Apple previews Mac OS X Snow Leopard with QuickTime X - Page 4

What if it's there on purpose?

R O S E T T A

If you don't believe me google "boot powerpc leopard on intel"

You're just making guesses, aren't you?

Think about you said. A 32 bit kernel supports a 64 bit OS, but a 64 bit kernel doesn't support 32 bit programs of processors.

Nonsense.


Complete and utter hogwash.


Er, no, well-known champion of Linux.


Indeed, Linus accused HFS+ of being "complete and utter crap" and "scary", without providing any supporting evidence whatsoever. There was a thread discussing it here and the general consensus was that he was talking out of his backside.

Why do you think that HFS+ is so awful? Give me the juicy technical lowdown, not some self-serving quote from the purveyor of a competing OS.

Honestly, HFS+, far from being shit, is one of the best file systems out there. It supports journaling, long file names (with only one non-allowed character), case sensitivity (although third-party programs tend not to like that), meta-data, multiple data forks for files, has very high limits on number of files in a directory (2.1 billion), file size (8 EiB), partition size (8 EiB), and has low susceptibility to fragmentation.

Yes, ZFS is a more advanced file system, but that brings with it some drawbacks such as higher CPU overhead. Just because there are other more advanced file systems (and there's very few that are more advanced than HFS+), that doesn't make HFS+ shit.

I did not say that.
Both PPC64(G5) and x64(like Core2 and Xeon) processors support 32bit Operating Systems so Mach can be 32bit and still run on them, while CoreDuo(x86) and PPC32(G4) cant run 64 bit kernels so we can have both 32bit and 64bit mach for G5 and x64 and 32 bit mach for Core and G4.
We have a 32bit mach in Leopard with 2 architectures:
mach_kernel (for architecture i386)tMach-O executable i386
mach_kernel (for architecture ppc)tMach-O executable ppc
I'm just guessing that it may not be possible to have 64bit and 32 bit mach on the same package.
Even if it is possible to have 32bit and 64bit mach at the same time it should have 4 architectures: PPC, i386, PPC64 , x86_64! It wont reduce Snow Leopard footprint for sure
Edit: by saying "it may not be possible to have 64bit and 32 bit mach on the same package" I dont mean Fat binaries cant handle that(Fat binary supports this), I mean it may not be possible to have a single Snow Leopard release that contains both 64 bit and 32 bit kernels.

I imagine that Betrand provided a bit more detail on exactly how the increased performance and lower footprint would be achieved. It'll be a while before we get to find out, because that session was delivered under NDA.


Yes, you are missing something. Whilst multi-core systems will probably receive the biggest boost, the performance optimisations are unlikely to be focussed solely on multi-core. From the limited information provided, it seems that they will be doing some serious work to the kernel, and that is no small undertaking. The improvements are likely to cover other areas such as disk I/O, and virtual memory.

It certainly does!

It'll reduce it on disk, which isn't really important as you'd gain back maybe a few GB of hard disk and nobody really cares about that. Where it's important they reduce the footprint is in RAM and CPU cycles and since the PPC code isn't loaded on an Intel CPU, that's not an issue either.

I think/hope they mean they're reducing how much RAM some of the OS's frameworks consume, which have been getting larger over time. Just kicking the old Quicktime code into touch (no pun intended) would be a great start.

QUOTE=archer75;1262231]It boggles my mind that they are releasing Snow Leopard which is primarily a maintenance release.
Bug fixes, security and stability should always be patched in and seeing as how we are only on 10.5.3 we have alot of numbers left to go.
I wonder if they are simply running out of ideas to fit in an OS. I think Tiger was their shining achievement and they didn't deliver near as much with Leopard.

Or perhaps there is more than they are telling us. Rumors of dropping PPC support. They do mention improving compatibility with modern hardware and I did read an article just today that suggested Apple may indeed be moving to opening up OSX to run on PC's. Whatever the case I think something is certainly up.[/QUOTE]

You have a point. Perhaps no ideas are forthcoming. Perhaps ideas can't always flow constantly to the grey matter at Apple labs but there are a number (maybe a lot) of things they could fix/re-instate.

