Originally Posted by pik80
In a ways Apple already has a hierarchy free finder. iPhoto and iTunes is like a finder for photos and music because they look a lot like the finder except that they have dumped the hierarchies.
It's too complex to do this in a general sense though. Website URLs should make it clear that hierarchical systems cannot disappear easily. When you visit www.website.com/images/
the machine running the site translates the domain into a folder located on the server and there is an images folder inside that. If you dump the hierarchy, you will have to either rewrite 125 million websites or find a way to map a hierarchical request into a different filesystem.
The latter would be the necessary way IMO because not every other system will dump their hierarchy so there needs to be a translation.
But here's the thing, let's say that you rethink the way files are organised and go metadata only, how do you logically associate them? You group them as you do with everything in life - you put underpants in the underpants drawer and socks in the socks drawer and shirts in the closet. You don't have them strewn over the floor along with books, CDs, food, toiletries, rubbish, bedding, keys etc and then pick out what you want when you want.
You index items so that you can very quickly access a subgroup of data. This is where iTunes works well because you want to play music so you filter out anything that's not music by opening iTunes. Where this doesn't work so well is for mixed media like projects which may have audio, images, text, movies, application-specific documents etc.
Inevitably you get lots of projects so you group those too by archiving old ones or categorise them. Before you know it you have a hierarchy anyway. So you get rid of one form of hierarchy just to recreate it in a different form and break compatibility in the process while having to build converters to stay compatible with everyone else.
What Apple did with Spotlight was the best compromise - maintain hierarchy for compatibility but use metadata for advanced grouping.
The mini-Finder idea works ok like the photos and iPod apps on the iPhone whose data can be accessed by any app. The principal would be to group files by the main types: movies, images, audio, documents. Documents would be a catch-all. Application-specific files would remain app-centric. If you delete an app, there's no reason why those files couldn't go as no other app can use them and the media they referenced would exist in the mini-Finder.
You may have 100,000 images on your filesystem though so dumping them into a pile won't be very helpful but you can build tag clouds and instead of choosing images/pretty girls/short girls/ and getting one set of results, you could choose tags <images> <pretty> <girls> <short> <tall> and you can exclude mid-sized ones.
Here's where the system breaks down though: when you absolutely need to identify a file precisely, tags aren't enough to distinguish two files and if you make them that way (i.e prevent duplicate tags or tag groups), you restrict the flexibility of the metadata. Applications need to find data in a precise location all the time as do websites. Users don't have to as groupings update all the time.
You can add a unique identifier in metadata that lets an application access that file directly but that's what a hierarchical path is - it's just fixed for every app. Files on a hard drive are scattered everywhere. The filesystem is a logically grouped metadata index - the supported metadata is just limited. ZFS supposedly expanded on current systems to allow arbitrary metadata and this may solve the issue of speed when accessing arbitrary metadata.
I would assume that the people who write the filesystem drivers know the best way to work things so I have no doubt that what we think up now will be scribbled on a sheet of crumpled paper at the bottom of a wastebasket many revisions behind where they're at now.