OS X Snow Leopard Details 489
JD-1027 writes in to kick off a discussion of OS X Snow Leopard. Apple's stated goal: "Taking a break from adding new features, Snow Leopard — scheduled to ship in about a year — builds on Leopard's enormous innovations by delivering a new generation of core software technologies that will streamline Mac OS X, enhance its performance, and set new standards for quality." The technologies: Grand Central to get better use of multiple processors and multicore chips, OpenCL to tap the power of the GPU, 64 bit so we can finally have our 16 TB of RAM, QuickTime X for optimized modern codec performance, and built in Exchange support in iCal, Address Book, and Apple Mail that most likely will help get Macs into corporate environments. We've previously discussed ZFS in the server version of Snow Leopard."
How about NTFS read-write? (Score:2, Interesting)
Re:How about NTFS read-write? (Score:4, Interesting)
Who is in charge of codenames at Apple? (Score:2, Interesting)
It's almost as if Apple is trying to prove that FOSS projects don't have a monopoly on horrible names.
Yeah... "Leopard"... "Snow Leopard"... that's not gonna cause any confusion, right?
Re:One wonders... (Score:4, Interesting)
Or so I've heard.
Re:One wonders... (Score:4, Interesting)
Hahaha. IRIX back then was so buggy I'm amazed that the user experience was as good as it was.
Strategy? (Score:3, Interesting)
The idea would be to stop Redmond from using Apple as the R&D labs, as many suspect winds up being the case ("Start your photocopiers"), and deny MS even the opportunity to borrow for Windows 7.
The more I think about it though, the more obstacles I see to this. But it would be sweeeeet...
Re:One wonders... (Score:5, Interesting)
That said, Apple has done amazing things with every release of OSX and I look forward to Snow Leopard as much as every other release. I simply didn't read it as something that anyone should treat as a Really Big Deal, even to the point that Jobs barely mentioned it in the keynote, unlike Leopard that got its coming out party twice.
Therefore, if a 10.6 box just appeared in the Apple stores, but didn't get much mention, it would probably be missed by most. Sure it would be pre-installed on new machines, but where would be the hype to get everyone on it as quickly as possible? This is why I was thinking about the 10.0->10.1 upgrade; if this is the first Intel-only release, how would they sell a version that offers no new features, and is unavailable to everyone who doesn't haven an Intel machine? I, personally, wouldn't want to be in the marketing department trying to sell 10.6; if they just make it available as a download, they might ultimately save a lot of $$$ that would have been spent trying to market it, then explain it, correct the marketing, etc.
Re:One wonders... (Score:5, Interesting)
Re:End of PowerPC Support? (Score:3, Interesting)
Re:One wonders... (Score:5, Interesting)
Sure, the boosts in efficiency and stability will be welcome, but I for one am very excited about full Exchange support in iCal and Address Book. Heck, the Exchange support in Mail is a bit spotty as well, so touching that up would be great as well.
But what would really be great (and very much in line with the whole "embracing enterprise" thing) would be native support for Cisco IPsec VPN connections. As it stands, you have to use Cisco's own clunky client; if you could use the built-in client you could connect via a menubar icon. (Shimo does this pretty nicely, but it just became crippleware.)
It seems like an obvious addition, given the iPhone 2.0 OS is supposed to have it. Anyone know if it's on the docket for Snow Leopard?
Re:Grand Central -- details, anyone? (Score:1, Interesting)
The closest thing that I got to additional information about Grand Central is some article specifically stating that it's a scheduler. It might be that's the assumption of the author, as they may be unfamiliar with Apple's affinity for adding or changing APIs.
This is a rather on-point announcement to me, as I'm writing software for OS X and have been doing "back to basics" readings which include Butenhof's POSIX thread book. I'd hate to settle on a particular approach to multithreading in my software and then find out Grand Central undercuts that decision.
