Hacker News Reader

Comment by the_other | Hacker News Reader

As a comp-sci graduate, as someone who grew into computing through the late 80s and 90s, the files-based paradigm makes a lot of sense to me. But is it really the ideal model/UX for most people? Files are hard to manage. Most people don’t do library work, or habitually create relationship hierarchies. (I do it all the time; I love a taxonomy, but it’s work, and the general pattern for memory is association, not hierarchy).

I worry the “apps” paradigm focuses users on brands and encourages lock-in, walled gardens, and so on.

I think we need a third way.

View on HN

I'm 100% with this idea. Managing the file system in a decent way is extra effort and a skill in itself. One that is just adjacent to the concept of using a computer for work.

In fact I'd dare to say having the filesystem as first and foremost UI, is bad UX. We just grew with it, so now it feels like the most natural way to work because we've spent decades getting used to it.

This became apparent to me when trying to teach computers to some people who were like a "blank slate" so they brought a fresh perspective to everything. The concept of applications seemed about right for them. But opening the dreaded File Explorer and swim through drive letters, virtual and real folders, having to be well organized, having to worry about backups of specific folders, etc... was (and still is) a daunting task for them. And one that is just tangential to the work they wanted to do in the first place.

> The concept of applications seemed about right for them. But opening the dreaded File Explorer and swim through drive letters, virtual and real folders, having to be well organized, having to worry about backups of specific folders, etc... was (and still is) a daunting task for them.

Dude, I grew up with computers in the mid-90s and that's all a daunting task for me, too. In fact, it may be WORSE for me than for a "blank slate".

I had switched to Linux on my personal machines for a number of years and then tried to help someone with a computer issue they were having with their Windows computer- I had NO IDEA where their "Pictures" folder was, or why I kept seeing the same file in multiple places as I clicked around the file explorer. I also feel like I was confused about something to do with "Desktop", but I don't quite recall what it was. It was honestly pretty disorienting- I thought I was in a weird carnival house of mirrors.

I don't honestly know --right now-- if you can save a file to one of these virtual "Pictures" places or not. Or, if you do, where it would actually go.

To be honest, I don't know either because I avoided them since the first time they got introduced.

Those folders is precisely Microsoft acknowledging there is something wrong about the files-first approach, but as usual, in trying to make it more approachable, they ended up with an even more confusing solution, virtual folders that try to abstract the actual filesystem hierarchy with some virtual collection folders based on file types.

> Those folders is precisely Microsoft acknowledging there is something wrong about the files-first approach, but as usual, in trying to make it more approachable, they ended up with an even more confusing solution, virtual folders that try to abstract the actual filesystem hierarchy with some virtual collection folders based on file types.

I don't think those are "virtual collection folders based on file types"; they're just virtual folders (=directories) that point to two (or more?) actual folders (=directories). AFAICR you can save .JPG or .PNG files in other folders, and they won't show up in any magic "Pictures" folder; you need to save into one of the actual "Pictures" folders -- your personal one, or the shared/public/all users (WETF it's called) one -- for that to happen.

But yeah, it feels like the "solution" is at least as confusing as the problem was to begin with.

You're right, of course. I was talking from memory, it's been years since I haven't seen those folders.

> ...tried to help someone with a computer issue they were having with their Windows computer- I had NO IDEA where their "Pictures" folder was, or why I kept seeing the same file in multiple places as I clicked around the file explorer. I also feel like I was confused about something to do with "Desktop", but I don't quite recall what it was. It was honestly pretty disorienting- I thought I was in a weird carnival house of mirrors.

As I understand it, there are actually N+1 of all these categories of "semi-magical" directories on-disk -- where N is the number of different users registered on the machine -- but each user only sees (and has to care about) two of each category. And the confusing bit is that Windows tries to make it look, to each user, as if there is only one of each.

That is: There is a "Pictures" directory for each user, stored in something like C:\Users\<<TheirUserID>>\Pictures. Then there is also a shared C:\Users\All Users\Pictures directory (though see further below at [1]!) for, well, pictures that you want all users to be able to see... And finally, to top it off, there's the "virtual Pictures directory", kind of an alias (or Shortcut, in Windows lingo) to two directories at once: the users own, and the shared one. Just to make it look all un-confusing and user friendly, "here's the Pictures folder, where you keep your pictures". Road, paved, etc...

So, N+1 Pictures directories where each user interacts with their own + the shared one -- and likewise for the other "magical" per-user-plus-a-shared-one directories: Desktop, Start Menu, Music, etc.

> I don't honestly know --right now-- if you can save a file to one of these virtual "Pictures" places or not.

Oh yes, sure you can.

> Or, if you do, where it would actually go.

