OS X had better kick ass.

September 12, 2000 1:00 PM

OS X had better kick ass. For me this is Apple's final curtain call. I have been a life-long, devoted follower of the Great Apple and now it is time my devotion paid off. Cool hardware is great Steve, now lets see the goods on the inside.
posted by Brilliantcrank (16 comments total)

User name and password required on reboot? Stupid, stupid, stupid; let's hope that's dead in the final version. 128 MB RAM required - well, that's no problem, as long as I don't do anything rash like run an application.

My favorite part of the story: Virtual PC will not run in the "Classic" environment. Asking an emulator to run an emulator does seem a little cruel. But Q3 and Unreal Tournament supposedly run 33% slower in the Classic environment, which makes X a non-starter for Mac gamers. (Both of them.) It also doesn't support AirPort yet - so no more answering email from the iBook on the sofa for a while. Also sounds as if it has trouble with FireWire devices, so there goes the interface with my digital camera. Maybe it'll access the camera in Classic mode, but digitizing video in emulation mode is just asking for a three fps stutter-fest.

I'm a diehard Mac user, and I can't possibly think of a reason to upgrade now. Can't see one good reason yet. A Dock full of throbbing icons is not sufficient enticement.
posted by lileks at 1:48 PM on September 12, 2000

"user name and password required on reboot?".

Although I'm not an OSX engineer, I can practically guarantee that this will be in the final version. Why? Because it's a fundamental component of the Unix Way. No Unix-type operating system would EVER dispense with this fundamental security feature. If this is a problem for you, don't go anywhere near OSX. It's Unix (well, BSD, same thing).
posted by aramaic at 1:57 PM on September 12, 2000

OK, it's a beta, not a full release. Maybe it will get better before the formal release.

It better damned well, because what's been presented now is distinctly unimpressive.

It requires 1.5Gb of free disk space to install. (And people complain about Microsoft bloat.)

The ZD article describing it had some other interesting facts:

If you're running multiple OS9-compatible (i.e. classic, non-carbonized) programs and one of them goes bozo, it can take out the entire Classic environment and kill all the other classic programs which are running with it, because there's only one virtual machine. In that case, the Classic environment crashes and is shut down, although OSX itself continues to run. It becomes necessary to start the classic environment again, a process they describe as taking "several minutes".

This is in contrast to NT and OS/2, for instance, which create a separate executable environment for each legacy (i.e. DOS or Win 3.1) program. On those systems, if one of those legacy programs goes bozo, it has no effect at all on the others because it is living on its own virtual machine. It literally thinks it's the only program running, and taking out that virtual machine has no effect at all on any others which might exist. Those virtual machines are set up on an as-needed basis, and creation of them is nearly instantaneous. On OSX, on the other hand, they recommend that you start the Classic environment at boot time and let it be available in the background so that it's there when you need it and you don't have to wait for it to begin.

Apparently "carbonized" apps don't have this flaw, but just how many of the most important Mac apps have been "carbonized" yet? And even if they have, have the customers purchased new versions?

No, the fact is that for the next two years, the majority of programs running on OSX will be "classic" and neither "carbon" nor "cocoa".

I'm sorry, I don't find that very impressive. What the heck is taking the Classic execution environment "several minutes" to initialize? For crying out loud! It takes NT about one second to create a virtual Win3.1 machine.

There's also this: all the people running OS9 who finally were able to replace their piece-of-shit ATI display cards with something more modern from 3DFX, must go back to the ATI cards for the installation. (They better hope they can find them. Then they better hope they can figure out how to make OS9 work with them.) No word as to whether they have to remain with them after installation is complete. It depends on whether 3DFX has managed to come up with an OSX-compatible driver. But whether they have or not, it ain't on the installation disks, which only speak "ATI". And apparently it doesn't recognize a lot of auxiliary disk drives as really being there, either.

