Observed: The Death of the File System?

With their February 24th revelation of more features in the upcoming OS X Lion operating system, Apple may have taken its first steps toward an unfamiliar future… a future in which the file system does not exist.

Credit for this observation goes to Mike Rundle, who tweeted about being able to imagine “a future in which the Finder does not exist”. Documents would be associated with the apps that created them, like on the iPad. Mike went on to describe his vision in more detail, a vision in which users simply have apps. “Documents associated with them appear magically. Presto.” While this might sound like some kind of user experience utopia, I have a grave concern that eliminating a file system in this manner misses a huge audience.


While opening Pages to work on the family newsletter might make sense for casual home users of a computer system, it does not make sense in a professional context. In the professional world, we work on projects. Projects are composed of many different types of files. And yes, we might have the same apps open all day, but do we want to be forced to duplicate a hierarchy of information in every single application?


Besides, “projects” are just one type of organizational scheme. As a user experience designer, I’ve seen a lot of professionals in other fields organizing a lot of stuff in a lot of different ways. So even attempts at inter-app organization around the concept of a project, such as Microsoft’s Project Center, are not effective replacements for an infinitely flexible organization scheme like simple folders.

Some Wheels Need Reinventing

The conversation that Mike’s comments sparked led us both to the conclusion that we still need a high-level organization system of some kind. And that is the challenge. It’s a challenge because that problem has already been solved by the file system. The challenge is to solve it better.

At Interaction11, Tim Wood called for designers to reject the “complacency artifacts” of the past, design patterns that have lost relevance in the modern world but continue to be used simply because that’s how things are done. He encouraged us to be bold enough to reinvent wheels that need reinventing, and that’s exactly where we’re at with file systems.

Gestural user interfaces, effortless portability, and ubiquitous network access… All these things and more are redefining how people interact with technology. UX designers need to recognize this and push themselves beyond the limits of our vision. Yes, we absolutely must continue observing people interacting with technology, analyzing those interactions, and synthesizing solutions that work in context. But what’s even more important now is that we rely on our raw creativity for that last part. There are old problems out there that need to be solved in new ways, and the file system is one of them.

Header image courtesy of Tim Wood.

Fred Beecher

Fred Beecher is a Lead User Experience Consultant at Evantage Consulting in Minneapolis, Minnesota. Fred has been working in user experience design for 13 years, doing user research, information architecture, interaction design, and usability evaluation for a diverse array of clients like Medtronic, UnitedHealthcare, 3M, RBC Dain Rauscher, General Mills, Thomson Reuters, National Marrow Donor Program, and more.

18 comments on this article

  1. Rob Jones on

    Alan Cooper predicted this about 10 years ago in the first version of “About Face”. It’s taken a long time to get there, but his vision of it was pretty clear. His opinion was the filesystem was for the computer, not the user, and should be transparent to the user.

  2. Neil Cadsawan on

    It’s both funny and sad how Apple has taken a 180 turn on this subject – as a proponent of the document-centered OpenDoc to today’s application-centric iOS. In a doc-centered world, no one would care about, or even have any notion of an application. An application would be thought of more as a capability of the OS. This to me is more of what a “user experience utopia” would be – never having to think about an app in order to do something. Simply knowing what you want to do and being able to do it without having to think of what application you need sounds like a pretty interesting notion to me.

    But that ideal didn’t happen for various reasons and now we live in an app-centric world where if you delete an app, you delete all files associated with them. Unless of course you’ve saved them somewhere else, but that would need some sort of file system – even the cloud. This also creates the issue of one app not knowing that files exist within another app without some other sort of facilitation of a 3rd party or a file system.

    We value the sharing of information from a professional standpoint, shouldn’t we apply that as well with the technology we use? Apple’s iOS makes sharing information between apps difficult at best. It’s not to say that it can’t or won’t change in the future, it’s simply how it is today. This is fine for a consumption model, but erects barriers for the creation process. Shouldn’t we be looking for ways to make that easier?

  3. What about tagging:
    You can make your file organization as complex as you want (with multiple tags and stuff).

  4. Eric Basford on

    Interesting problems and nice tie-in with Tim’s excellent IxD11 presentation. I think Google has made some effort to eliminate our notions of the file system with Gmail labels first (eliminating mutually exclusive “folders”) and now Google Docs (which gives me an unpleasantly long list of different documents on the “home” screen but allows for tagging into multiple “collections). I’d be curious to see how the Chrome OS handles these problems.

  5. Stephen on

    Surely you just need a projects app? Most of a project is management, messaging, deadlines, todos. The projects app just needs a way to access the files from the Documents app, or Wireframes app. Tagging could probably achieve this at a lower level much better than a hierarchical filesystem, but a filesystem may do just fine so long as the projects app can abstract this away as efficiently as the Documents app.