AFAIK, the default is the user's own directory; if they want to make it shared for all users on their machine, they have to jump through the hoop of explicitly digging out the "All Users\Pictures" directory to save to.

[1] Now, that said, this is apparently based on out-of-date information: As I looked through my Windows 10 box to check on the directory names right now, I noticed that "C:\Users\All Users\" is apparently also, in turn a "shortcut" (=alias) nowadays, pointing to somewhere in C:\ProgramData... So the above is as of now an oversimplification. But AFAICR that was actually how it worked, with the shared all-users directory being an actual physical directory on disk, in Windows NT, 2000, and I think XP. Dunno when it changed; maybe with Vista, 7, or 8.

> the general pattern for memory is association, not hierarchy

But we organize our real lives in hierarchies, too. We have our clothes in their own wardrobe maybe, socks in one drawer, underpants in another.. not the green shirts and socks with the green books and the vegetables, and the forks and knives in one box with our screwdrivers, pliers and pencils.

It's never perfect, of course, some people are messier than others, but the general idea everybody knows and sees the merit of. We navigate and create hierarchical taxonomies so much, every waking hour, that we barely acknowledge them as such. The digital version of that is not fundamentally different from that.

IMO the question shouldn't be "where did I put this, so I can have it?", but "if I had this, where would I put it?". If you know your own or an agreed upon structure for something, it's faster and less taxing, both mentally and time-wise.

Also, I don't just look for things I remember, I also remember things by browsing the shelves so to speak, and that becomes more fruitful and enjoyable with some order :)

There's a difference between the way our minds work and the way reality works. In reality a physical object can only exist in one place which naturally leads to hierarchal organization being the dominant paradigm. On the other hand our minds operate associatively.

When we started to build out computer systems we based the UX on metaphors of real-life tool equivalents. I would posture though that as we become more of a computing-literate society and as computing moves closer to becoming an extension of our minds as opposed to an extension of reality, the dominant paradigm will shift towards associate models of organization.

You've eloquently expressed something I've been feeling for a long time, but couldn't put into words myself. Bookmarks and file systems are frustrating ways to organize digital thingies.

So what would be a less frustrating way -- tagging everything with as many (or as few?) keywords as possible, and then ad-hoc searches as looking through "virtual folders"[1,2] with the tag as the "folder name"? And more systematic ones as (SQL-style?) queries of tags with "AND" and "OR" logical operators? Or something else?

___

[1]: https://news.ycombinator.com/item?id=27916648

[2]: https://news.ycombinator.com/item?id=27916467

Yes - extremely rich metadata combined with robust search features.

For example: "I want to see all..."

- excel files

- that I myself have created (not downloaded)

- anywhere on my hard drive

- tagged with a particular project name

- containing a particular word inside the file

- sorted by "last opened" date

Then ideally you'd be able to save this search/filter/view for future use.

This isn't impossible to accomplish today, but it's so painful and slow that you're basically forced to fall back to folder hierarchies for organization.

> On the other hand our minds operate associatively.

The mind doesn't just "work" one way or another (certainly in no way that can be boiled down to a single adjective) and does everything it does that one way.

And hey, that we consider a thing a physical object distinct from its environment is a distinction we made with our minds, rather than the other way around, and that I have folders like "docs/images/photos/people/$city/$person" or "docs/images/created/cheatsheets" is also not because I'm used to that it being that way in real life.

The whole point of a taxonomy is for things to belong somewhere in it, roughly or neatly. You don't have to remember the place, you just have to be the same person with the same heuristics. That enables a very fast binary tree search, so to speak, both when looking for a thing and when looking for a place to put it.

That's a positive, not a limitation or baggage from the physical world, that makes it so great and quick. That's why we used them since forever, including for purely platonic ideas even that have no resemblance to physical reality, and way before computers.

And are we really becoming that much more computer literate, or are machines that can read the books for us becoming more widespread? There is a difference between proposing a better way to store and find things, and proposing someone else do it for us (until they don't).

> When we started to build out computer systems we based the UX on metaphors of real-life tool equivalents.

They started out as tools that were sold and learned and used, now they're turning more and more into a foot in the door for companies to capture consumers however they can.

https://theconversation.com/the-internet-of-things-is-sendin...

Taking care of one's own documents and personal affects is part of what makes a person an adult, and if these documents and media get moved into the digital realm, where vast amounts of them can be ordered and re-ordered within seconds, that is all the more reason to think about what they are and how to order them. If that's hard, then that's even more reason to not put it off.

How do you tell the difference between a.) the company gets to be the middleman not being good enough to find what you are searching for, b.) that company censoring you, or c.) you having misremembered that thing existing?

