Actually, that transparency is exactly what will make or break this scheme. If know you can't make it transparent, then you shouldn't try. But the companies that succeed (probably Apple and Google) would make billions.

At some point, old data is going to have to be pushed out of that cache, and if you don't make it clear to users they're never going to know what they can and can't run.

That's why it's a >fat< cache, as opposed to a thin client. If you have a big enough cache, then it won't happen too often. It doesn't work absolutely all the time, but if it works almost all of the time, customers will be happy.

And if it doesn't work, and your company also makes the hardware, then when the customer comes to the store, you tell them they need to buy a machine with more memory. Cha-ching.

Bullshit. When stuff doesn't work people are often unhappy as hell. And a computer that just deletes old stuff, it will really piss them off.

You've just demonstrated you don't understand what's being discusses here.

I don't think so. You seem pretty attached to the idea of the behavior of CPU caches, while steadfastly refusing to extend your own metaphor to persistent storage. Odd.

And if one has a Terabyte of data, how is putting it in the cloud and accessing it at web speeds better than a hard disk at bus speeds?

With a processor cache, populating the cache predicatively is far easier due to the limited domain of alternatives, the logical structure of instructions, and trillions of cycles per core per hour available for testing alternative predictive schemes.

On the other hand, a user may ask for a rarely requested file once every ten to thirty fortnights, or never. And predicting that request would require parsing a joke told on WJMZ's morning show fourteen minutes ago.

The point is that a cache, even if it is a fat cache, will remove files eventually. Files the user may be expecting to find.

Of course, if someone offered a service where the ultimate storage backup lost data, it would be idiotic. However, this as certainly not my suggestion. Rather, you selectively applied the functionality of hardware caches to the "Fat Cache" idea, while fabricating ridiculous semantics that go against your own hardware analogy. (Where the cloud would be the hard drive.)

This, as the comment to which I responded proposes, creates an issue because the state of the device is unknown to the user and any particular piece of data or particular program may or may not be available at any time.

Perhaps I am just dense, but I do not see how this is advantageous compared to local storage. In other words, just because you feed it a case of Krispy Creames, a thin client is still a thin client.

For a given amount of storage, a Fat Cache may beat a thin client. But it's turtles all the way down, if the amount digital information people store keeps increasing, i.e. the first box of ten 720k floppies lasted me the better part of a year.

Reorient your thinking. The central thing is no longer the "device" it's the user's account and sessions in the cloud. The user's data never goes away, it's just that access to it is sometimes good and sometimes not so good. Mostly, it's good.

Nowadays, if a user's mobile broadband is broken, they just think of going somewhere with better reception. In a few years, when a seldom used piece of data has fallen out of the cache, they'll just think, oh well, I'll get back to it in a bit. (Or, if it's really important, they can select "Keep Available" so it'll be there next time.)

However, if you go back to a time before mobile broadband was commonplace, and you tried to introduce someone to it in a bad reception zone, they would've been very "meh." That seems to be where you are.

A strong case can be made that having access to mobile broadband intermittently is an improvement for the user over the way things worked before.

I haven't seen anyone even attempt to make a case that intermittent access to one's data and apps is an improvement over the current systems.

In other words, why would anyone willingly switch to this model? And among the unwilling, how does it better meet the needs of business than a thin client or other conventional solutions?

Having to select "Keep Available" requires the user to accurately predict the future value of access, cache space for it, and adds another step to the process of saving stuff for later use.

It's not availability for next time that's the issue. It's availability now, when I need it that matters.

A better analogy is a allen wrench sitting in a toolbox out in the garage. It's not at hand when the disposal is stuck, but access is predictable. What you are proposing is that someone comes by once a week and swaps out the tools in the toolbox based on which one's you have been using most recently.

Again, where is the advantage to the user?