I’ve spent the last year intensely exploring this new world of cloud computing, especially the multimedia side of it, and I’ve come to a startling conclusion. We are drowning in a sea of barely organized pictures, videos and music, and our computer operating systems are not doing enough to help.
Forget the desktop search tools and half-assed attempts to integrate search engines with web browsers. We are in dire need of a new OS paradigm.
Danbooru meets Windows?
We’ve recently seen new OS updates to popular PC platforms such as Apple and Ubuntu Linux, and we’re on the cusp of the Windows 7 release. However, all of these platforms have failed to make a significant leap away from the basic file-n-folder hierarchy that has driven random-access disc systems since the dark ages of the 1980s.
I say it’s high time to ditch hierarchical folders and move the desktop into the semantic tag cloud. This is not an unblazed trail. Yahoo’s Flickr site and Apple’s iTunes software have made great strides in educating the general public on how to tag and group their photos and audio.
And the tag cloud paradigm drives some of the best underground media services like 4scrape, which siphons the contents of the 4chan image board’s wallpaper channels and stores them for search and download later, and Danbooru, which appears to largely be a repository for, well, Japanese cartoon tits. Even if cel-shaded bosoms aren’t your thing, you can still appreciate the genius of the Danbooru content management system. To search engine code junkies, I’m sure it’s nothing special. But I think its simple search box and tag listing is a glimpse of the future of the computer operating system.
“But Fuzzy,” you might say, “how will we directly find our data if we can’t search through folders to find it?”
How do you find it now? Your data does not actually live in the neat little folders that your OS represents to you. Those file paths are illusory. Your data is scattered willy-nilly all over your hard drive. Your OS just keeps a table of where everything little chunk is stored, and then presents it to you as a more rational, less expedient system.
There is, of course, a few drawbacks to converting our beloved desktop paradigms to shiny new cloud computing ones. First, there’s all that tagging. Every new install or download could end up resembling the clunky Flickr process if the design is not done right. Second, we as fickle humans aren’t always sure what it is that we want. Menus and stacks of folders give us easily understood frameworks for casual, aimless browsing.
Finally, there is also the problem of motivation. With computer sales tanking and stripped-down, glorified web browsing devices like the iPhone and netbooks taking off, why would anyone invest the time and money to create an OS focused on tag managing local data storage? Why not just recycle the old girl one more time with a new dress?
Well, I say netbooks are just a cyclical fad that will diminish with the next leap in processor technology and the next big online security breach. Google is just one good court case or one good hacker away from having its Docs service severely tarnished. Data control is data power. There will always be a market for local storage and local control, even if it ebbs and flows. That market deserves a more modern OS.
Well, just ran across this, and would like to state for the record two points. First, folders are NOT illusory. Most modern filesystems are based around the concepts of inodes to store information. These inodes are typically separated into two types – file and directory nodes. Directory nodes can store pointers to more inodes, which can be, once again, directory nodes or file nodes. File nodes can just store pointers to blocks of data on the disk. Thus, the ‘folder’ paradigm is necessary for data stored on harddisk – it is a limitation of the filesystem that makes it possible to store data on a harddisk without very significant performance penalties.
Second, while I agree that this would be a neat idea, it is not something that should be addressed by an operating system – rather, it should be addressed as a new exercise in filesystem management. As a side note, this is actually somewhat implemented – run a google search for tagsistant.
Sorry for the lecture – ran across the article, and thought it was cool, so figured I’d drop in my two pence for the sake of accuracy.
tl;dr – more cartoon boobs plz.
Forgive me, but didn’t you just describe how the illusion works?
Not really. It isn’t an illusion, it is a necessary artifact of storing data on a hard disk. It is possible to circumvent it at the user level, as seen in database programs (mySql, postgresql, oracle, etc) , so implementing a file tagging interface isn’t impossible, but would require a major shift in filesystem paradigm.
The point of a filesystem is to store data on a disk (which is very slow, in comparison to accessing data in memory) in such a way as to optimize both sequential read time ( reading a single file that spans multiple memory blocks ) and random read time (reading multiple parts of various files at random places on the disk). The inode paradigm allows for good random access by splitting the data into small blocks, and simply pointing to those blocks, and good sequential access by allowing those blocks to be located physically close together on the disk.
Walking away from inodes would require you develop something similar — the two ways of storing data without it (static table and linked list) are restricted in maximum partition size and performance, respectively. I could give you a larger lecture on it, but it would probably bore you more than I already have.
Too late, I’m totally geeking out now.
So, if I’m understanding you right, you’re saying that the hierarchical paradigm of file organization extends to the actual memory block level and is necessary for fast sequential loading. I see. So, I guess what it is that I really want is an operating system that merges that tagging and organization features of Danbooru, iTunes or Adobe Bridge into a more coherent interface with the file system.
My vision: Imagine if you will being prompted to tag photo sets, music and videos as you save them. Sure it’s tedious at first, but then you go down to your task bar search and type “anime tits,” and BANG! all the cell-shaded bosoms you ever saved are hauled up in their own window. You could then save this search as its own “folder.” Multiple virtual “folders” could share the same files in different contextual groupings. Managing large collections would be far easier, and you wouldn’t need to launch a separate process to do it.
I’m pretty sure I’m not the only person who thinks this this would be amazing.
What you describe is possible, but once again isn’t the purview of the operating system and isn’t really functional as a fundamental-level filesystem. What you seem to be thinking about is a relational database (the backend for things like Danbooru) — and indeed, storage paradigms depending on relational databases are in development. I posted a reference earlier to tagsistant, which is a Linux implementation of what you describe, and Microsoft was at one point developing something called WinFS, which was something like an additional protocol on top of NTFS that added relational database capabilities.
Actually, as far as I can see, that really is the only way to implement this – as an additional layer put atop an established filesystem. There are too many functions that a filesystem specializes in to create a new one simply for tagging.
I agree that it would be very useful in certain applications, but the problem is developing a tagging system that does not add too much additional overhead, either in the user experience or in access time. In order to work correctly, it would probably require a pretty sophisticated Intelligent Agent to guess the correct properties without prompting the user — otherwise, the user would just bitch about having to enter too much information and would go back to using the standard folders paradigm.
That said, I’ve only taken one Computer Architecture class so far, so my knowledge of filesystem structure is mostly theoretical.
Hey, don’t knock your knowledge. It’s likely more informed than mine. I’m just a communications guy.
I think you have a more narrow definition of “operating system” than I do. I use “operating system” to describe everything from the file system to the GUI – the whole package that enables users to manage their programs and data. I realize that Linux users have a much different perspective with all the GUI options available to them. What happens at the disk level concerns me less than how people interface with it. And, yes, I’m saying that our GUIs need to be more like searchable databases with, as you mentioned, better AI. I’ve seen this concept creeping into Windows Vista, but it’s still clunky.
This thread (the original post and then the line of comments) is very informative, and I will link people to it whenever I fervently argue with them about the positives of tag clouds over categorization.