Linked by Thom Holwerda on Fri 29th Jun 2012 22:55 UTC
OSNews, Generic OSes "Whenever there is a conversation about the future of computing, is discussion inevitably turns to the notion of a 'File'. After all, most tablets and phones don't show the user anything that resembles a file, only Apps that contain their own content, tucked away inside their own opaque storage structure. This is wrong. Files are abstraction layers around content that are necessary for interoperability. Without the notion of a File or other similar shared content abstraction, the ability to use different applications with the same information grinds to a halt, which hampers innovation and user experience." Aside from the fact that a file manager for Android is just a click away, and aside from the fact that Android's share menu addresses many of these concerns, his point still stands: files are not an outdated, archaic concept. One of my biggest gripes with iOS is just how user-hostile the operating system it when it comes to getting stuff - whatever stuff - to and from the device.
Thread beginning with comment 524656
To view parent comment, click here.
To read all comments associated with this story, please click here.
poladark
Member since:
2009-07-15

Alfman,

This says nothing of the inability to make metadata an integrated feature of an OS built from scratch to combine hierarchical storage and metadata tagging.

I think such an OS does need to be designed from scratch. A smartphone or tablet platform that deals less with text and more media would be a good candidate.

I don't see what's so special about hierarchical file systems though. Hierarchical storage needs to be able to be extended in flexible ways too and the best way to do it is just to base the whole thing on tags. Locking yourself into an overly simplistic hierarchical system is a problem too. It's all just ways of finding pointers to data in a database anyway.

Heck even linux suffers from legacy posix stuff, look at how little support our file managers have for extended file attributes. The lack of integration totally discourages their use.

I've been very frustrated with this example as well. While that's just one operating system, i think similar things happen on all platforms.

Fundamentally though just having multiple categorised piles offers some weak organisation, just don't shuffle all the papers into one bin and then expect to make sense of it later.

I don't think we should underestimate the visual advantage aspect of sorting into piles either. If you put things in a subdirectory it's invisible. "Out of sight, out of mind." With piles you get visual cues for what's there. I've seen both new and experienced users use their icons on the desktop in this way. In some ways it's a very rich way of organizing content.

Like a pile, I ought to be able to leave a virtual desktop untouched for weeks and then come back to it again later. Also, I ought to be able to hand the whole pile over to someone else.

Actually, this sounds like an awesome idea to me.

It's a great idea. Having persistent information about piles would be very useful.

"A lot of people do have problems with tree-like file systems today. If it comes naturally, why do people evidently have a problem with it?"
Maybe MS Bob was before it's time? Today it might get a more receptive audience...Not even kidding, if the aim is to dumb down devices, why reinvent the wheel?

MS Bob was antiquated from the start. The idea that we can show some friendly familiar pictures to solve our data management needs is preposterous.

Part of the problem in this whole discussion is that some people on both sides of the debate think that it's about "dumbing down" as if people who don't use computers for a living are less smart. They aren't.

"The trick is to balance the needs of people looking to use data in different ways. Forcing everyone to use the same approach regardless of their needs is not optimal either."

Yes, but the thing is hierarchical storage is a superset of having a giant heap of things. There will be those who choose to keep everything together in one place, but so what? I don't see how the support for hierarchies has hurt them at all. If ever they're ready to start organising stuff, the hierarchy should be there for them.

There are many potential supersets of having disorganized data on a disk and i would argue that the best fit depends on the type of data and what type of work you could potentially perform on it.

Editing source code for instance is a royal pain on a standard hierarchical file system. I need to use a separate system to keep track of the changes I've made such as CVS or GIT (or a clever IDE program) because the file systems on Linux, MacOS or Windows don't support versioning. Some file systems do, but not these and I can't extend them. While i certainly can see the use of organizing my project into hierarchies i may end up with complicated hierarchy structures just to compensate for the shortcomings of the file system.

Edited 2012-07-01 11:19 UTC

Reply Parent Score: 1

Alfman Member since:
2011-01-28

"Locking yourself into an overly simplistic hierarchical system is a problem too."

But I don't see how supporting a hierarchy necessarily comes at the exclusion of alternate metadata/searching mechanisms? They seem to be complementary.

Of course in existing operating systems, metatags will always be second class citizens, which is unfortunate. But when designing a platform from scratch I think the problem is solvable.


"Part of the problem in this whole discussion is that some people on both sides of the debate think that it's about 'dumbing down' as if people who don't use computers for a living are less smart. They aren't."

Well, I wouldn't have suggested they were, but it was kind of implied when somebody claimed folders were too difficult. I feel that's exaggerated, but even if it is too steep a learning curve, there's no reason new users would *have* to learn about them before they had a need for them anyway. So it seems silly to me to argue "steep learning curves" as a justification for eliminating file folders, as some people have.


"Editing source code for instance is a royal pain on a standard hierarchical file system. I need to use a separate system to keep track of the changes I've made such as CVS or GIT...While i certainly can see the use of organizing my project into hierarchies i may end up with complicated hierarchy structures just to compensate for the shortcomings of the file system."

Who hasn't been there?

However this same problem would happen under a tagging system which lacked versioning as well. The solution is obviously to incorporate versioning.

Q: How do you work with large multi-source projects on a platform that doesn't expose any kind of a file hierarchy?

The trouble with transparent/automatic versioning is garbage collection - when to delete old copies. For source code it's not a big deal, but with photos & video...it can quickly escalate resources. It could be configured differently in each application, but that's inconsistent. What are your thoughts?

Reply Parent Score: 2

poladark Member since:
2009-07-15

However this same problem would happen under a tagging system which lacked versioning as well. The solution is obviously to incorporate versioning.

Absolutely, but part of of the problem is also that we won't know in advance what kind of user scenarios might appear in the future. What we need is something that is completely extensible in every possible way.

Q: How do you work with large multi-source projects on a platform that doesn't expose any kind of a file hierarchy?

You could complement or replace the file hierarchy by a basic project hierarchy. That may seem like splitting hairs but it might make things more easily managed. I won't deny that software is developed in hierarchies to some extent and that there may be a benefit in that.

The trouble with transparent/automatic versioning is garbage collection - when to delete old copies. For source code it's not a big deal, but with photos & video...it can quickly escalate resources. It could be configured differently in each application, but that's inconsistent. What are your thoughts?

People keep saying that this is a problem but in reality on systems like OpenVMS it very seldom is. OpenVMS doesn't have an automatic mechanism for doing a "purge" of old versions (at least that i know of) but a couple of solutions could be used. (I usually just ran a scheduled job to back up old versions of important stuff and just purged the rest.) With a more flexible tagging system however you could specify purges of old versions depending on the type of file you were dealing with.

Reply Parent Score: 1