[Ti] Powerbook MacIntel... could be, could be... what do you think?

[Kynan Shook]

Chris Olson <chris.olson at astcomm.net> writes:
> Fact:
> With the billions of lines of code that have been polished on
> PowerPC, all the problems are not going to be found in the port to
> x86.  For instance, an integer divide-by-zero exception in some
> obscure subroutine results in a crash on 86, while on PowerPC the
> same operation returns zero.
>
> The first purchasers of MacTel computers will be public beta
> testers.  And I do know that for a fact, because unless Apple changes
> their methodology, there will be no public beta testing prior to
> release.
...
> Releasing a new operating system version is trivial compared to
> releasing a new architecture port.  I laugh at the comparisons when
> people say this is no different that going from 68k to PPC.  Here's
> another fact: back then there wasn't much code.  Today there's
> literally billions of lines of it and Mac OS X is a complicated beast
> that takes up so much disk space that it couldn't even be *stored* on
> the supercomputers of the 68k days.


Ahhh, Chris, always the pessimist.

There are two distinct issues in this transition: Apple's support,  
and third party support.  Third party, obviously, depends on how  
incompetent your favorite software vendors are; and if they don't  
make a good product, then somebody else will push them out of the  
market.

So, really, the issue is going to be Apple, and from what I have  
seen, there is absolutely no reason to fret.  They have been working  
on this transition since OS 10.0.  Seriously.  While talking with an  
employee at WWDC, I was informed that Apple has had a policy that  
every software product made (since OS X; so things like iPhoto, etc.)  
must be architecture agnostic; that means that endianness and  
exceptions like your divide by zero example can't be relied on.  This  
way, the Marklar program that has been responsible for the Intel port  
has been able to run all these applications on Intel hardware for  
years, possibly even without the knowledge of the programmers.

For that matter, one of Apple's programmers told me that he has been  
using a Dell box running OS X (probably versions 10.2 and 10.3) for  
quite some time because, at least for his particular CPU-intensive  
task, the Intel hardware was significantly faster.

Many developers have found that, if they follow good programming  
guidelines, the transition to Intel is trivial.  This is especially  
true for cross-platform products like Mathematica that are,  
presumably, designed modularly with architecture differences in  
mind.  I'd assume that the cross-platform OpenOffice was designed  
with similar ideas.

So, seeing as how OS X has been running on Intel for years, and they  
have been planning on it since the beginning, I have a feeling that  
the vast majority of the bugs will have been squashed.  There's  
really not that much to worry about, and I wish you'd just wait to  
pass judgement until there is an actual product to pass judgement upon.