OK, as I said, it's a beta. But two comments:

1. It better damned well get radically improved before release. If loading the classic environment takes more than five seconds on a healthy computer, then they've blown it. There better be a whole lot more in the way of device driver support, too. Full release better be able to install on a computer with a 3DFX card in it.

2. Things won't REALLY be healthy in the Mac world until the majority of apps being sold are cocoa-compatible and classic/carbon has gone the way of Win16.

For Microsoft, the full transition from Win16 to WIN32 (which amounts to the same thing in terms of commercial transformation) took about five years, and that was with Microsoft spending hundreds of millions of dollars per year encouraging and helping its ISPs to make the change PLUS it didn't begin until Win 95 had become sufficiently established to represent a viable market for WIN32 apps. Serious commitment to WIN32 design starts didn't really begin until about mid-1996, when there were 40 million Win 95 systems around and it was clear that a WIN32 app had a viable market (because of course they couldn't be sold to Win16 users).

The transition from "classic" to "cocoa" won't begin in earnest until there are at least 20 million Macs running OSX. That's when the ISPs will begin to make their design-starts using cocoa instead of classic, because they'll feel confident that there's an adequate opportunity for sales. But those apps won't appear on the market for another two-to-three years because of the lag between design-start and product shipment.

It's also going to go a lot slower because Apple simply doesn't have the resources Microsoft had to pour into encouraging the ISPs to convert. Microsoft's effort in this regard was immense and it remainsimmense.

And I'm not talking about blackmail, I'm talking about MSDN. MSDN is a huge expense for MS, and membership fees don't pay 1% of how much it costs MS. Immense amounts of sample code which you could use for free; huge amounts of documentation, quarterly updates, and even MS engineers who would visit your site and help you with intractable problems on MS's nickel -- THAT is what MSDN represented, and why the transition succeeded.

Apple will do their best; we'll see just how "best" that is. There's just no way Apple can match that level of effort, and it's not at all obvious that what Apple will do will be enough to make the transition from "classic" to "cocoa" rapid and complete.

posted by Steven Den Beste at 2:44 PM on September 12, 2000

OSX is NextStep with some sugar coating. Yes, Classic apps will be second-class citizens, but if Apple (well, NeXT) have done their job right, you won't have much more trouble with Classic apps under OSX than you do now.

As for the delay in Cocoa apps: NextStep apps should still compile and/or be able to run in OSX pretty easily. Although NS has been pretty much dead (or on life support) for years now, there are still tons of NS apps around (e.g. the original HTTP browser, called "World Wide Web" :-)...

I am not a Mac user; I sysadmined Macs for a while right during the transition to PPC and I abhored the underlying technology as much as I liked the consistency of the UI. OTOH, at the same time I was hearing the rebooting chimes of at least a Mac a minute, I was sysadmining NextStations.

To this day, 5 years later NextStep 2.0 (an 1992 OS if memory serves) was a better overall OS for my money than anything else I have ever used (let me see: Windows from 3.0 on, Mac from System 5 to around 8, SunOS and Solaris, Irix, FreeBSD and Linux, OS/2 and GEM, HP-UX and OSF/1, Unicos and Dynix, CP/M and Amstrads, in no particular order). I have great, great hope for OS X. If Apple doesn't botch it up with high prices both on software and/or hardware, the post-OSX Apple can be the Mercedes/BMW to Microsoft's Chevy.

I am even thinking of dumping this Win2K laptop and going back to the candyland... 'coz I want my NextStep back :-)...

posted by costas at 4:40 PM on September 12, 2000

Does anyone have any experience with / opinions about BeOS? How does it compare with the Mac OS?
posted by wiremommy at 4:49 PM on September 12, 2000

I'm running the BeOS on one of my cpu's. It is nice, very stable and very intuitive. I was hoping it was gonna be bought up by Apple a few years back before the candy machines came along, and become the default OS for all Apple products. That was not the case despite all the rumors at the time. I certainly love it more than the mess that 9.x has been. Big hopes for X.
posted by thirteen at 4:56 PM on September 12, 2000

I would move to Linux in a hearbeat if I could run Macromedia and Adobe applications. I know that I can emulate Windows but it is to damn slow. I have been rather impressed with Windows 2k so far. I prefer MacOS but I am tired of all the memory errors and applications that just quit because of a flaw in the OS.