I've also been using computers since the age of DOS and felt like I loved taxonomies but once I grew up I found out I hate them. A non-system/code file should better only have meaningful (and also easy to view, edit and query) extended attributes (long and short titles only targeting humans, optional description, creation and last update timestamps independent of when you have technically put it into to the file system or changed some metadata about it, content author field, tags letting you "put" it into several categories at once, type-specific properties like the type itself (so we wouldn't have to use stupid "extensions"), picture size, music bitrate etc). Creating strict single-place (symlinks don't help much in this context) short-named taxonomies for real-life non-programmer files is a huge pain and they often feel unnatural. Coming up with file names targeting both humans and traditional file system (e.g. where you can't use colons and have to keep names rather short because of full path length limits) while also optimizing them to leverage searching isn't a pleasure either.

I am sad WinFS never arrived and I am in a constant search of a good cross-platform tagging file manager.

I've been using hash tags in file names (e.g. "eat more veggies #todo.txt").

To group by tags, I use the smart folders functionality on windows/macos/googleDrive to save a search for items with the tag in the name.

Definitely not as smooth as an OS-level tagging feature, but it works in a pinch. Should be easy to port to other systems with some scripts in the future when/if this functionality becomes mainstream.

There is the problem of one document not necessarily being one file. Sometimes you have sidecar files or some meta data that is stored somewhere else in a database of sorts. So it can be messy, especially dealing with antiquated formats like dcim (easy file name conflicts), gopro leaves extra thumbnail files scattered around, wtf is even DS_STORE, etc etc. I also think that file managers are not powerful enough for searching and managing all kinds of data, the fact that we need photo management software proves it.

It's not perfect, but the macOS way of creating "bundles"; where a directory looks like a single file if it has a particular extension is 99% of the way there. The average user double-clicks and it opens in the relevant app. The power user can right-click and choose 'view contents', which opens it like any other directory.

This can also be done via compressed files with an application-specific extension. We use it at work for Python zipapps, and of course everyone probably knows .docx and .xlsx

> file managers are not powerful enough for searching and managing all kinds of data, the fact that we need photo management software proves it.

I don't want to be responsible for manually managing all my photos, copying them off cards or downloading from online services or whatever. I also want to view the photos in various ways: I'm regularly browsing them via the map view, or person tagged in an image, and so on. Hence the photo management software that handles all that for me.

But the photo management could certainly be better integrated into the OS, ie open your Photos folder and rather than seeing raw image files, it would load the photo management.

Something that is evolving for certain common types of files (images, for example) is an application-mediated storage, but where apps "share ownership" (or rather, declare awareness) of files rather than owning them exclusively.

So e.g. my camera roll, Google Photos, and Facebook comments can all give me a window onto my photos.

Those "windows" are currently inconsistent and kind of painful, but I see the germ of an approach there.

> Those "windows" are currently inconsistent and kind of painful, but I see the germ of an approach there.

There is an approach here, but I can't see it as anything other than a hostile takeover.

The "shared ownership" of application-mediated storage is not under user control. Depending on the implementation, it's either controlled by the platform and otherwise sandboxed (i.e. you can't get at the data from outside of a participating app), or - more commonly - it's controlled by mutual business agreements between app vendors. In that latter case, you still can't get at the data directly, and the app you're using can suddenly stop being able to access the data, because the vendor that provides access revoked API access to the vendor of your app.

For a random example, see e.g. Spotify revoking API access to SongShift for no good reason, and then restoring it later, possibly due to GDPR pressure by a fellow HNer[0].

Regular files are bits on your storage. They stay forever independent. Application-mediated storage is just a window into business partnerships.

[0] - https://news.ycombinator.com/item?id=24764371

I don't think your criticism is accurate. The shared Photos folder on my phone is a bit different than the database of photos Google Photos has access too. But don't get me wrong. I want to criticise the shared ownership model. I just think the model the PP intended to mention was different than the one you criticised.

The Photos folder is just a folder. Applications that have requested and been granted permission can access it using a user interface of their choice. This is the "shared ownership" location.

The Google Photos database is stored on Googles servers. Third party applications do not necessarily have access to it. It is thoroughly owned by Google.

The "shared ownership" model somewhat blurs the lines between the two of them. I think this is a considerable part of why this approach is poor: it is considerably more confusing than the old filesystem approach.

With the filesystem, you know exactly where the file lives. Most people know how to copy files and send files and can understand putting something onto a shared drive. The shared ownership notion blurs the lines with exclusive app ownership, and the idea that there are files which I can manipulate becomes obscure. It is hard even for experts to distinguish between local files and remote database entries. The distinction between copying a file and using remote tools to share access to the same file (or a copy of the file) becomes frustrating.

