My personal take on this is that need to make the data layer in computers transparent.
Files are currently opaque sets of bytes which need an application specific parser to interact with them. And when you have programs that can interact with a file, the only operations are read and write. Realistically, you can only have one program editing a file at once. Even having multiple programs reading a file at the same time is often a bit broken.
Instead, what if files were structured (think JSON or SQL). And instead of overwriting the file when you hit save, the application could send granular updates to the files’ content as edits happened. (Eg update Foo.bar.baz = 22, or the sql equivalent). And applications listen for changes in the data model, so another application (maybe on a different computer)! Can edit something in the file, and the changes get propagated live.
There’s all sorts of benefits to this:
- We can enable collaborative editing in all applications for free
- It’s super easy to make plugins for applications - just make another program which interacts with the data. No plug-in api needed.
- The Unix philosophy is strengthened because multiple small programs can work with the same data at once. We can split out IDEs into a lot more small tools.
- Users have lots of devices. This sort of thing would make data able to move between their devices in a way that’s transparent to the application. Like live collaborative editing Dropbox.
We need CRDTs to do it properly. But the technology behind that is more or less solved at this point.
Honestly I think the hardest thing will be getting OS people on board. I think it’s about time we shook up how the filesystem works. open/read/write basically haven’t changed in 50 years.