[DM-MUG] Apple moving to Intel - confirmed by Jobs at WWDC

[Rod Ragner]

Adding Intel processors is not a big deal. When I was a NeXTSTEP/ 
OpenStep (the predecessor of Mac OS X) developer, we built multi- 
architecture binaries for NeXT (Motorola 68040), HP, SUN, and Intel  
(486 and Pentium).

There are a few caveats: You have to write processor agnostic code  
and refrain from writing code that is tied to a particular byte-order  
(little-endian versus big-endian), but this is not particularly  
difficult and is arguably what "high level programming" should be  
about -- let the compiler deal with the bits...

Any application actively being developed/maintained should be  
recompiled (perhaps with a few changes to remove processor specific  
code) rather quickly.

Smaller developers might take longer, but their applications are  
typically smaller and will require less changes, so they too should  
be able to modify them rather quickly.

The problem is with applications that are not currently being  
developed/maintained. These 'deprecated' (perhaps owner-less)  
applications will probably run fine via Rosetta (the processor  
translator), but not ultra fast. Remember that Rosetta only emulates  
a G3, not a G4 nor G5 and will not support Altivec nor Classic  
applications.

This, of course, presupposes that these are Macintosh applications.  
Windows applications are not Macintosh applications and will still  
have to be completely rewritten to run on the Macintosh. Whether  
Windows-only developers decide to also develop for the new Macintosh  
computers requires that they develop for both PowerPC and Intel-based  
Macintosh computers, just as they would have before. Therefore, we  
may only see more people developing for the Mac if they feel that the  
market is growing enough for them to be profitable.

This new migration will be relatively painless for any users who have  
already migrated to Mac OS X native (Cocoa or Carbon) applications.  
People who are still trying to survive on Classic applications are  
going to feel a lot more pain.

Since Apple has said that the first Intel-based Mac will ship in June  
of 2006, we will hopefully continue to see the existing PowerPC Macs  
and hopefully some with more powerful processors for one year.  
Whether we see the dual core PowerPC 970MP is anyone's guess, but if  
the processor has already been "tapped out" then Apple will ship one.  
If not, then we will only see incremental speed improvements. That  
said, IBM and FreeScale now have even less interest in catering to  
Apple, which could mean that they will just stop any development they  
may have been undertaking and we may be stuck with the models that we  
now have at the speeds that we now have.

Could Apple ship an Intel-based Mac sooner than one year from now? I  
would imagine that they could get one designed and in production in  
less than six months. If they have been contemplating this change for  
a while, which I suspect they have, then they have probably been  
designing them for a while now and could ship one much sooner. But,  
Apple would not ship one until enough of the most important  
applications have been built with "universal binaries," which could  
be the limiting factor. If Apple, Adobe, Quark, etc. build their  
universal binaries quickly, then we could see the new computers much  
sooner. But, I would guess that the big development houses will wait  
and incorporate the changes in their next major revisions, which  
might take one year...

In the end, I am not particularly concerned about the process change.  
It is not a very large impact on good developers. Macintosh users  
should not be too concerned, since their favorite applications are  
likely to be modified quickly.

This migration is a double edge sword, though. Apple no longer need  
be concerned with people thinking that their computers are slow when  
compared to Wintel computers. But, they will NEVER BE FASTER either!  
At best, Apple will now be at parity with the other hardware. Sure,  
they can innovate superior support chips, memory subsystems, device  
protocols, etc., but the processor will no longer be an advantage.

A potential bright spot is that this multi-architecture (actually  
dual-architecture) change permits them to add new architectures in  
the future with similar minimal changes to applications. Consider  
AMD: We do not know the details of the 'deal' with Intel, but Apple  
may be able to use AMD and other processors when they choose. They  
might even support the "Cell" processor someday (although that would  
require significantly more developer work). It also means that the  
PowerPC is not completely dead; IBM or FreeScale could someday choose  
to make a great processor that Apple could use.

"What, me worry?"

Rod