Re:One wonders... (Score:4, Interesting)
But what would really be great (and very much in line with the whole "embracing enterprise" thing) would be native support for Cisco IPsec VPN connections. As it stands, you have to use Cisco's own clunky client; if you could use the built-in client you could connect via a menubar icon. (Shimo does this pretty nicely, but it just became crippleware.)
Re:One wonders... (Score:3, Interesting)
Can't read from a Windows partition (Score:3, Interesting)
It's actually really nice to have a Mac around when pulling files from a possibly infected NTFS drive. You're not going to pick up anything that will infect your machine, and you can pick and choose through the files you want at your leisure after reimaging your Windows box.
Re:Jubeezus Folks get a grip (Score:5, Interesting)
And no, I'm not a switcheur nor a noob. I've used/owned Macs since System 7, I've been using Linux for 8 years now, and I started with DOS 5 on an 80286, and ran every Windows and Mac version from then to current.
XFS is a fast format, ext3 takes a few minutes depending on the size of the partition, and NTFS is a few seconds in quick mode. Quick format has been there for quite a while (even DOS) and without it I always assumed format was zeroing the partition, which is slow of course.
Re:That's why I'm going to buy it. (Score:5, Interesting)
Apple didn't just add bling - they made the operating system more stable and fixed a lot of bugs. So, be fair - we didn't pay $120 for a new dock.
Full list of new features in Leopard: http://www.apple.com/macosx/features/300.html [apple.com]
Streamline It Simple Again (Score:2, Interesting)
Maybe it's Apple competing with Windows that's somehow gravitationally moved the Mac experience closer to the Windows one, even as Windows has sucked ever closer to Apple's innovations. But it used to be easy for a beginner (or just an "uninformed expert" like me) to "just do it" with a Mac, with a much shallower, barely noticeable learning curve.
What we need is a GUI revolution. The iPhone offers one, with its multitouch innovations. As does Nintendo's Wii, with its unconventional new controllers. The Mac, like everyone else, is still stuck in a transitional metaphor to an office/desktop physical environment that's now been totally replaced by its simulation on the Mac. That metaphor doesn't really help people use "documents" and "tools" from past experience with the real things, liberating us from them. It's now a trap that constrains us to only the small set of characteristics that both the real and the virtual versions share in common.
I hope Apple will spend the next year "streamlining" MacOS into something more simple and immediately usable, the way Apple has delivered in the past. Because usually Windows, Linux and everyone else follows and improves likewise. But if it doesn't, then I hope that inspires people to do something really new that's really simple, yet delivering the vast power of all our new devices. Because those people will inevitably be the ones to drag everyone else along into the new, simpler paradigm. And probably get rich along the way.
Re:Jubeezus Folks get a grip (Score:2, Interesting)
So: it would be really nice if Apple could get *any* file system working other than HFS+ working. There is practically no chance of this until they abandon HFS+ completely. If (parts of) OS X weren't screwed up so bad that they depend on HFS+, then NTFS, ZFS, UFS, any FS(!) support would be easy.
OS X makes me hate Unix. In the sense that it makes me long for the systems i learned on: Solaris, FreeBSD, Linux, A/IX and yes: A/UX. "Thinking Different" is bad in this context. We have man pages and RFCs and POSIX so people are all thinking the same. I'm sick of Apple making up new solutions to problems the *nix community solved years ago.
Re:Jubeezus Folks get a grip (Score:4, Interesting)
NTFS is an integral feature in win xp, which is an upgrade for most informed vista users.
As such ntfs is the future of the pc market.
The Zune, however, is to music players what the edsel was to automobiles.
When the comp usa's went belly up in my city and had their closeout sales, even the shelving units went before the piles of zunes left sitting in the middle of the empty salesfloors (I wish I had photos, it's not an exaggeration).
Re:Jubeezus Folks get a grip (Score:2, Interesting)
However, I wasn't trying to bash Apple (if you'll read above, I've been using Macs for a long time). I merely was responding to a comment that there is a WoW client for Mac, and yet I use a Windows laptop for it.
