I was 10 years old - I'm not sure why.
Furthermore, though storing things along a temporal axis might be useful for one person but I'm not sure how useful it would be to someone who had just come along. It seems like you'd almost have to "re-live" history through examing the whole timeline to really have a good grasp of how to retrieve information from a stream.
I'm not sure really though how different this is from a folder (which could be a stream) sorted by date, holding all of the files for a particular project. I already store lots of things that way, and it does work pretty well.
In general I think I like the approach of a multi-demonsional file storage, like what BeOS had - meaningful attributes of any sort that could be attached to a file.
I also think some sort of sym-link system is important, to be able to keep a file or sub-project in more than one category.
As an aside, I'd like to say I'm pretty excited about finding this website (which I was led to through a post on /.) as really new UI designs have been an interest of mine for some time, and I'm excited to see a place where people can present new ideas and talk about them!
|
|
|
|
|
|
|
|
|
|
|
|
|
|
LifeStreams has been out for years now, somehow it never has made it into mainstream computing. It has been repackaged a few times, and has been getting better in the process.
The major problem with LifeStreams is that it puts you in front of a pile of files. So the metaphor has been changed (from a spacial desktop to a timeline), but the problem is the same : 47 385 items on my disk. Why should I have to be put in front of 47 385 files ? Lifestreams does use user defined 'streams', but basically it all breaks down to a (large) list of files.
For the moment, file systems are organised in a hierarchical tree. That means that one object (file) can only belong to one category (file path + name). A few layers of metadata have been added over the years ('families', 'types', 'creators', etc), and 'Aliases' or 'Symlinks' can be used for a handfull of defined files (but get broken easily).
What we need is rich extensible metadata. A filesystem organised like a database. BeFS (the file system of BeOS) has been doing this for years. Apple invented the Meta Content Framework to deal with the problem. Unfortunately, they made a cheesy GUI layer over it called HotSauce (remember that?;) that was so bad it killed the project.
Today, new radical ideas are comming from (belive it or not) microsoft! Steve Capps on the new MSN Explorer (possible future .Net interface): "Folders are ridiculous!" he says with a snort.
"Computers have 20 things that are important, 10 things you use often and a bunch of crap. Let's put it all on one screen -go for it!".
The new interface of Explorer in XP is completely based on metafiltered content : my clients, my projects, my lifestyle, etc. But I don't think they let you create your own categories.
So what happens after you organise your information with rich extensible metadata ? Well you want custom, personnal tools to organise it for you : you want modular interfaces, called by the data you manipulate. Imagine a very specific interface by Adobe for graphic designers to organise their image files, etc. And to implement that, you need a standard protocol to let your data, applications and your modular interfaces to communicate. XUL and XForms are a start, but are very much focused on GUI widdgets for the moment.
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
I think time based attributes can greatly simplify storage and retrieveal of documents. For example, is a current project open or closed? Is an e-mail thread open or closed. When was the last time the doc was updated? There are currently no uniform tools to use this kind of system. There are beginnings - the history sidebar in IE, the Journal in Outlook, but they need to improve on them.
|
|
|
|
|
|
|
|
|
|
|
|
|
"Open" or "closed"? How does a time-based system tell you this? Some projects languish forever and never close, whereas some projects bloom and then die in two weeks. That's another layer of metadata, entirely separate from time.
Organizing files temporally isn't necessarily any better than organizing them spatially or hierarchically. Computers are smarter than that. They can do all of the above and more simultaneously. The challenge is to design an interface that doesn't make the user manipulate files - the computer can do it for them - but makes it easy for the user to manipulate metadata.
|
|
|
|
|
|
