Amiga Kickstart Mystery Solved

[Pentad]

Hello all,

In another post I wrote that I've been following the Usenet posts from 30 years on oldusenet.net.  You can read the posts daily as they came up 30 years ago.

I read a post from RJ Mical this week that solved a mystery for me about Kickstart that I wanted to share (and hope you find interesting).

In late 1986 or very early 1987 (January) I was at a Commodore demo for the Amiga 500 and 2000.  The computers were pre-production and were so early they had the Commodore and Amiga keys instead of two Amiga keys next to the spacebar.  (My Amiga 500 was so early it had the Commodore and Amiga keys).

During the demo the Commodore Rep talked about how the Amiga had Kickstart in ROM (1.2).  During the Q&A, I (along with others) asked about loading older Kickstarts because 1.2 broke a lot of software.  Many EA games did not work with Kickstart 1.2 for example.  Archon crashed upon loading if I remember correctly during this demo.

The Commodore Rep said without missing a beat that you can load an older Kickstart into the Amiga 500 and 2000 by holding down special keys and turning it on.  He turned off one of the Amigas and tried holding down different keys to get a Kickstart hand screen.  When it didn't work he tried the other Amiga.

After a number of failed tries he actually called somebody at Commodore for help but they didn't know so he called somebody at EA but they didn't know either.  He apologized and said that these were pre-production machines and that production machines would be different.

Of course neither the released 500 or 2000 did this and it raises all sorts of issues about where the 256k would come from to store Kickstart (memory was not cheap), the memory location address, and other issues.

I often wondered if this gentleman was just confused or if Commodore had considered this at some point during development.  Outside of this demo, I had never read or heard of CBM doing this again.

Well, RJ posted this on 06/05/1986:

As a point of interest, even if Commodore ends up putting the ROM Kernel
system code in ROM, you will still be able to put a Kickstart disk in
the internal drive and have that code be loaded as the system code.
This allows further releases (as well as custom releases) of the Kickstart
even after the Amiga goes to ROM.

RJ was replying to a post about concerns with putting Kickstart in ROM on future Amigas.

So, at some point they had considered this for production.

All the development versions of Kickstart post Amiga 1000 were addressed at $200000 instead of $FC000 which makes a lot of sense.  

I just thought was interesting.

-P

[Pentad]

A couple of early A3000 offered the option to load a "super kickstart" from disk, which is - as you say - relocated to $200000. The kickstart sources include a special flag that performs this relocation and adds a couple of modifications to make it run.

A couple of problems remain, of course. Namely, expansion boards will typically not work.

Thomas,

Something else I thought was kind of interesting,  they have been pushing for a final bug list for Kickstart/Workbench 1.2 this past week or so (30 years ago).  Yet, nobody has pointed out that AutoConfig is broken so no autobooting HDs.  I remember Amiga World talking about how autobooting HDs were really being looked forward to in Kickstart 1.2.  I'm surprised it got missed, though, hindsight is 20/20.

Greg Donner's fantastic site (http://www.gregdonner.org/workbench/wb_12.html) has this info:

"Kickstart 1.2 (v33.180) did not have autoboot capability. Although it contained AutoConfigTM code, it was broken due to a simple bug: the autoconfig init code jsr had a wrong a6 register (the base register at the jsr time should be expansionbase but is execbase instead)."

If I've followed these threads correctly, it worked at some point in the development cycle.  So not to sound glib but you would think the "Huh, my HD is not autobooting anymore" would have been an issue someone would have spotted and mentioned.  Its a fairly obvious bug.  :-)

You know, I've always been so impressed with the AmigaOS.  When you look at other operating systems at the time (DOS, Mac, Atari) the Amiga is just so remarkable. Given the amount of money spent on the development of the Mac you would think the MacOS would be so much better, but it is far from it.


-P

[Pentad]

Apple had the honour of being the pioneer for this type of personal computer in developing it for a mass market. Pioneers must map the so far unknown territory and find solutions for the problems that arise as the product evolves. Apple spent an incredible amount of time and money on research and development before it produced the Macintosh in 1984.

This being a pioneering product, some of the decisions made in development led to operating system architecture design choices which were cemented for a decade, causing trouble along the way. For example, the original Macintosh had very little RAM (not more than a C64), and it leveraged the disk drive to compensate for that. The operating system design is strongly influenced by scarcity of RAM.

Anyway, the nice thing is that competitors can learn from what the pioneer did and travel along different roads. The Amiga certainly managed to do that, as Apple had, by mapping the unknown territory, done some of Amiga's R&D for them As did other pioneers which came before them, most notably SAGE Computer Technology on whose machines the Amiga operating system was built before Amiga had a hardware platform of its own.

I would disagree.  When you look at the MacOS (kernel) you would think it should be far superior to the Amiga's Exec.  Apple had a more money to pour into development, you had a longer development cycle (much longer than the Amiga's), and you had more people working on it.  Some of them had come from the LisaOS so they weren't exactly inventing the wheel.

While I have nothing but respect for Andy Hertzfeld, Steve Capps, Bruce Horn, and Bud Tribble, I think the kernel design of the Mac (and it flaws) are a direct result of Andy Hertzfeld writing a kernel for the first time.

Again, I think Andy is a great programmer and his work on Switcher and Savant really tried to fix some of the limitations of the MacOS, however, it really was an issue of scaling.

You mentioned memory (twice that of the C64), but in reality they were looking to 512k during development (just like the Amiga).  Burrell Smith was no fool and they knew 128k would not suffice.  Even the first Mac, though shipped with 128k, supported 512k with a simple upgrade and Apple shipped 512k Macs in the fall. (Burrell was a genius, it is unfortunate what happened to him)  

The Amiga was being developed at the same time as the Mac (it isn't like the Mac came out first and the Amiga team drew inspiration from it) but had far less money, people, and resources.  However, they were able to leverage a far superior kernel because of Carl Sassenrath's background in kernel development.

When you look at the two kernels there is just no comparison. Amiga's Exec supports preemptive multitasking, scales incredible well in all sorts of memory configurations and supports some advance features for the time.

-P