The GMA 950 is optimized for video playback, not 3d acceleration. To the Macbook's credit, it is quieter and cooler than the Windows machine.
I bought the Macbook for its size, and at the time I played no games at all. A 13.3" mac with a real video card would have me sold, but they haven't released one yet.
Ironically, the best system for playing WoW would probably be the rebuilt server, with a 2.4GHz conroe, a real Nvidia card with 256MB RAM and of course several 7200RPM drives. The Windows WoW client also runs on Linux I hear.
Re:Jubeezus Folks get a grip (Score:5, Interesting)
And can you point to any standard??
Last time I was checking, only few applications were using Direct X 10. For any kind of productivity more or less everybody uses bunch of wrappers or some commercial library.
The whole point here that there is no standard. And M$ forces everybody to kiss PR ass of Direct X, though literally nobody directly uses it, except for hardware manufacturers (nVidia and ATI). Some proprietary half-arsed spec in .DOCX peppered with implementation details from actual version of Direct X (even is such document exists) hardly qualifies as standard to me.
On other side, Kronos group is something. They are slow on up-take, but generally deliver usable standards industry needs. They are vendor neutral what is also important.
Do not expect anything in particular from OpenCL. I'm pretty sure that it would try to appeal to wider audience - consequently it would be pretty dumb down. But still it would let any developer to access GPU chip. Knowing how Apple does things, with couple of extra objects in one's program and few extra checks on whether you can use GPU, many tasks would get a decent performance boost. It wouldn't be high-end nor exclusive - it would be something for wider audience.
Re:Streamline It Simple Again (Score:3, Interesting)
There are small steps that can be taken to make certain tasks easier. On my mac LC maybe 12 or so years ago, I had a broad and complex series of folders that I used to organize my various personal text documents so that I could find what I was looking for with relative ease. Now, on my Macbook I have only maybe two or three folders that I keep those sorts of documents in. I have significantly more of those documents now, but the processing power of my current computer combined with software like spotlight makes finding something very simple indeed. Activating spotlight isn't an intuitive process, but it's very simple and easy to learn. And that's also an ability that doesn't really have any metaphor in the physical world. Being able to instantly find a particular string of text from a pile of thousands of pages is pretty awesome.
There might be some sort of amazingly innovative way to simplify all of that without dumbing it down or reducing capabilities, but I kind of doubt it, short of some sort of direct neural interface that lets me control the computer with my thoughts. As long as I'm feeing the computer data with my hands and the computer is sending me data via my eyes(and occasionally ears), I cannot imagine effectively using a computer without someway of quickly inputting text (a keyboard) and without a pointing device (a mouse/trackpad/touchscreen/etc). Once you've got that, icons are obvious, as is the need to contain tasks (windows). Add in the various ways of inputting and displaying text/images, and you've got your basic GUI.
The thing is, I'm not sure that "intuitive" is even the best thing to strive for. Humans are generally pretty good at learning, so requiring someone to grasp some basic actions in order to interact with a computer isn't that big of a deal. More important is that the system be internally consistent, so that those basic actions can become automatic and thoughtless.
The Wii is only a partially useful example, because while it does allow for new ways of interacting with a video game, there are also sacrifices. You might think an Xbox360 controller has a ridiculous number of buttons on it, but there are games that use all of those buttons. You might be able to find some way to map all those buttons onto the control methods provided to you by the Wii-mote, but there's no guarantee that that new control scheme will be superior.
Multi-touch on the iPhone is really just an evolution of our current gui, and one that as much as anything only makes up for the lack of a keyboard/mouse, rather than being some sort of evolution beyond them.
Re:Jubeezus Folks get a grip (Score:3, Interesting)
For some reason, Apple feels forced to use Intel in everything even in Graphics which Intel has no clue about. I wonder if there is some kind of agreement involved considering they are basically ignoring 64bit/multi core/SMP G5 userbase in 10.6. Hopefully it is false rumour.
Apple should have nothing to do with "integrated graphics", "integrated" anything. They aren't some no name Taiwan company, they aren't in cheap laptop market.