  6. Rich on

    Quit thinking the machine is the interface to be focused on. The real focus is the computing power of us – remove restraints and augment our limitations. Isn’t that what we’re striving for.

  7. Diego on

    One thing that I like of the open source KDE project is that they try new ideas in their desktop environment (not radical ones, as you can find in a research project, but still new for a matured desktop environment).

    One of the things that KDE implemented in his plasma desktop is the ability to have “tasks”. So if you have a “project” you can associate multiple applications with a task and open them with one click.

    The task metaphor is limited in the KDE implementation, but still looks like a nice idea.

    I also like the approach of the old Squeak Etoys and his zooming interface (www.squeak.org, or for examples of etoys projects: http://www.squeakland.org), with a better implementation (and the ability to search) seems like a good way to work in a “file less” environment.

  8. Greg on

    My knee-jerk reaction is of course to exclaim what a horrible idea this would be. Coming from a unix world where the developer and user are at one with every aspect of the O/S, this just sounds wrong.

    But I am willing to grant that many forms of abstraction have turned out very well, and I can understand where abstracting the concept of inodes, file objects, directory hierarchies and so on would be beneficial. I certainly don’t generally know how my phone keeps its files during the course of its operation. And though there may be a filesystem on my phone, I’m geneally able not to know about it as an end user.

    Therefore, I have only two concerns about this concept:

    Cross-app usage of files. I know that I occasionally need to open files with apps that would generally not be associated with it. If the system over-abstracts, then I won’t be able to do my task as I need to.

    “Trust us…you don’t need to do that” Some systems work better if you drink the kool-aid and go with its design paradigm. But they do allow you to deviate if you need to. Other systems just take away all the functionality assuring you that you don’t need to do those things. If the implementation is the first (like phones) great. If it’s the latter..it would be a disaster.

  9. Thanks for the comments, everyone. Responding in one big one, myself…

    Rob: Absolutely. Hell, the notion of a *file* is for the computer. But what’s going to be challenging is to solve for the need of an infinitely flexible taxonomy outside of this established metaphor. Because while the file system is mostly for the computer, humans have made use of it too.

    Neil: My current thinking is much along the same lines as yours, but I’m more focused on content rather than capabilities. To me, an application as a set of capabilities is intuitive. An application is essentially a tool. When I want to build a prototype, I open Axure. When I want to write a letter, I open Entourage. What I’m most interested in is breaking down the *content barriers* between applications. Some do this already, like how you can access your iPhoto taxonomy from within other applications.

    Remco: While tagging can be useful, it places a pretty high cognitive load on most people that I’d rather burden a machine with. It might be a component of the solution, but I’m pretty sure it won’t be the solution itself.

    Eric: Yes, labels in Gmail are intriguing. They have the one-to-many attribute of tags, but also the locational aspects of folders. Assigning stuff to folders is a pretty low cognitive load task (hence the challenge of moving away from the filesystem), and rules (which most people probably WON’T set up) make it even easier to take advantage of their dual nature.

    Stephen: If projects were the only organizing principle, sure. What I was saying in the article is that, in my research, I’ve seen people organize things (and not organize them, natch) in infinite ways to support both personal idiosyncrasies and overcomplicated business processes. The need for a taxonomic structure for digital data is clear, and my observations have led me to understand that it must be infinitely flexible. Maybe there are “presets” where you can choose an organizing principle, such as projects, events, people, dates, etc. Who knows… the future is wide open.

    Rich: You’ve either been reading too much Bruce Sterling or playing too much Shadowrun. : ) But seriously though… isn’t that what a computer is? What a network is? It already augments our limitations on memory, communication, etc. For now, we’re much better at coming up with clever machines than we are at coming up with clever neurons. At our present level of technology, that type of thing is a great focus for research for neuroscientists. However, I’m not one. Besides, you wouldn’t want me to be one. I *have* played too much Shadowrun. : )

  10. Andrew on

    It depends. It depends on what kind of user you are and what you want to do. The more casual user is less file management-centric, whereas someone who is doing work would be more file management-centric. Work vs play.

    For many people play is a big part of their interaction with devices/applications and for them the file management issue is less relevant if at all.

    But some productivity apps assist you with doing away with file management per se. For example, some apps keep track of “recent files.” It’s likely that the most recent files interest you most, and they keep circulating to the top. And apps like Lightroom to an extent hide file management from you; if you choose to you can view ALL photos and browse for what you need. I can even do this with the finder – browse photos – but only folder by folder.

