If Google can produce a Linux based operating system that addresses the needs of casual users... [they can] simply eat everyone elses lunch.
He also addresses the elephant in the room that the major Linux UI pushes that consciously and grandly targeted those goals were grand steps backwards. That the things the UI prognosticators think users want, like abstraction and simplification, alienated real Linux users like Linus. That normal users missed features the UI cognoscenti banished as "too confusing," such as "easy mouse configurability for things like how to launch applications."
When you talk about abstraction, metaphors, making the filesystem invisible to users, and (most telling of all) appealing to "casual users" it sounds like more of the same talk that we've had for many years now. The more grand and self-conscious the community gets about user interfaces, and the more they target non-techie users, the less useful the desktop environments become for their actual users. That missing configurability that Linus laments, which was sacrificed for the sake of "casual" users -- did we really attract a bunch of casual users? Did we eat anyone's lunch, take market share from OSX or Windows? Or did we just make our desktop environments less usable for the people who actually use them?
Linus says GNOME was "useless" without that configurability, and that this new device might be usable as a laptop if it had a terminal and a development environment. That doesn't sound like a call for Linux environments to appeal to "casual" users. I don't think Linus is asking for even more simplifications to get in the way of his usage, like abstracting away the filesystem.
The elephant in the room is that Linux desktop environments did a better job of giving Linux users what they wanted when that's all they tried to do. Trying to give us what we didn't know we wanted has been a failure. Targeting non-techie users has gained us nothing that we didn't already want for ourselves. It's time for the community to get off its high horse about UIs and go back to what it's good at, which is humbly (and very successfully) catering to its own needs. And nobody should feel ashamed of creating or using an operating system that never eats Apple's or Microsoft's lunch (isn't that such a ten years ago obsession?)
(P.S. You know what's good at abstracting away the filesystem when and how it's appropriate? Applications.)
I agree that applications are good at abstracting away the filesystem which is why I would argue for a lighter weight OS (which I presume chrome OS is).
Now whenever there is an innovation, it is reviled and everyone asks for the interface which looked roughly like Windows 95.
So I guess they listened to the wrong messages.
Targeting "casual" users has resulted in some valuable work such as better stability and completeness in many programs, simpler configuration interfaces for many things (such as wireless configuration GUIs that were better than Windows'), and better support for multimedia. However (leaving aside the fact that those were all things that techie users were begging for anyway) there was a downside. Customizability and configuration options disappeared. People made assumptions about "normal" users that implicitly labeled a large chunk of the Linux community abnormal. Case in point: if you configured your wireless connection using NetworkManager and then switched to a different desktop environment, your wireless connection might not work anymore. That might sound like a bug, but it wasn't. It wasn't designed to work that way. How could that be considered remotely acceptable? Because normal people don't use alternative desktop environments. (Thankfully, I've since read that NetworkManager accepts plugins that write to the correct system configuration files, though I don't know if distros test and install them.)
And where's the payoff? What was the payoff supposed to be, anyway? Back in the nineties and early 2000s many people assumed Linux had to make it on the mainstream desktop in order to be successful, but I think we've put that that misconception behind us. Yet people still hold up the "casual user" as the gold standard that Linux is supposed to cater to, as if it were a moral imperative. Few Linux users fit that stereotype. It isn't that we're concretely knowledgeable about Linux. If you put ten randomly chosen Linux users in a room, chances are that for every component of a running Linux system there will be somebody in the room who is utterly unfamiliar with how it works or how to configure it. Obviously we should cater to ignorance, if only so we have the luxury of remaining ignorant ourselves. But there is a general savoir faire with computers that it is acceptable to assume. By savoir faire I mean whatever factor it is that explains why I am the person in my family who always gets a call on the phone when someone needs help with Windows 7. Even though I've never developed on Windows, haven't used Windows for anything more than e-mail, web browsing, and Office in ten years, even though I haven't used Windows personally in six months, even though I've never used Windows 7 or Vista at all. They still ask me for help with Windows 7, and usually I can help them.
All I'm saying is that it's acceptable for Linux UI designers to assume that the people using their UIs are likely to resemble current and past Linux users. Take a break from innovating for projected, postulated, hoped-for users. Innovate for the ones we already have instead.
In regards to remote resources, currently there are various FUSE plugins which can present remote resources in the filesystem (e.g. flickrfs). These are of course issues with the abstraction (how do we represent tags in a hierarchy?), but the use of these plugins is mostly transparent to the user (ignoring performance). It's certainly possible to build a layer atop of these that would integrate better with the system (e.g. add a Facebook selection to the open dialog).
You also seem to want to conflate applications with the data they manage. For a large number of users this is fine (iTunes is already their music, IE is the internet, etc), but for many others this is problematic (how do you manage moving data between these applications without a common intermediary. consider scanning an image, touching it up, then including it in a document - where do the original and altered versions live, how are they accessed by applications?).
Conflating data with applications is an interesting issue. On the one hand I feel that there may be a move towards this, for example it would be advantageous to (say) Apple for every photo you took on your iphone to be tied into the iCloud ecosystem never to escape.
On the other hand , storage is such a commodity that there's no reason that you couldn't hold your data where you want but hook into the functionality of a particular application to organise it.
For example maybe you use gmail to look through your archives of old emails that you had from a previous provider that are stored on your HDD. Spotify already does something like this with it's local music functionality.
He's really just addressing the elephant in the room by saying that as of yet the whole open source desktop mess has failed to produce anything that looks like a compelling , easy to use desktop.
If Google can produce a Linux based operating system that addresses the needs of casual users and power users/technophiles (like Linus) and get it on enough hardware (i.e not just netbooks but "serious" desktops as well) then they can (combined with android) simply eat everyone elses lunch.
What is not so important necessarily is whether or not local storage or local apps will "go away" but that Google understands that a modern OS needs to treat "the cloud", i.e webapps , online storage , social networks as first class citizens inside the OS. This is a metaphor that is missing (in a really intuitive form) mostly from even new OSes like Windows 8.
Hard disks are not dead, but I think the idea of having C:\Program Files\.. is. Even as a technical user myself the directory hierarchy is mostly just an abstraction that gets in my way. My work (i.e source code) is in git, my games are in Steam, my email is at google , my music on Spotify and all the "other stuff" is in dropbox.
What I need is an OS that ties all of these things together seamless regardless of whether the bytes themselves reside on the disk inside my computer or on some website somewhere but can provide sufficient tools when needed to fix leaky abstractions.
Let's suppose you could buy a new workstation soon that ran a nice clean Chrome OS desktop but had sufficient memory and hardware level virtualization that simply clipping in a "proper" Linux or all the MS libraries such as DirectX that you needed to run Windows apps was a trivial activity that could be abstracted away from the end user if required. Would you not be curious to buy one?