Re:One wonders... (Score:5, Interesting)
I noticed that 10.5 seemingly has more stability problems than previous versions of OSX since 10.1. Is it unfair?
I don't know if it is unfair but it sure is incorrect. Did you use the Finder from 10.0 through 10.3? It got slightly more stable with 10.4 but it was only 10.5 that a network outage didn't take down most of the Finder.
OSX wasn't even usable until 10.2 and not really preferable until 10.3. (IMO)
Now I will say that 10.5.2 was the first point update that I thought caused tons of problems. I ended up having to reinstall Leopard from scratch and then apply the updates. I haven't had to do that since the old XP SP1 days.
Re:Streamline It Simple Again (Score:3, Interesting)
Note, I am by no means an Apple/Mac fan. "User" at best; I consider my mac like the rest of my music gear: as an appliance. My primary OS is Linux.
I call BS. The changes to OSX haven't been that big in past years---in fact, I've found them to be rather minimal tweaks at best. I recently moved from early 10.3 (on an original mac mini) to the latest 10.5 (on an imac). The changes I notice? The dock looks slightly different and has stacks now. Multiple desktops are builtin. There are probably a few other minor things.
Comparing this to "the windows experience" (which I sometimes must deal with), there are far more things I notice: in OSX, it's obvious and easy to find how to do stuff, especially configuring the system. Everything is in one place. I don't have to hunt through 3 or more different control panels and hope to happen upon the dialog that does what I want. Everything pretty much just works in OSX, and the complexities aren't hidden, they're simply organized in a very accessible fashion.
I call more BS. There is little "innovative" or "unconventional" in either of these examples. The iPhone is, for the most part, single-touch-oriented with a conventional touchscreen interface. It has pretty graphics and scaling, and there are a few multitouch things you can do (that often work poorly). There is a bit of gesture recognition, which is hardly new. The Wii, likewise, has nothing particularly innovative in its UI. It's almost entirely mouse-like point and click in its interface components and the better games. The few games where it manages to use motion sensing in an "intuitive" fashion, it's anti-innovation: natural mimicry is what the GUI has been about for decades.
OK, so you're bored and you want a flashy new "streamlined" toy that doesn't work like anything else, but somehow magically delivers usability. That's not how it works. There is no magic. If you want a revolutionary, innovative, streamlined UI, go try out blender [blender.org]. It's quite unlike anything else, and once you learn it, it's extremely fast and easy to get things done. And it's got a hell of a learning curve to get there.
If you want something you can use without a lot of effort, it's going to be conventional. Maybe candy-coated so you don't notice so much, but it's going to be as conventional as anything. People are used to mice, clicking on icons and buttons, and menus. The problem is, once people are used to something, getting them to change or accept something new is difficult to impossible.
For what you want, Apple is doing the right thing: releasing new, more polished versions of its OS, with enough new shinies to keep your attention. Unfortunately they haven't talked about them yet, and you're starting to wander, but I'm sure we'll hear about something soon.
Re:That's why I'm going to buy it. (Score:3, Interesting)
I'm skeptical that Grand Central will help as many applications as some suggest. But it's anything but vaporware. It's on the 10.6 developer edition given out at WWDC and there were sessions on it.
Not being behind the NDA I have no clue exactly how Grand Central functions or what kinds of processes it'll improve. But those who have seen it seem reasonably impressed. Even if mum about the details.
Re:Streamline It Simple Again (Score:3, Interesting)
That's why the PC environment I want [slashdot.org] has all UI widgets standardized, across the OS, with no individual "application spaces" in which different UIs are available for the same operations, on the same kinds of data. Throwing out the "application" layer in the UI altogether, so the OS directly supports the operations on the data, the data can interoperate with other data in the same UI context. An OS that new functions plug into, like apps, but which functions present their own UIs only right where their relevant data is exposed in the OS UI.
So if you see something that looks familiar, it will be, both in its ways of consuming its content with your senses, and operating on the content with the UIs coming with the content.