    Finally, don’t many of us simply have too much stuff on our computers (and devices)? Are the file structures really that organized and efficient. How many simply us search to find the file they need? I know I do a lot. If I can search or browse everything then how it is organized and hashing through that organization becomes less important.

    But,as you point out, I would be concerned about longevity if docs are correlated only with apps. They should be separate unless they are exclusive to each other. If a better app became available then how could I transfer app-related files to the better app?

    I like the idea of tagging to create dynamic groups, however that requires effort from the user and most people are lazy (I have nearly 20,000 photos in Lightroom and most are untagged). Passive tagging would be useful here.

    Lots to consider.

  11. DGEbel on

    Bah. Neil is dead on. Vendor lock-in is growing worse instead of decreasing. Apps and data *should* be independent of each other. Replacing an viewer or editor should not delete all your old creations.
    Users need some kind of “resource”-centric view, whether tags, folders, lists or searches. Any type of user could want, say, a speadsheet, a letter, some pictures & a video collected together, especially for backups or transport.
    More thoughts on my News blog.

  12. DGEbel on

    I agree, the apps should not be the focus, except maybe for very limited-tasked devices that only do one or two things… like phones used to be. General computing should be about “resources” like documents, pictures, etc.
    I can see tags or searches or similar becoming the normal document-centric user view though.

  13. Thanks for keeping the great comments flowing, everyone. I’m going to respond in one big one again.

    Andrew: You are right on. I think that what’s coming (partially heralded by the iPad) is a very strong separation between personal and professional computing. There was recently an article on Daring Fireball that articulates this view very nicely: http://daringfireball.net/2011/03/the_chair

    In a personal context, a file system and structure really doesn’t make a lot of sense. But in a professional environment, anything can be required. You mention passive tagging, which is a great idea because you’re right, people can’t be relied on to manage their content. What I’m thinking, though, is that we need to place the burden on technology and create some sort of protocol, an uber-XML that allows us access to content of different types within EVERY application.

    DGEbel: I think you and I are thinking along the same lines. I’d go further though and abstract things beyond even the document level and into straight-up content. See above.

  14. Pingback: All This ChittahChattah | ChittahChattah Quickies

  15. Pingback: from chaos » Blog Archive » “Tablets are a fad” and other failures of insight

  16. Pingback: Why the file system’s death is greatly exaggerated « Explain Technology

  17. Hi, I only just read this article, tho I was aware of Apple’s new direction (it’s also evident in iOS).
    It’s a topic I’ve been thinking about a lot over the last year. Readers of this post may be interested in my article on O’Reilly Radar, “Why files need to die” which explores why files are holding us back and what might help us replace them in some depth:

  18. Hi Alex. Your article had actually been in my Instapaper queue for a while. I just now got to reading it.. I’ll respond here because the comments on yours have gotten totally nuts. 🙂

    I enjoyed your article, but I think it misses the scenario I talk about above, “professional” computing. A family trip is one thing, but building a complex business critical web application is totally another. Besides, there are already systems that allow us to organize on those principles, like Evernote. Even so, they take effort. I have to think about the content I’m generating, decide what’s a notebook and what’s not, and think about what tags to assign.

    And even if we had some magical timeline that I could go back along via semantic processing, that would be difficult too. First off, I usually dont remember what I was doing when (much to the chagrin of people who have to parse my time sheets 🙂 Second, I’ve got so much junk going on at once that any semantic processing algorithm would think that a chat with my wife about vacation, a proposal for new work, a prototype for my current project, and an email thread about my SXSW submission were all related because they were all open at the same time. It’s just too hard to teach a computer the complexities if human meaning-making. That’s why we have folders.

    A folder says, these pieces of information are related. Simple. People get it, computers get it. I think it’s the ease of folders, and not the file metaphor, that keeps the filesystem stuck in people’s minds.

    Now that I’ve installed and begun to uses Lion, here’s what I think is going to happen. I think we’ll have files and folders for a long long time, but what will go away is *file management*. Imagine storing everything on the cloud by default. Hard drives would be relegated to backups for thte increasingly few situations in which we are offline. Versioning and saving have already gone away in Lion, the big hurdle now is collaboration. You’d create a folder, add a spreadsheet that is the project estimate, a word processing document that contains the proposal, and a diagramming document full of various charts. Then, somehow, you’d indicate who you are collaborating with, they’d be notified, and would then have access. You would never have to save multiple versions again, because you’d all be working on thet same files by default. Then the folder can be moved around as it’s own discrete object as the project lfiecycle changes.

    The benefit here is that that it remains simple. It’s not a database, it’s just a folder. And really, any other method of organizing project work like this is just a folder by another name.