One of those that was a feature of Spotlight/Finder searches/Smart folders: the Search view. This was the view that grouped file types by kind, separated by blue separators and fitted out with some tools (Play slide show, top five, less etc...). This was the most attractive aspect of the Spotlight paradigm for me. It meant that a Smart folder with a project for 20th Century fine art would always have images of paintings by Picasso, Dali and Matisse in one place and ready to be presented in a slideshow. Now I am guessing that Apple dropped the search view because it was the default view in searches and may have been a hog on resources causing a general sluggishness in Spotlight. All very well and fine but for me the priority was functionality, not speed. Apple could just have the Search View, not as default, but as a view that one could use to tidy up the search and perhaps they may do this in 10.6 to clean up the mess that Spotlight can be.

The other feature Apple could polish are the stacks. In an early developer build of Leopard one could navigate through folders to different directories, much like hierarchical folders in the Dock. The stack HUD would change to reflect the contents of the folder. This was dropped in the final release of Leopard. For me when I saw a leaked video of this in action I was impressed. It was a killer. It reminded me of those pop-up folders in Mac OS 9. I was disappointed when they did not transpire. For me stacks are visual, a much quicker way to work if you know what you are looking for and where it is located.

One last thing. I hope that the romours are incorrect; that carbon may disappear. I use Apple Works for a certain task. It does the job well where other apps are more cumbersome.

Add a complete overhaul of OpenGL and the video card drivers to the list.

In any case...

Apple is cleaning up the OS. Apple is taking a step back and reviewing/rewriting code. Slashing code out. It's a VERY wise choice and it's one that very little software company (I'm looking at you Adobe) actually undertakes.

As much as some people would like to see an OS X update with tons of feature, it's absolutely essential that Apple takes this 'feature' break to concentrate on optimization/stability/clean up or risk ending up with bloat software like XP/Vista/Photoshop/Office.

People that aren't happy with the thought of this can skip 10.6 altogether and there will be no harm done. Leopard is excellent for what it is and will be good enough for the next 3-4 years.

I don't think carbon will disappear, but I think it will be discontinued on the programming side.

One thing. Forgive me if I am incorrect (currently on vacation and head generally not in Mac things) but did Apple state that Snow Leopard would be released as retail some time next year? Did I hear that Snow Leopard would only be available on new hardware and that it has not been formally named as 10.6? Would this make sense? It would be like there not being a retail version of Tiger for Intel Macs. Hence, the low profile announcement of Snow Leopard. There may well be a Mac OS X 10.6 or 10.7 named Lion or Tabby or may or may not run on PPC? Who Knows?

No, they didn't state explicitly that it would be released to retail, but I see no reason why it shouldn't be. There will be plenty of machines out there right now that will benefit from the improvements. I expect it to be available for $50 or less.



I don't think you did.


Indeed it has not been formerly named as 10.6 yet. But they have referred to it as the "next major version", so one would have to assume that its full name will be "Mac OS X version 10.6 (Snow Leopard)"

As a code developer, I'd welcome a machine that doesn't hang-up as frequently as 10.5 does, especially 10.5.3. Apple should pay attention to fixing what they've got and not adding more crap that we don't need, won't work, and we possibly can't afford, Just make it work.

This is because 90% of people will gladly accept mediocrity as the standard.

That's an idiotic statement. The quad G5s are as fast as most of the models apple is selling now, and they're not that old. Or do you think Apple customers should just blindly buy new machines even if the old ones are more than fast enough just to line Stevie's pockets?

So was there any mention at the event about intel/PPC? If not, is there any word of anyone attending the event asking Apple folks about that?

What's the green button?

16 terrabytes sounds like a lot, but since 10.5 already supports 4 TB of physical memory, is it really that big a deal? 10.7 or 10.8 will probably be out before Apple ships a machine that can hold more ram than 10.5 can access. Same with multicore - it would sound much more impressive if Apple didn't already claim that 10.5 was a multicore powerhouse. They have already promised things like this and failed to deliver, why should we have any faith in these claimed improvements until they actually ship it and show the benefits with benchmarks?


That's a pretty big assumption to make from such a vague statement. For all we know they're just going to switch to a smart installer that leaves out the unnecessary code, or at least gives that as an option.


Are you sure about that? In the case of APPS, including apple ones, they include the PPC code in many cases. Are you sure they don't do the same with OS code?


Because there are fixes and optimizations specific to PPC. Devs would be pissed if they put effort into a version that's not supported. Apple has to tell devs that soon if that's what they have planned.

I for one am looking forward to Snow Leopard, whether it's a paid update or not. I don't think it's simply about optimizing the code (though the Safari team has shown that they can squeeze a lot of performance improvements out of doing that - Squirrelfish is much faster than the current Javascript implementation in Safari), I think it's about a whole new way of converting the code to binaries.