Re:Kick the Finder. (Score:3, Interesting)
Tiger for Intel was "Snow Tiger" (Score:3, Interesting)
With Tiger they said "come get Tiger" and with Tiger for Intel they said "come get Intel". With Leopard they're selling Leopard and with Snow Leopard they'll sell a larger number of processors and more memory than Leopard can support. One release they sell the software then one release they sell the hardware. They don't have to worry if Snow Leopard in-a-box doesn't sell all that well, because Snow Leopard in-a-Mac will sell really well, it'll be designed to drive new Mac sales. They already mentioned ungodly amounts of RAM in their first PR about Snow Leopard.
Re:Kick the Finder. (Score:2, Interesting)
Why do they need so much software? Why don't they have it all in [Computer name]/Applications or their home folder?
Users can't find a particular window? Expose and Spaces. The dock is too cluttered? You can shove stuff into a folder or get rid of icons you don't use. Drag the Applications folder there, instant Start Menu.
I mean, good grief, Apple gives them a home directory complete with "Documents" and "Downloads", in addition to the desktop which they can clutter just as much as Windows/*nix users can. Why is it Apple's fault that your users can't organize their stuff?
Re:One wonders... (Score:3, Interesting)
Now, they've got a fairly substantial "additional" product line - ipodTV, airport, etc. which all need support. And...
Now they've got other considerations for OS X: they're running all their products (aside from the ipod touch) on the OS X core technologies. That requires reduction in size, additional efficiency, and so on and so forth. Fixing, and making OS X better overall (sans additional features) pays dividends, because it spans all of their products. Doing tihs will make future feature additions to the platform more tenable, as well as make the platform a tenable long-term project.
There's really very little "small" stuff they can do to OS X, I think. I've only used it a little, and I'm no fan boy, but there are substantial benefits in almost every area for normal desktop use. In order to make OS X viable (and superior) in the other arenas, they've got to fix what ails - in this case, some of the underlying infrastructure.
As a system and database administrator on their over-priced platform, these changes excite me a lot more than 10.5 did, because they open up more fully the possibility of actually having a system which has a full suite of integrated sysadmin tools that can be leveraged for efficient db driving.
Re:One wonders... (Score:3, Interesting)
Case(s) in point:
Windows 98 -> 2000. People jumped on that ship pretty quickly, even though 2000 offered diminished graphical performance. The only people who stayed with 98 were people with low-end hardware, people who'd been bit by upgrading MS software too soon in the past, and by those who were hardcore gamers and didn't mind the stability for an extra 5fps.
Enlightenment window manager (.16 or
I'm curious: what new features do people actually want in OS X which are obtainable? I'm aware of the stupid things like "transparent Windows emulation" or "run Windows without any performance hit". Those are kind of stupid. In my mind, 95%+ of the features which can and be delivered in an OS and are significant to the vast majority's user experience are present already. The only things significantly lacking are not the wiz-bang user features, but the nitty-gritty which is important to the more technically inclined - the kind of things that linux users bicker about ("2.4 had better performance than 2.6 for aquatic parallel computing!", "the new i8x0 Xorg drivers are borked!" or what have you). That's important, because a lot of nitty-gritty stuff is missing...
However... If there was one feature I'd ask for in the "presentation" parts of OS X, it'd be the ability to do window vs. task/app based window management. Not something hardwired, just a damn option. Unfortunately, features like this will never come about, because they break the Apple UI Use Guidelines, or some such nonsense...
Re:Kick the Finder. (Score:3, Interesting)
Finder is a complete mess. It appears to be a ported application from an OS from 1978, or something equally antiquated and quaint: being certain of what you're doing (copying? moving?), and in which directory you're doing it (damn it, why did it put it at the filesystem root, AGAIN) are just the start of what makes finder frustrating. Why does the 'maximize'/+ button not do as it does in most other applications? Why is there no "cut" option? Why do I not have an "address" bar, particularly now that we've got full and proper UNIX file paths? Why do Finder windows not stack/organize themselves in such a fashion as to make having more than (say) 3 open at any one time frustrating and confusing?
