I'm not sure we really need the concept of a 'file system' in the coming web era.
As long as you have the ability to move your data around, eg 'export photos from flickr to facebook', adding some file system only complicates matters.
This gets me wondering: what if we made everything orthogonally persistent in the cloud and used the 200 GB of local storage as a cache? There wouldn't be any notion of a hard disk. Instead, I'd have 200GB of persistent RAM (simulated by about 8GB of dynamic RAM, backed by 200GB of MLC Flash) directly synced with the cloud.
I would never have to think about saving anything to disk. I'd just have "My Apps" and "My Data." They'd just always be there, forever. If I ever lost my laptop, so long as I have network, they'd still be there, just everything would run slower for a few days. This is a far cry, even from the Cr-48. Such a machine wouldn't even need to boot. [1] Everything would just always be there. (Apps would have to be rewritten to eliminate the notion of the disk.)
Clearly, we already have the hardware and software to support this. At this point, it's just a matter of putting it all together. This would take some serious engineering, but whoever succeeded would basically be the next Microsoft.
[1] - An easier way to accomplish this would be to maintain a small boot partition, but use the majority of the SSD for 200GB of swap. The kernel would have to be changed so that local swap is treated as a cache for pages stored in the cloud. This prevents us from having to engineer error correction directly in RAM. This would only be a stopgap, however, as eliminating the notion of a hard disk entirely would have significant benefits.
> If I ever lost my laptop, so long as I have network, they'd still
> be there, just everything would run slower for a few days.
* It would also always be there for governments and corporations to spy on.* It would also always be there to some extent even if you tried to delete the data (archived by design or by request of government/corporation).
* It would also make it that much easier for governments/corporations to secretly data-mine everyone's data for suspicion of subversive thoughts
... etc
I suppose what I really should say instead of "we've solved this" is that storing things "in the cloud" doesn't necessarily mean you give up all rights to privacy. There has to be some middle ground here.
As long as you have the ability to move your data around, eg 'export photos from flickr to facebook', adding some file system only complicates matters.
I'm not convinced that having to learn how a dozen different websites import and export data is an improvement over having a file that I can do whatever I want with. That seems like a step back to the 90s where every app had a unique format and your options for manipulating and moving data around were limited to what each app explicitly supported.
When disruptive technologies achieve the minimum required feature set to meet market demand, market competition shifts to other criteria. One case study in the book describes how cable actuated earth movers companies felt save as hydraulic actuated earth movers could never move as much earth and initially had zero commercial applications. They still can't. Yet hydraulic actuated earth movers have almost completely replaced cable driven ones, except at the market edges that require extreme performance, by establishing a foot hold beneath the existing market (ditch diggers) and innovating upwards.
Web applications are simple compared to desktop applications. They are constrained by web standards and run in sandboxed environments. The question is not whether web applications can exceed Desktop applications but whether they can meet market demand.
Software + Services (a term Microsoft created when faced with the Internet Services Disruption) is a hybrid, transitional strategy for desktop application companies that makes sense in partially connected environments. This is the model of Dropbox and every other data replication provider. Over time, perhaps with ubiquitous 4G networks, we'll see how this fares against low maintenance operating systems and zero maintenance web applications.