BeOS is cool but it lacks any applications.

I think I might go back to Lincoln Logs.

posted by Brilliantcrank at 5:13 PM on September 12, 2000

The startup time for the Classic window can't be 5 seconds. You're basically running a completely different OS on top of OS X. Windows could launch DOS windows easily and quickly because Win95 and 98 were all built on top of DOS. Also, since DOS was text based and OS 9 obviously isn't, there's going to be a huge difference in startup time there. Compare it to VPC starting up on a Mac, rather than DOS on Windows.

I agree that it still has a long way to go before it's ready for prime-time, but this is just a beta. They've got at least 4 months, probably more until they release a full version. Hopefully they'll be able to bring down the RAM and HD sizes in that time as well.

posted by mzanatta at 5:42 PM on September 12, 2000

The market is in applications. OS's aren't profitable anymore cause they're free. Admin's and applications are what cost $$. You all keep talking about stability, and cross-platform portability of: your applications. Not the kernel, not the GUI, but what you use to get the job done. My guess is that when MS splits, the app division is where the money and investment opporunities will be and if MS doesn't port to Unix and other OS's they'll just lose customer base to cool stuff like Mozilla and other open source projects. Anyone agree? Anyone have insights or disagreements with me? I too have high hopes for OS 10 but to think it's going to be different? No way. Mac's have never been different, just better looking.
posted by greyscale at 6:57 PM on September 12, 2000

mzanatta missed the point: I never mentioned Win 95 or Win 98. Never once. Why are you talking about them? (Perhaps because you can't excuse what I was really talking about?)

NT launches a Win3.1 session in about a second, and NT is not "built on top of DOS" or any of that other crap. Also, Win3.1 is graphic, not text. And Win 3.1 is "a completely different OS" than NT. (NT was a complete rewrite.) I don't see any qualitative difference between the relationship between NT and Win 3.1 and the relationship between OSX and OS9. Thus I don't see any excuse for why the performance should be so radically different.

[And if you won't accept NT as an example, OS/2 is just as good at bringing up Win 3.1 sessions as NT is, and it doesn't even come from the same vendor.]

I see a very serious case of strawman here: someone's changing the subject. I talk about NT bringing up Win3.1 sessions rapidly and you respond by talking about Win 95 bringing up DOS sessions rapidly. I don't understand what relevance your comments have to mine.

If NT and OS/2 can bring up a graphics Win 3.1 session in a second, why does it take several minutes for OSX to bring up an OS9 graphics session?

posted by Steven Den Beste at 7:53 PM on September 12, 2000

By the way, folks, a small point of usage:

To most of the world, "X" means "X-Windows". It might be a good idea to refer to this operating system consistently as OSX and not use the diminuitive "X", in order to avoid confusion and ambiguity.
posted by Steven Den Beste at 7:57 PM on September 12, 2000

Ah come on, what are you all gripping about? Look at the lovely bejewelled interface... look... look... it's drawing you in... aaahhh... shiny beads... I will trade them for my homeland...

See. It worked on me.

And anyway, it's a damn sight more impressive than the equally long-awaited k10k redesign.
posted by James Bachman at 8:22 PM on September 12, 2000

"user name and password required on reboot?"

I don't work for Apple, but for a company that has been working on on new versions of our apps for awhile now and have had many developer copies and many discussions. The idea is right now that their will be two versions of the OS. One that you'll have to supply a login for and will still give access to a terminal window. The other will be for the general population and will remove the login and access to all of the Unix that it's built on. That's the last we were told from Apple.
posted by monkeyboy at 4:37 AM on September 13, 2000

why does it take several minutes for OSX to bring up an OS9 graphics session?

