4 ms·

Have you considered a file system organized as a timeline that _also_ supports tagging? I find one of the key concepts that's not a first-class concept is _whe

Have you considered a file system organized as a timeline that _also_ supports tagging? I find one of the key concepts that's not a first-class concept is _when_ the file was modified. Rather than a file-and-folder physical analogy for the file system UI, I think a timeline-oriented UI could present some advantages for the way that humans actually think and work. Tags would be a helpful orthogonal organization scheme, but I don't think they work as a primary UI for navigation. This is great work, though! I love the compilation of various other works, the references, and the way you've dug into the details!

foobarbecue 9y ago
This might be a good idea if clocks were reliable.
- cesarb 9y ago
  That used to be more of a problem in the past. Nowadays, it's very common for a computer to have its clock synchronized via either NTP or a cell phone network.
  - foobarbecue 9y ago
    Sure, it's common, but I work with devices all the time that have no access to the internet. Cameras, robots, and scientific instruments. Many, like the raspberry pi, can't even keep time when they are turned off. I also go on expeditions where I have no internet access for weeks. A time - based filesystem would be useless in all of these cases.
nayuki 9y ago
I have thought about file tagging for over a decade, before setting out to write the article. But a timeline-oriented file organization only came to my awareness near the end of writing the first draft. A year has passed since I wrote the article, and the idea of timeline presentation has grown on me a lot. Especially because I use numerous data systems daily that are already time-oriented: Every chat program, email, Twitter, Facebook personal profile timeline, the "recent documents" view in major popular applications like Microsoft Office or Adobe Reader. You should find this blurb in my article helpful: > The Lifestreams Software Architecture http://www.cs.yale.edu/homes/freeman/dissertation/etf.pdf http://www.cs.yale.edu/homes/freeman/dissertation/etf.pdf (185 pages) > Comprehensively designs and tests a system for workflow and archival, based on chronological presentation plus keyword and attribute filtering. Thanks for your compliments on the thoroughness of my exposition. I did a lot of research and thinking as preparation for writing. I wanted to see how other people viewed the problem of file organization and what kind of solutions they proposed. I wanted to find weaknesses in my arguments and to avoid repeating unnecessary work, and of course I wanted to move toward the best solution.
- hvidgaard 9y ago
  If all you want is knowledge about when a piece of data was created and/or modified, then tags do that just fine.
TeMPOraL 9y ago
Not sure if it generalizes for the entire filesystem - not all files are modified due to explicit user action. Software keeps logging all the time. Many applications update their config files when closing, or during runtime. Saving a document in a program may trigger saving another 3 files elsewhere. All in all, it seems like a recipe for seeing the least relevant data first.
- jessmartin 9y ago
  You've highlighted one of the primary challenges with such a filesytem. Determining what files have been modified as a _meaningful user action by the user_ (determined by the user thinking it was meaningful) and _what changed_ would be a very hard, but incredibly valuable problem for the user. I think fundamentally the success and adoption of a timeline-oriented filesystem leans on new user experiences that have yet to be designed.

foobarbecue 9y ago

This might be a good idea if clocks were reliable.

cesarb 9y ago
That used to be more of a problem in the past. Nowadays, it's very common for a computer to have its clock synchronized via either NTP or a cell phone network.
- foobarbecue 9y ago
  Sure, it's common, but I work with devices all the time that have no access to the internet. Cameras, robots, and scientific instruments. Many, like the raspberry pi, can't even keep time when they are turned off. I also go on expeditions where I have no internet access for weeks. A time - based filesystem would be useless in all of these cases.

nayuki 9y ago

I have thought about file tagging for over a decade, before setting out to write the article. But a timeline-oriented file organization only came to my awareness near the end of writing the first draft. A year has passed since I wrote the article, and the idea of timeline presentation has grown on me a lot. Especially because I use numerous data systems daily that are already time-oriented: Every chat program, email, Twitter, Facebook personal profile timeline, the "recent documents" view in major popular applications like Microsoft Office or Adobe Reader. You should find this blurb in my article helpful: > The Lifestreams Software Architecture http://www.cs.yale.edu/homes/freeman/dissertation/etf.pdf http://www.cs.yale.edu/homes/freeman/dissertation/etf.pdf (185 pages) > Comprehensively designs and tests a system for workflow and archival, based on chronological presentation plus keyword and attribute filtering. Thanks for your compliments on the thoroughness of my exposition. I did a lot of research and thinking as preparation for writing. I wanted to see how other people viewed the problem of file organization and what kind of solutions they proposed. I wanted to find weaknesses in my arguments and to avoid repeating unnecessary work, and of course I wanted to move toward the best solution.

hvidgaard 9y ago
If all you want is knowledge about when a piece of data was created and/or modified, then tags do that just fine.

TeMPOraL 9y ago

Not sure if it generalizes for the entire filesystem - not all files are modified due to explicit user action. Software keeps logging all the time. Many applications update their config files when closing, or during runtime. Saving a document in a program may trigger saving another 3 files elsewhere. All in all, it seems like a recipe for seeing the least relevant data first.

jessmartin 9y ago
You've highlighted one of the primary challenges with such a filesytem. Determining what files have been modified as a _meaningful user action by the user_ (determined by the user thinking it was meaningful) and _what changed_ would be a very hard, but incredibly valuable problem for the user. I think fundamentally the success and adoption of a timeline-oriented filesystem leans on new user experiences that have yet to be designed.