Honest to god, I've resorted to just using iTerm with multiple tabs for all file management (short of multiple selections). It's quicker, easier, and less confusing, as I never have to wonder "where am I?" I don't want to be forced to feel that way, and I don't intend to feel that way at all until I'm well past my 50s.
The task management - application switching instead of app switching, and no way to change it - is equally irritating. This includes the parent-child window lock-out situation. It results in all kinds of irritating context problems, where you're trying to perform work, but are unable to do so without repeatedly closing and opening a specific context window, as you're unable to switch and/or remember the content of said window between switches in completion (I end up printing shit out and referencing it that way, sadly, more often than I'd like). That isn't reasonable, at all, and it's like no other operating system or windowing system I've used.
Finally, combining those two problems seems to result in an inefficient use of screen real estate. There's a good reason mac workstations have large displays: they need them to be effective at multitasking. I don't imagine that was much of a case when the Mac was just a graphic workstation or something like that (when macOS multitasking sucked/didn't really exist, and there weren't many apps/users), but now, it's kind of ridiculous. I don't want to have to buy a larger screen just to get basic work done because fancy widgets are taking up too much space; I want a bigger screen because I need more space. Compared to pretty much other UI, OS X definitely seems to need more space by default. (Sorry, I can't quantify it better than that.)
It wouldn't be such an issue if focus context switched properly when going from "Space" to "Space", but doesn't, so that potential way of managing things is kinda of another irritant that's got to be worked around...
Re:Streamline It Simple Again (Score:3, Interesting)
I thought the Monad shell that Microsoft dropped from Vista sounded like a step in the right direction. Although it still operates as components running in a shell it formalises how to encoded the types that each tool can operate over.
Implementation of your idea would require extending this type information beyond stdin and stdout to interfaces for using windows / widgets as input/output.
It's a persuasive idea and I've liked it each time I've heard it. From Amiga back when they added the datatypes library to 3.0, from Microsoft when they first started describing Cairo (and again with Vista and again next time).
But still, one day it will happen...
Maybe they can make the file system layout easier. (Score:2, Interesting)
I'm revealing my age here, when I say that my first unix based OS was NeXTstep.
One of the things I liked about Next was that an application kept to itself: Wherever you installed it, everything was in a directory "ApplicationName.app" This made uninstalling easy. It also meant that installing an application on a network file system made it available to all NextStations on the local network. (In some cases a 'dwrite global applicationname value' was needed for licensing for individual machines.)
Apple has not insisted on this. While many applications will work this way, now files are also stuffed into various Library directories. Uninstalling applications manually is no longer trivial.
Furthermore, some applications insist on writing to their own program directory.
I wish that apple and other OS's would implement a new security model regarding file spaces.
1. There are three file spaces: OS, Application, and user. Each can be divided.
2. The OS space consists of the distro along with applications from the distro vendor. For Windows the OS would include WordPad, but not Office (sold separately) For Mac it would include Mail, but not Aperture. For linux it would include
2a. The OS space has at least the following three subsections:
3. User space.
By default user space has a directory for each user, with access restricted to and controlled by that user. This is pretty much the way things are now.
3a. User space/group space. Methods for collaborating and sharing documents.
4. Application space.
app space is done on 1 top level directory per vendor. Acrobat reader goes in
The key here is that the adobe installer does not have write privileges outside of the
Just as user smith can't write to user jone's files, nor should Adobe be able to write to microsoft files.
This implies that some program equivalent to Next's 'buildservices' needs to periodically run to pick out what programs provided services for other programs.
5. In a general setup, no user should be able to execute a file in a directory they have write access to. Some mechanism for installations, and for developers needs to be made, but as a general rule this would go a long way to intercept malware. For users (as opposed to developers) having executable code in their directories is not a benefit.
