Re: The Apple II is crap

[mjmahon@aol.com (Michael J. Mahon)]

Roy wrote:

>Linards Ticmanis wrote:
>> M. Pender wrote:
>> 
>> 
>>>>The only computer company to successfully make the transition from their
>>>>8-bit lines to their 16/32-bit lines, with any degree of compatibility
>>>>/at all/ was Acorn. And even then it was pretty kludged.
>>>>
>>>
>>>
>>>*ahem* and of course the IBM PC
>> 
>> 
>> Huh? The PC started out as (albeit crippled) 16-bit system from the very
>> beginning. Yes, the bus was 8-bit like, but the processor was internally
>> 16 bit which made it easy to go for real 16 bit with the XT.
>
>The only thing "16 bit" about the PC OR the XT was the internals of the 
>8088.
>
>XT was no more 16 bit than the PC.
>
>AT was 16 bit, including the 16 bit AT version of the ISA bus.

Since performance rises as a function of technology while keeping
the same software base, I do not think that performance is the
primary consideration when evaluating an architecture's "width".

The critical thing is the addressing model presented to software.
All the applications and systems developed for a computer will
be dependent upon that addressing model, and changes to the
model will _not_ be easily or quickly adapted to.

The 8088 presented 16-bit linear spaces embedded in a 20-bit
segmented architecture--much better than bank switching
but not as simple or versatile as direct linear addressing.  This
led to a confusing set of "memory models" for application
developers, and hindered "mix & match" for libraries written
to different models.

By the time that the 386 came out, with flat 32-bit addressing,
the former models were so entrenched that it took a decade
to jump-start a 32-bit application base.

The 68000 was the first widely-available microprocessor to
offer a 32-bit linear address space architecture, of which
only 24 bits was actually implemented on the initial
chip versions.  This encouraged the development of larger
applications, but (as with the IBM 360) the prevalence of
24-bit implementations led developers to use the upper byte
of addresses for other purposes, which again delayed
exploitation of true 32-bit addressing when the platform
began to support it.

The move to 64-bit computing platforms, now starting, is
similarly hampered by the entrenched base of 32-bit
software.  The ubiquitous "punning" by developers between
"integer" and "pointer" is a major obstacle to simple porting
of code to a new addressing model, as it was in the past.

The bad news is that these experiences demonstrate that
programmers as a whole do not concern themselves much
with the longevity of their code, happily limiting its portability
to simplify their lives.  (Of course, they don't spend too
much time thinking about buffer overruns, either.  ;-)

The good news is that 64-bit addressing should last for
32 doublings of memory capacity, or about twice as long
as the growth from 16-bits to 32-bits lasted.  ;-)

-michael

Check out amazing quality 8-bit Apple sound on my
Home page:  http://members.aol.com/MJMahon/