[johnvschmitt]

I appreciate your inputs, but IMO this use case is kind of inevitable.

Yes, it's hard, and not quite doable today. But, there are millions of people working on making the components of this solution better.

It's so attractive to carry around a computer in our pockets (our "phones") and it's really not much of a stretch to envision a near future where the thing in our pocket is our "local data/projects" that we can plug into any larger display or keyboard /IO on demand.

Before you know it, it's going to seem archaic to have local working data on any other device than the one in your pocket. And it's going to seem archaic that any nice large interface (screen or keyboard) is tied to only 1 device.

[rab_oof]

I agree and thought of that a couple years ago. The issue is that it takes a loong time to converge APIs of desktop and mobile. Plus, there are power and cooling issues. Next, stuffing in a CPU and GPU that can satisfy both requirements is basically asking for completely new cores that can do high power and low power. Cooling for 3D desktop games would require tradeoffs.

The issue is whether it's worth having a bicycle that can turn into a car and a car that can turn into a bicycle. Most of the time, the execution on these is terrible at both.

[walterbell]

> Small local storage (working cache) with the rest in the cloud.

> Before you know it, it's going to seem archaic to have local working data on any other device than the one in your pocket.

These requirements seem to contradict each other. Do you want a thin/cloud client or offline compute?

Moka5 has an offline compute offering, http://www.moka5.com/2014/10/21/comparing-moka5-livepc-vs-vm...

[rab_oof]

Does Moka5 have something new eg they're still alive?

Moka5 was basically running apps on packaged VMware hypervisors on Mac and Windows endpoints, but centrally-managed like Citrix Metaframe.

The better approach is simply wrap the app up in a read-only bundle, use an fs driver to redirect writes to a shadow vol and deploy it to endpoints without having N addl VMs to manage (so it can work ThinApp or client-side.). Even the OS minus config should really be a verifiable, read-only archive.

Disclaimer: I've had mtgs with Monica at Stanford. The similar word from both Citrix and Vmware is that that neither could develop sustainable professional relationships for an acquisition event to occur. Sad, but typical.