Erm, because it takes several minutes for the MacOS to do it right now???
Why do you expect anything else? It's not Windows, and by design is intended to keep MacOS 9 and MacOS X as far apart from each other as possible.

When all the main apps are carbonised [which with the exception of Office will be pretty soon] many users won't see the classic environment ever again ['cept for legacy games]. The transition to Cocoa is a completely different matter, and many apps will remain Carbonised at least until its time for a complete rewrite, probably longer.
posted by theparanoidandroid at 6:20 AM on September 13, 2000

When all the main apps are carbonised [which with the exception of Office will be pretty soon] many users won't see the classic environment ever again.

IF they replace their entire installed base of software with new copies. Do you really think they'll do that? The fact that they've become available (and it's questionable whether it will be as fast as you say) doesn't mean the customer base will instantly switch over. On the contrary, most won't even if the upgrades are free (and in many cases the upgrades will cost money).

And the fact remains that NT and OS/9 are capable of bringing up virtual machines to run Win 3.1 apps in under a second; I don't accept why it should take several minutes for a Mac to do the same thing for "classic".

OK, now I'll answer my own question: it's because the Win 3.1 emulation for both NT and OS/2 isn't actually Win 3.1. Rather, it's a shell which emulates Win 3.1, but most of whose functions are actually implemented by calls to NT and OS/2 (outside the virtual machine). The Win 3.1 emulation environment is an entirely new piece of code; it isn't actually Win 3.1. It loads fast because it's small. (It also has a small memory footprint.)

"Classic" emulation under OSX is slow to start because it's creating an entire emulated Mac, and then actually loading all of OS9 into it. It loads slow because no-one wrote a thin OS9 emulation layer; they're actually using all of OS9 (possibly in slightly modified form). It's footprint must be huge, even before you load any apps into it.

Under OS/2, it's possible to create a VDM (virtual DOS machine) and then to actually load and run real Win 3.1, unmodified, within it. And that process is much slower to initialize, taking a minute or so. That is the analogy to what the Mac is doing. But the OS/2 users who demonstrated that did so just to show that it was possible; no-one in their right mind actually uses it that way. Everyone who wants to use a Win 3.1 app uses that thin emulation layer which loads real quick.

If the "classic" environment under OSX was only going to survive in common use for a few more months, this might be excusable. But any estimate to that effect is hopelessly naive.

OSX users will be using classic (non-carbonized) apps heavily for years. Any claims to the contrary are wishful thinking. And that means that the current implementation of the classic environment under OSX is scandalously bad.

posted by Steven Den Beste at 8:54 AM on September 13, 2000

OK, You're so smart, how would YOU have done it?

The other reason that the Win 3.1 environment can start so fast under NT and OS/2 is that there's no initialialization to be done; the image which is loaded is already initialized and ready to go.

Most of that time getting classic into OSX is spent letting classic do all of its normal initialization. But why is that necessary each and every time? We've got a perfectly good virtual memory system (finally!) so let's use it.

Here's what I'd do: During the OSX installation, load classic. Let it take its time. Let it do all its stuff. Once it's finished, SNAPSHOT ITS MEMORY IMAGE to disk. When you need classic, load the snapshot, already initialized, already ready to go. Load time is however much time it takes to clear enough memory and load the file from disk -- very rapid.

Provide to the user a utility for updating the snapshot. If he adds a new (I don't know what they call it, those funny background programs everyone is so fond of) or a new device to the system or something else which requires the classic environment to change, the user then runs this utility. What it does is let classic do a formal load sequence from scratch AND THEN UPDATE THE SNAPSHOT ON DISK which can still be loaded when needed in just a few seconds.

Why was this so difficult to implement?

posted by Steven Den Beste at 9:18 AM on September 13, 2000