Re:Streamline It Simple Again (Score:3, Interesting)
The main difference between the Unix model and mine (at least how it appears to the user) is that my model doesn't have applications. It's got libraries with UI code, that work on specific datatypes (including multitypes). The OS presents the data within its datatype's UI, with its datatype's operations exposed, in display frames with OS-wide operations on all datatypes. The OS could be Linux, running a single app, the UI, which exposes those widgets, and those presenting OS features. Like pipes: every process has exposeable widgets for its filehandles, which are used to redirect its STDIO, among other processes.
The result would start to look a lot like what the original Mac started out looking like: all data would have the same UI that all the data of that type did, like the Mac started with just the MacWrite app frame for all text data, MacPaint look & feel for all graphics, etc. Knowing how to do an operation on a datatype would work in any context with that data. And the original operations for one datatype would serve as templates to be "overloaded" (C++ style) in equivalent (or analogous) operations on other datatypes. So using any new datatypes/UI is leveraged as much as possible off existing skills with existing datatypes and their UIs.
I note that what broke that tradition was Excel, which was Microsoft's first Mac program (well before a Windows version). Excel was such a great program (and still is, mostly) that it was excused from smashing Apple's otherwise strictly enforced GUI guidelines. I take that lesson in thinking of the basic template for all "data frames" being a multilayer spreadsheet grid, with each sheet composed of subsheets including a data tier, an "algo" tier, and a presentation tier (the default view is the presentation, but any tier can be viewed). Each tier holds multiple layers, as each tier has its own connection to data, its own algorithm, and its own presentation to the next layer. The default view would be something like a spreadsheet grid, but stylesheets in the presentation layer would make the graphical rendering arbitary and customizable. But in standard terms, so any object or collection with the same datatypes or overloaded operations could use any stylesheet that agrees with those types.
That could all be built on a Linux/x86 PC right now. I'd prefer to have the HW upgraded to a RISC/DSP/FPGA, like a PS3 with PCI-e for the FPGA. Because that architecture is inherently parallel, which my SW model directly supports. Everything is an object messaging other objects, so the RISC needs only to schedule which objects run in which execution unit, whether DSP or FPGA (depending on their arithmetic or their logic demands). The OS would pull the code from the object's datatype's class library, then either init and interconnect the DSP, or config some FPGA off buses or registers on the chip. Probably the whole OS would start as Linux running on the Cell's PPC, then gradually port each function to FPGA/DSP, gaining performance and flexibility along the way. I'd probably want to reimplement a filesystem under the VFS API, but implementing it in a relational engine that wrote directly to inodes of raw storage, before factoring FPGA/DSP out, but a good team could develop them each in parallel with the other. For good measure, I'd run the whole OS under a Hypervisor (virtualization like Xen or VMWare), which is how Sony runs its Cell PS3 already. With that architecture, arbitrary constraints on what a user can do that are just artifacts of how features have piled up on the Desktop over 30 years can go away. Which would let a lot more complexity under the hood tie everything together a lot more symmetrically, so the default presentation could be a lot simpler, as would exploring under that hood.
If I had a $million and a couple-few years with a dozen developers, I'd release it around 2010, and change the world.
Re:first post (Score:3, Interesting)
Actually, I'm fairly comfortable saying that. Deep underlying changes, complete rewrites... those are great ways to break the living heck out of a system that is mostly working very well. Whereas adding tools for the end-users (even kids) that don't yank the entire rug out from under every program in the system and replace it with a brand new rug which may be slippery, a fire hazard, contain uncounted numbers of weevils, and - by accident of course - is missing the rubber backing so you slip on it every time you step on it...
But really, I'm not worried about it. You know why? Because what I actually think we're going to get a year from now is an announcement that there's new iPhone software available. Perhaps accompanied by the news that there's a new iPhone, too. If we do get an OS X that has been substantially rewritten internally, I will (a) be astonished, and (b) let you test it for a couple of years before I make even the slightest move to upgrade. Because momma didn't raise no fool.