Apple is the biggest developer of LLVM, after it hired Chris Lattner who developed it at the University of Illinois. LLVM (http://llvm.org/) is really aimed at optimizing code as part of the compilation strategy, and is partly what Apple will use to replace gcc. gcc, while it has been fabulous, is showing its age, in that it has a lot of difficult code to maintain and improve, and it's easy to break things. Apple realized that if they rely on gcc as a compiler, there's not going to be revolutionary improvements in the produced code. So instead, they decided they needed a new back end for compilation, and that's llvm. They have also decided they need a new front end, which is the actual code parsing itself. Currently llvm relies on gcc for this, but Apple is developing a new compiler, called clang (http://clang.llvm.org/) which is already showing useful performance gains, though is early in its development. Looking at the sessions at WWDC, one of them on compiler technologies says:

"Xcode 3.1 introduces two new compilers for Mac OS X: GCC 4.2 and LLVM-GCC. Learn how the new security and performance improvements in GCC 4.2 can help you produce better applications. Understand the innovations in LLVM-GCC, and find out how you can use it in your own testing and development. Finally, get a preview of future compiler developments"

I suspect that Snow Leopard will be completely compiled by LLVM, though not sure whether they will have clang ready to act as the front end, so it may use gcc for that part. The result will end up being dramatically faster executables, even with no changes made to source code of the applications. However, given the APIs that they are developing for taking better advantage of the multiple cores and graphics chips, apps that take advantage of those (e.g. Quicktime X) will show even more dramatic performance gains. Another interesting thing about LLVM, is that you could ship code that is partway compiled, and then have the compilation finish on the target machine, meaning that it is optimized for a user's particular hardware (this is similar to what Gentoo Linux does, though you get the source in that case, rather than some intermediate compiled form). If it knows about your processor, and your graphics card, then there are certain additional optimizations that can be made at compile time, rather than having to figure out what code path to take at runtime.

While there may be no new "user visible features", as in a new app, or new UI, or new paradigm such as Expose, the amount of work that will have to go into developing Snow Leopard will be no less than the amount of effort that has gone into past releases. This is much more than a 10.5.x release - I'm expecting *some* applications to show more than 2x increases in performance. If you offered me a processor that was twice as fast for the cost of an OS update (who knows what or if it will cost, but it will certainly be no more than $129), I would buy it in an instant. I believe Snow Leopard will be just as good an investment.

Clang doesn't sound like it is something you would bet an operating system on. It hasn't been out there long enough. I don't think Apple is using gcc for the kernel.

From the clang page.

I agree clang is not ready for prime time (and said I didn't expect it to be in Snow Leopard), but I do think in the 2-3 year time frame that LLVM+clang will be what XCode uses to compile everything. AFAIK, Apple uses gcc for everything currently, except for the Open GL stack, which in 10.5 started being optimized with llvm (http://lists.cs.uiuc.edu/pipermail/l...st/006492.html). Also see http://arstechnica.com/staff/fatbits.ars/2006/8/17/5024

I used to be a CIO and I disagree. The Intel Macs that won't run a 64bit kernel will be four years old by the time Snow Leopard will be released. Apple typically continue to provide security fixes through two major releases i.e. 10.5 will get security support until the release of 10.7.


The point here is to transition from a 32bit kernel to a 64bit kernel. In theory, Apple could ship both but it would be a development and support hassle as well as Apple's desire to use 64bit kernel features in user space. The consequence of a 64bit kernel is that it will not run on anything older than C2D. Both 32bit and 64bit userspace apps will continue to run.

Apple uses GCC for the Kernel. They use their GCC and ObjC/C runtime augmentations for their kernel.

PDF Slide presentation on what LLVM does:

http://llvm.org/pubs/2006-04-25-GelatoLLVMIntro.pdf

Page to Lattner's MS Thesis:
http://www.llvm.org/pubs/2002-12-LattnerMSThesis.html


Clang Overview


Note the reference to Commercial. The GNU Compiler Collection has been a struggle for Apple regarding licensing issues and getting changes uploaded to trunk for several years.

KEY FEATURES:

End-User Features:

* Fast compiles and low memory use
* Expressive diagnostics
* GCC compatibility

I'm glad someone else besides me noticed that blurb on the Apple website.

