Otaku, Cedric's weblog

Can you say "Steve want's to sell films and needs therefore a DRM compatible system"?

I've got one: Apple will still keep their high standards for third party parts, possibly even with a custom motherboard/chipset. This will hinder the efforts of those who try to make it work on generic PC hardware.

On a related point, there was a lot of emphasis, both during and after the announcement, on the "universal binary", which, as far as I can tell, is simply a PowerPC and x86 binary smushed into a single file. If so, this means that the "universal binary" must be around twice as big as each single binary.

This might be no big deal for shrink-wrapped apps that ship on CD or DVD, but these days, more and more apps are downloaded, so this means twice as much bandwidth, and twice as much download time.

Maybe this is not such a big deal, but it is definitely something that "wasn't said", as you point out. Suddenly, your favorite app is a 50MB download, where it used to be only 25MB, and what do you get for the extra 25MB? Nothing.

I guess the additional delay is for other vendors to catch up: applications to be ported, drivers for some new hardware to be written, etc. Yet, it's very promising and I'm sure the long time fans, the ones booing, we eventually be happy with this.

If you watch the keynote video you can see a nice ad from the 90s in which Apple bashes Intel's performances.

Overrall what I enjoyed the most in this announce his the confession from Jobs: "we did not manage to fullfill our promise, so we're going another direction to make it happen". At least they're not stubborn.

Now we just have to wait one year to see exactly what all this is about. If we don't expect too much (after all they speak about a 2 years transition), we shouldn't be disappointed. Quite the contrary with some luck :)

Oh and if Apple ships x86 based computers as appealing and as neat as their current PPC hardware, it's all I need to be convinced.

Cédric: you were right, you're pessimistic :)

Danny :
The applications size won't be doubled since only the binary need to be universal. All de the resources stay unique.

Stephan : Intel has denied that its new chips include DRM features. Steve Jobs has said he doesn't believe in any kind of digital protection scheme, the one in ITunes was put there only to get the rights from some majors :

http://www.rollingstone.com/news/story/_/id/5939600

>> Suddenly, your favorite app is a 50MB download, where it used to be only 25MB, and what do you get for the extra 25MB? Nothing.

Not really. This assume the app was all code. If that 25 meg app was 10 meg code and 15 meg resources it would only grow to 35 megs as a universal-binary file.

The thing that I think is even funnier is that people seem to think this sort of transition is hard or one that hasn't been done before.

All of this is actually a rerun of what happened in the NeXT community in the early 90's. At that time Steve and company decided to stop building their own hardware and supported running NeXTSTEP (then OPENTSTEP) on Intel, Sparc, PA-RISC while still supporting the old Motorola-based hardware. (Yes, these were "fat binaries" which are basically the programs for each architecture just glued together. And, yes, it made for big downloads. Thankfully, you could always strip down the executable (albeit after install) so you only kept the version you needed)

Since today's Mac OS X shares more common ancenstry with NeXTSTEP than it does with Mac OS 9, it's not surprising that OS X has also been building for Intel. What Steve probably isn't saying is that while they might have been compiling it for Intel, it probably only ran on a very specific Intel machine due to extremley limited driver support.

I would guess that the lag time is to build the Apple-blessed, Intel hardware, right appropriate drivers for it, and then get 3rd parties to move their software to Intel.

I also expect the 1-2 year timeframe to be an exageration and that some hardware will come out before that. After all, I can't imagine folks are going to be interested in buying the PPC boxes when they were just announced to be end of lifed.

Universal binaries are available for Mac OS X for the last 10 years. Ok, it was called NeXTStep at that time though :-)
So go and google for 'fat binaries' and 'NeXT' to see what is available.
No one prevents you from supplying an app in three flavours:
- fat
- ppc
- intel

I guess, for Apple this is one of the few ways to survive. With the nice centrino chpisets, even Win* notebooks are not too bad these days. And they are way che-aper anyway.

The porting bit is not too hard if you ever contributed to an OSS like NetBSD. So I beleive that the Mathematica port only took two hours.

You don't have to come up with hundreds of explanations: the G5 is a decent chip, but it will never fit in a notebook due to heat issues. Jobs knows notebooks are the future of the market, and the G4 is long in the tooth indeed.

@Guillaume: Yes, and Steve said "PPCs have better performance." And I think the MPAA is much more paranoid about films than the RIAA about music.

>>Since today's Mac OS X shares more common ancenstry with NeXTSTEP than it does with Mac OS 9, it's not surprising that OS X has also been building for Intel. What Steve probably isn't saying is that while they might have been compiling it for Intel, it probably only ran on a very specific Intel machine due to extremley limited driver support.

Well, That wouldn't be a surprise at all. Even Darwin pretty much ONLY runs on Intel chips in either a 440BX chipset motherboard or on a Thinkpad. You certainly can't take Darwin and run it on any old beige box.

And yeah, all the points about NeXTStep are very fair. They had pretty much this same .app structure long ago and it did work.

It's going to take a year or two for IBM to put the PPC onto the same die as an Opteron ;-)

Well, I guess I will see in 2 weeks when I get my Mac OS X86 system through the ADC. What this whole thing came down to is that IBM couldn't deliver a laptop chip. Not surprising since the G5 has so much in common with the Power architecture which isn't called LowPower for a reason I guess. At the end of the day the G5 is a decent chip but there is a reason my DP 2.5 ghz machine is liquid cooled.

As for universal binaries, my guess is that most people that have applications for download will just produce two, like people produce different downloads now for Linux. For instance, there are 4 different downloads for Eclipse on Linux right now. In the cases where it doesn't matter, its nice to have the convenience of a single app though.

As for your comment about machine translation not being rocket science, I would say that doing it well is a lot harder than just doing it. It looks like the IP they picked up from Transitive is a lot better than the PearPC emulation that we have already seen on x86.

Other more subtle questions do loom large though. What types of machines will be first? Laptops? Servers? Will there be support for x86-64? Will there be Opteron based machines? What is the performance of applications on the two architectures now that we can really compare Apples to Apples?

Why a 2 years delay before the switch ?
I think the main reason is that Apple could not break his deal with IBM in one day.
I heard that IBM was informed of the switch just a few days before the keynote.

What you have to remember is a universal binary does NOT make the file 2x the size. Only the executible part of the file has two files -- one for PPC and one for x86. All the rest of the files are shared.