The way forward needs to be about respecting and empowering the user. The user wants to feel in control of their data - that is why we have the GDPR and other comparable laws. The user doesn't want to subordinate usability to data trust; they want companies to fulfil both, and since there is a hole in the market, they have turned to regulation to help plug it. Unfortunately, companies have not responded in ways that give us grounds to trust them.

A new model based on user-controlled resources and programs that can access resources (given to the program by the user) without gratuitously logging in should be the way forward. This doesn't mean a reversion to the old approach. It had its problems. But the future should look more like the past.

File-systems work very well for people who understand how a computer works. And I'm afraid your right about the "apps paradigm", it looks easy within the app but the extended usage is difficult.

Looking at Signal (the messenger app) on iOS we see the problems, they struggle creating save backup. The are not allowed to store the backup on the file-system only inside the app, which counteracts the idea of a backup. The other option is a cloud backup, most likely and third-party servers, which counteracts the idea of security and avability. Okay, but through some weird menus you could copy maybe something the a crippled third party app, let us say Files (not a file-browser!) from Apple.

PS: I still wonder how many people could have grasp how file-systems work, if Microsoft Windows didn't provided that weird desktop metapher. Which was placed within the the "own files", which incluced the "Desktop", all accompanied by the drive letters. And UNIX? This is a tree!

iOS backs up application data normally. The Signal developers blocked it. And they said it's likely they'll remove file export and import from the Android app once it has history transfer like the iOS app.

Storing an encrypted database in iCloud isn't a problem. And you can back up iOS to a computer instead.

Honestly, I think a third way would be good, but sticking with a way would be even better.

A huge, huge chunk of my last year was spent teaching dozens of teachers (long story) who were used to, and very productive with, the '90s way of using files were everything, how all these new cloud things work. The whole thing was extraordinarily confusing. Lots of things that they depended on didn't quite fit with the "let's do away with files" approach, and the fact that they couldn't do it more than compensated for any productivity gains due to easy collaborative editing and the like.

We're all computer nerds here and love to learn new things about computers but most people aren't. People have treated the sick, built the largest buildings in the world, gone to space and developed new vaccines using this clunky, inefficient, Stone Age file-based interface. I'd say it's not too shabby.

Unless a hypothetical third way enables measurable, verifiable improvements (and I mean measurable, not "we asked 10 interns to rate how easy it is to perform these actions", that's an anecdote, not a usability study), that third way would be a net loss for virtually everyone except us.

FWIW: I think the constant UX churn in our industry is just the symptom of an identity crisis. It's what happens when a company (or, as is often the case, an entire industry) is increasingly unable to produce meaningful fundamental improvements, so it's churning visible, but practically inconsequential changes in an effort to justify the money it burns through. In our bubble we all go ooh and aah over the latest UX improvements of whatever piece of software we run into again, but truth is most people just go aww fsck, how the hell do you do X now, over and over again?

I agree completely. We need a way to facilitate interoperability without requiring normal humans to know/care about bytes on a disk in some hierarchy (directory tree) where, frankly, your hierarchy WILL eventually be wrong.

Remember back in the old days, organizing your music files into a directory hierarchy- often organized by band? Then, inevitably, two bands you listen to release a song where they collaborated. Now your system is fucked- where do you put that song? It's why "tagging" things became a thing (emails, magical "Pictures" directory in your OS's file manager, etc).

There has to be some kind of extension to the idea of "tags"...

From what I’ve observed working with users, a pretty huge chunk of them (if not an outright majority) really struggle with file management. Some might not be motivated enough to organize files, and others just fundamentally do not understand how the abstract concept of hierarchical file systems work (programmers are great at dealing with these abstractions; end users generally are not)

My boomer dad used to hate folders. His home directory was completely trashed with files. Impossible to find anything, and his aversion to folders became a self reinforcing problem.

But with the lockdowns he was spending more time behind the screen and actually started to structure the mess. Now he’s even happily using a NAS to archive stuff.

I guess the principle of having apps manage files is even easier to grok than folders. However if my dad can do it on his own then it’s not like people just can’t learn folders. It’s really only a matter of how do you motivate them.

We're living in the age of search engines.

So ... just dump files in one big folder, and use a search tool to find them?

It's called Search Everything, and it's probably the best and most useful Windows utility ever written. I easily use it a dozen times a day, there's likely been days with over 100 searches.

That's more or less my approach.

It doesn't have to be "dump everything in one big folder", traditional file systems are still useful as a way to uniquely identify a file (literally a URL), but more and more, I use search to interact with data.

And that's the way we are going. For example, the Start menu in new Windows versions is best used with search. You can still do it the "old way" but for me, it is clear that it is now secondary.

We're so far into the age of search engines that you're even ending statements as queries

Reply on news.ycombinator.com