[ThomPete]

I really don't hope that is the future of the desktop. I mean really? Is that what we have to look forward to?

In my world there can be no talk about the future of the desktop unless:

The desktop metaphor and the current filesystem disappears.

My machine starts to monitor what I do and actually use this (The Ghost Protocol)

The machine starts to connect everything I do and build contextual maps automatically. For instance, I receive a picture in my mail and throw it into photoshop. When I then want to retrieve it I can not only look for name.psd but also for the context (Phil send it to me by mail)

Then we can talk about a the future of the desktop.

[sliverstorm]

Why is everyone interested in the filesystem going away? As long as users posses more than 20 files, organizing them is helpful.

Notice one of the most requested features on the Kindle was folders for the user interface. The result they gave us isn't folders, and the actual filesystem is abstracted, but it's effectively the same.

[cookiecaper]

A few other commenters have explained it pretty well, but I think the main problem is that things get lost way too easily. Most filesystems are just way too vast and even the strictest user will eventually end up misplacing a few files. And when that happens, it sucks.

I think the solution is not to destroy the concepts of files and folders, but to become increasingly dependent on tagging and context-sensitive searches. Most DEs have shifted to this approach; there's Spotlight in OS X, whatever that search on the start menu is called in Win7, and various daemons/systems in Linux (Beagle, Nepomuk, etc.). There are third-party utilities too, like Google Desktop Search. As those things improve, and as we see more work to narrow the filesystem concept to only things that are usable to the user and allow automatic tagging of content, usability issues will go away.

[megablast]

The majority do not understand file systems, that is why. What do you NOT understand about this? Why do people keep making this comment.

I like file systems, I understand them, but I used to program dos interrupts, and actually reading sectors of the drive. But I can understand filesystems, a lot of people don't.

What people want is an all their photos grouped together, all the vides grouped together, all their documents together. They don't understand the different between the desktop, user folder, trash can etc...

[extension]

So when I want to attach something to an email, I have to go to the app for whatever kind of data it is and find the "share" button, which will be in a different place for each app, if it's even there at all. Ditto if I want to copy it to a USB stick, add it to my Dropbox, upload it to a website, etc.

The fundamental ideas behind a filesystem, packaging all data in a generic container and allowing arbitrary grouping of the containers, are extremely straightforward and intuitive, not to mention incredibly useful.

What throws people for a loop is having to share their filesystem with magic invisible gremlins that leave inscrutable files all over the place. Most filesystems come pre-loaded with mountains of this junk, which the user is made keenly aware of when they are banished to some small niche directory. But the gremlins won't even stay out of the niche; they are constantly creating folders and dumping mysterious config files there, behind the user's back.

The solution to this problem is clear and simple. A filesystem, from the user's perspective, should be empty when it comes out of the box, or at most contain a few items that the user knows exactly what they are. From then and forever, nothing should ever appear in the filesystem except what the user explicitly put there. The user is free to create, move, copy, rename or delete any file without unexpected side-effects.

It seems obvious to me that filesystems should work this way and there would be no big usability issue if they did. The problem is, we have all this old baggage attached to the filesystem, both conceptually and practically, and we are throwing out the baby with the bathwater.

[ams6110]

I think you've hit on why so many users keep everything on the desktop. They can have files and folders there (which most people CAN understand, by the way) but they don't have to deal with or see the rest of the file system (which they will inevitably stumble into pretty quickly using Explorer (or whatever file navigator app).

[extension]

They either stumble out of the Desktop or get lost in the pollution. They download a photo and edit it in Power Goo. When they want to find it again, do they look in Desktop, Desktop\Downloads, Desktop\My Pictures, Desktop\My Documents, Desktop\My Documents\Power Goo, etc?

[ThomPete]

I think you are missing the fundamental point here.

The current file systems whether MS or OSX isn't made for the amount of files that the average person deals with today.

A better system would be to allow you to search on multiple axis. But more importantly a better system would follow what you are attempting to do and limit the option space quite significantly.

[megablast]

WHen you want to share something, it will ask you what type of file you want to share, you pick photo, then select from the list, which you can filter in a number of different ways.

You could filter by name, date, who is in it.

[extension]

So you need special case code to handle each and every data type, even when you're just going to flatten it into an opaque blob in every case? And there would be a PDF store, and a Powerpoint store, and a Ruby script store, and an NES emulator saved game store, each with it's own completely independent API?

No, this is ludicrous and backwards, for so many reasons. Obviously you want some sort of generic data interchange layer, and voila: filesystem.

I can imagine a filesystem far more sophisticated than what we have today, with metadata, indexing, content handlers, searching, filtering, and so on. But the essential foundation for all of this is a generic package for user data.

[ThomPete]

The web is fundamentally a bunch of files.

Would you rather browse around the net in hierarchies or use google to find the files.

File systems like those we have today are fine when you don't have a lot of content. But it simply fails as a design system as soon as we move into terabytes of data.

Why do you think that Spotlight and QuickSilver are so popular on the mac?

[kaens]

Slightly OT, but what I really want in userspace is something that is good at searching/setting extended attributes for media files. I like my Artist/Album/Song dir structure and all, but it falls flat for some types of music, and I want to be able to search and group my movies by director or actor.

Add another entry to the "start learning what you need to to implement this" list, I guess.

[albertzeyer]

I have never actually seen any computer user, however incompetent, who has not organized his files in some sort of directory hierarchy.

[pohl]

Sounds like iPhoto and iTunes. Doesn't really pass the porn test, though.

[logic]

Your point about the system watching what you do and doing something useful with that information reminded me of some work a few years back by Nat Friedman on "Dashboard" that had me pretty excited at the time: http://nat.org/dashboard/

It was a shame to not actually see that go anywhere other than the prototype stage.

[ThomPete]

That sure is a shame as it seems like it would be pretty useful.

There need to be more thinking going into non-interaction. Actually having to interact with the computer as much as we do about the things we do, surely can't be the goal in itself.

I want my machine to make plenty of choices for me and present me with information not have me do it manually. I think that is what pisses me off the most right now.

There is something wrong and primitive about normal human beings still having to interact with their machines to get proper information.

[dkersten]

This idea is something I've talked about many times with various people. Even posted about it on stackoverflow and blogged about it once. Computers are better at boring and mundane maintenance tasks than we are - so they should do them. Users shouldn't have to think about such low level concepts as filesystems, directory structures, saving files and so on. It should be managed behind the scenes.

[ams6110]

I think anything beyond the simplest of automatic organization schemes will end up being confusing. I like to organize my files the way I want them organized. I can probably deal with all my pictures showing up under "My Pictures" but if it starts to try to do something like classify my documents by content it's going to depart from what I think "makes sense" pretty quickly.

[angrycoder]

"My machine starts to monitor what I do and actually use this (The Ghost Protocol)"

Yeah, the last time they tried this it was called Clippy.

[mgunes]

Take a look at Zeitgeist and GNOME Activity Journal, which haven't been accepted as modules into GNOME 3.0, but are already usable and will almost certainly be part of the 3.x series in the next few releases.

http://live.gnome.org/Zeitgeist http://live.gnome.org/GnomeActivityJournal

[anigbrowl]

It does seem like there will be some tentative steps in this direction. But I agree we need to see more - I like your idea of a content-neutral tagging agent that draws inferences based on my use patterns rather than a committee's agreed use cases.