While I agree with solipsism that PPC and Intel 32-bit are toasted with this release, might there exist any other way Apple could cut the Mac OS X footprint while leaving PPC support?

I'm not sure how they're doing it, but based on screenshots it looks like apps are much smaller AND they are still universal.

http://orchardspy.com/

... and it is running on a Core Duo which means 32bit support

Someone is paying attention

Indeed. The level of crap and stupidity being thrown around in this thread exceeds normal levels of shit-flinging.

Everyone's doing the same mistake. The apps aren't smaller. If you remove the localization from each app, you get very similar sizes to that seen in Snow Leopard screenshots. Try it yourself...remove all localizations except your preferred localization. In some cases you'll end up with an app that is exactly the same size as seen on the screenshot, in other cases you'll end up with something much smaller.

Some apps probably don't have the localizations yet. Simple as that. It makes a huge difference in some cases...as much as 400%. Remove Safari localizations and you'll see it go down from 50+ MBs to 9.2 MB.

I'd be extremely surprised if those apps had all their localizations.

I just think it is nice that Apple polishes off the OS. I am not going to complain about paying for it either. If $130 polishes off Leopard before 10.7 or X, so be it. It never hurts to slow down.

I have a G5, CD MBP, and several C2D machines. It would suck to see my MBP not have support on Snow Leopard (bad name). I could deal with the G5 and understand it but in reality, it seems that Apple would want smaller programs, more efficient code, and universal apps in order to keep their options open for future technology turns that may happen that we can't see but it would be nice for both hardware and software to take a major leap forward and leave the past in the past at some point.

You're not accounting for one obvious oversight: Those applications are in debug mode.

or....visions of PCjr running through my mind

True...10.6 will finally be the release where Apple remove all debug code from their apps (that's what they mean when they say they're going to maximize performance for existing apps ).

This has probably already been said...

I think Apple will push developers to write for cocoa only, but 10.6 won't drop carbon support.

As for PPC? I think 10.6 will run on Intel Macs only. I'd like some minor changes, like all scroll bars being the same (look at iTunes and then Safari)

Some new wallpapers wouldn't go amiss either

Leopard is an excellent reason to release Snow Leopard. Its incomplete state at release and its subsequent wobbliness reflect the fact that its code base had become too unwieldy to extend.

Conveniently, the iPhone team had to consolidate a lot of OS X to fit it on the iPhone. So Snow Leopard is essentially the lessons of the iPhone port rolled back into the main OS (note: lessons, not literal code), plus some further work developing Core.

Thank God they are finally, finally, finally going to retire the old QuickTime. (Non-trivial) QuickTime support was always where the most elegant Cocoa code would suddenly bog down into hundreds of lines of suspiciously Pascal-looking code. If Apple's multimedia layer is as pleasant to code for as it is to use developers everywhere will be happy. Also, that huge pile of transliterated legacy code cannot be easy to adapt to modern architectures.

Snow Leopard will be a great foundation for subsequent releases, and it will also lead the way forward for devices which are neither desktops nor laptops (but which are definitely Apple hardware—why on Earth would Apple break their seamless integration strategy when it is consistently kicking everyone's asses?).

Excellent points.

I haven't seen any confirmation yet that QuickTime X is a total rewrite.

It absolutely must be a total rewrite for several reasons:

Amorph mentions one...it has to be a pleasant and modern architecture to develop for
Almost a quarter of the security problems on Mac stem from QuickTime, a rewrite would allow Apple to take a step back and see where the security problems are.
It needs to be fast/efficient (I'm assuming this is what Apple means by lessons learned from QuickTime on the iPhone)
I suppose it needs to be 64-bit (hmurchison makes a point that 32-bit QuickTime is mostly deprecated)

If Apple doesn't rewrite QuickTime and only patches the old QuickTime, they've missed the boat and have led people into believing this is a brand new architecture with the name change.

We need more NDA breakers.

VERY VERY interesting. Much like Intel and PPC are parallel, so shall Leopard and Snow Leopard. I like this path. Although i'm happy to let my PPC die as to focus development on one roadmap. My PPC on Leopard runs rather poorly with Safari crashing daily, iPhoto running really slow and having issues, and many other little things.

I'm glad to see someone take the responsibility to harden their platform. Now this all could mean that Apple needs to re-write lots of code in order to allow touch in Lion, or other features as they have suggested, but either way, or both ways, i'm glad and will be happy to spend the $$ on SL.