I have a story to tell.

About the 1983 time frame I got involved as the hardware designer and chief/only hardware engneer at a start-up company that was embarking to build an "office automation" type product. (I won't get into the whole story of what the product was here). I worked my butt off for about 1.5 years and produced what I still think was a sensational 80186 based set of hardware for the product. I needed to be able to test the hardware to ensure that it worked properly so I hacked the DEBUG.COM program from DOS 2.1 and got it to run standalone on the hardware booting itself into RAM from a floppy disk by way of a loader that I programmed into EPROM on the hardware. In the mean time I fixed fixed up a simple clone structure of the DOS .COM file format for executable programs and was able to use IBM's MASM 1.1 to assemble programs on an early Corona Data Systems PC clone. These programs were linked into .COM format that could run on the target system hardware. I implemented a whole array of diagnostic programs for testing RAM, floppy drive reliability, I/O ports, a 2400 baud in-built modem, and a simple 4-line telephone interface that was on the product. For user interface to the diagnostic programs I used a Tandy TM100 early notebook computer running a terminal emulation. By now you may be thinking....that Karas is pretty clever doing all that...using off an off the shelf assembler/linker tool and a PC for offline software development and then being able to debug those programs on the target hardware with the hacked DEBUG program.....Not so was the thinking of the software partners at this startup company. They had the cockamamie idea that software development for a product had to be done on the target product itself. So three programmers spent 2 YEARS writing their own debug monitor for EPROM, A program text editor that ran on the program under the monitor that saved its files to the floppy disk as tracks. There was no directory or file name system....the programmers kept track of what source was on what tracks in a notebook. Then they built their own absolute assembler that loaded from a 'track' and would assemble another source 'track' and write machine code to another 'track'. There was no linking mind you!! Needless to say they started this ludicrous effort only after I had prototype hardware completed and tested using the scheme I described above. Can you imagine developing a word processor program or telephone management function using a development system such as this? Well the 2 years they spent screwing around with this stuff the venture capital market went from GOOD to NON-existent. We had hardware to show to potential investors....but it was running diagnostics. Three main partners.....me @ 10% interest ->hardware....main business guru @ 50% interest -> business planning and financing.....software leader @ 40% interest -> promoting this stupid development plan. I begged the main partner to get the software people off the stupid railroad they were on with its diverging tracks...but he simply said to me over and over for 2 years..."Am I telling you how to build the hardware?" Needless to say after a total of 3 1/2 years of peoples time, and almost $2MIL of seed capital and no site of a source of venture capital the operation ground down to a halt...due in large part to delays in getting enough of the right software running on the product to capture the interest of an early venture capital investor.

So........always use what is available as a base for what you do next. Never waste human talent doing crazy silly things that can be bought for a little money. Take the best good ideas around and build upon those ideas to make your new things. This is in a way a promotion of the idea of doing step wise development. Great benefit can also be attained by also maintaining compatibility with existing technology. I'll finish by suggesting that you look here to see what more I have to say about stepwise development.

Michael Karas