Hacker News Reader

Comment by PhasmaFelis | Hacker News Reader

I agree with most of your post, but:

> I thought dragging windows and applications across devices with a gesture would be a possibility in the not-too-distant future. Now, in 2016, this not-too-hard-to-develop feature seems almost impossible to imagine.

Moving a running process from one machine to another, seamlessly and instantly, is not at all easy on current architectures. Not that it's impossible (well, it might be impossible to do it instantly in all cases), but it would be a hell of a lot of work, even across a single platform, and have a lot of unpleasant snags to deal with. (What happens when important app/system settings are different on your desktop and laptop?)

View on HN

You assume a specific implementation. In the movie, we only see the display moving from one screen to the other. That is a problem solved since the 80s with X-Window.

More recently, all enterprise application UI are actually web UI. So that's just moving a browser window from one screen to another, so just a refined Apple "Handover".

On a more hardcore fashion, we can move a whole VM almost seamlessly. At some point that could be the case with containers too and that may not be that far fetched to move single process after all.

Exactly, it's like all already there, working seamlessly, even most of the time free. For the users who really want/need this functionality, it seems to be very cheap to just research those options and configure them already (even if they don't come preconfigured by default on a new OS installation). So maybe it's just not such a sought-after feature?

> all enterprise application UI are actually web UI

Yes, and so are most consumer apps, and "moving state" to another device is simply a question of IM'ing a link. Sure, the UI in the movie looks way cooler, but it's a movie.

That works for the barely double-digit percentage of apps which store all necessary state in the URL. In most cases, it's more like share a link, reauthenticate, get an unhelpful error page, use the navigation to get back to where you were, learn that the work you did first wasn't saved at all or that their eventual consistency means "same day" (e.g. iCloud), and hopefully they don't have some halfhearted attempt at locking which will prevent you from continuing. I suspect that if this feature ever arrives it'll be streaming rendered video like CarPlay/etc. because that's the only thing the device vendor can count on.

The point isn't that this is uncharted waters technically but that too many companies decided a good user experience isn't compatible with their desired profit margins. In some cases like security and bug fixes that might change due to regulation but that's far from certain and it's really hard to imagine that extending to broad interoperability.

Yes, for an app to successfully transfer state, it needs to be able to transfer state, that's a tautology.

But for webapps that want to be able to transfer state, the mechanism is the URL, and for those app, this works perfectly and unceremoniously well today.

The point I was responding to was your assertion that this was already true of most enterprise and consumer webapps:

> > all enterprise application UI are actually web UI > Yes, and so are most consumer apps, and "moving state" to another device is simply a question of IM'ing a link.

That's a great aspirational goal but it's simply not something which most people can assume will work – I still routinely find apps from major companies where you can't even use the back button within the same session!

X-Window from the 80s… Are you trying to give us PTSD flashbacks?

Let’s see how many of these words sound familiar: Xinerama, Zaphod, XRandR.

Not to mention network transparency. So slow, so fragile, so never ever going to support audio or USB devices or video acceleration. I see Microsoft RDP and I weep. I wonder how the Sun Ray protocol compares.

Moving processes is possible, though impractical when devices have different processors. See the enduring appeal of things like Continuum for smartphones. Moving just parts of application state, like Apple’s Continuity, tends to have vendor lock-in and third-party adoption issues. So, similar problems as what a practical IoT ecosystem faces.

Xen demonstrated this (for hypervisor-managed running processes) prior to 2005. It was one of the selling points of their virtualisation process, though it turned out that live-system migration had a few additional hiccups in it. Mostly works now.

If you're looking at back-end-mediated stuff, this is pretty much what happens when you synchronise browser sessions across devices, modulo the rate of transfer. The key is managing state intelligently.

This is also done, mostly, on server-side infrastructure, where front-end systems have little to no state on them -- individual client requests come through, state is managed usually in the datastore itself.

You're right. Maybe it's not as easy as I made it seem there. It's definitely possible, if you think of VMware's vMotion (live migration). That implies a shared hypervisor _by VMware_, of course, ... but it shows that technology-wise it's doable. When I saw vMotion for the first time, it blew my mind. Now all I can think of is how cool it would be to have that among different devices (laptop->desktop->smartphone->..).

vMotion is cool, but it works on isolated VMs, which is an easier job than migrating an application running on a desktop and integrated with the rest of the system (unless you completely isolate all applications like on the mobile OSs, but then you lose other things).

That's an idea. Wrap applications in a virtual machine and all gadgets are hypervisors. Swiiiipe.

https://criu.org/

I recall OpenMosix doing that years ago, but the tech never made it to 2.6 kernels.

https://en.wikipedia.org/wiki/OpenMosix

Reply on news.ycombinator.com