|
|
|
|
|
|
by nobody_nothing
1849 days ago
|
|
|
> Finally, the file system paradigm fails with shared content; you can't save a Google Doc to disk because then how can your friends or coworkers update it? Just curious, does anyone know of any hybrid file formats that store information both locally and online? It seems like one solution to this problem would be a document that stores an editable copy locally and a revision hash in its metadata, then decides whether to serve up the local or cloud copy depending on whether the user is connected to the internet. Sure, this could cause conflicts between online / cloud files if someone else edits the file at the same time as you, but that's true of any cloud sync service like iOS Notes. I guess in retrospect I'm just describing Dropbox which, while it's more a container for standard files than a file format in itself, has largely the same effect. |
|
|
A very insightful comment up thread observes that filesystem-centric computing worked for as long as collaboration was very limited. Once apps needed to move beyond that to collaboration at a finer grained level it fell apart and apps started needing databases, and in particular, databases that could link data from different users together, implying a shared privacy domain.
Was this change inevitable? The long since exiled and forgotten Hans Reiser wrote about this problem a lot back in the day (he murdered his wife and obviously his ideas lost any traction at that point). His thesis predated a lot of the concerns about privacy and central control that we see today, but briefly, he argued a part of why this was happening was that filesystem technology was not good enough because it couldn't handle very small files and because POSIX had some unnecessary limitations. Due to this lack, apps were constantly forced to invent filesystem-within-a-file formats, e.g. OLE2 and OpenDoc were both centred around this concept, SQLite obviously is one too, ZIP yes, but really most file formats can be viewed as a collection of small files within a file.
The idea was, if you upgrade filesystem tech, you can radically change how apps are written.
The problem is that operating system tech on servers and desktops has been stagnant for years. Microsoft and Apple lost interest in their primary operating systems and the open source world has never really been interested in going beyond 1970s design ideas, largely because cloning and adding small elaborations to commercial designs is the way the community stays unified. Look at the mass hysteria that followed systemd, which is one of the only upgrades to the core UNIX OS design patterns in decades. Actually making changes to the core of POSIX isn't something that's going to come out of that community. It'll probably take some company that wants to innovate on the core ideas again.