On Tue, 6 Jun 2023, Bill Degnan via cctalk wrote:
I thought the goal back then was not 100% hardware compatibility, it was MS/IBM DOS compatibility. To be able to load/run/copy files from one PC to another, dBASE, Lotus, Wordstar, etc. I don't think most manufacturers cared as long as the software worked and you could make a printout using a standard printer of the day like an Epson or whatever.
Most people, even those who explicitly stated that it was, were not intereste in 100% compatability, they just wanted something that would not give them unexpected incompatabilities.
I tested all of my software on real IBM hardware. If it worked on the real hardware, but failed on a customer's "compatible", it would be a drag, but not the disaster that it would be if it had worked on my "compatible", but failed on the real thing.
People who tested software on "compatible" machines sometimes got surprised. Trivial example: if the attribute of a text character is both reverse video AND bright, the PC CGA handled that differently than the Columbia CGA.
F'rinstance, one of my early versions worked on 5150, but failed on a customer's AT&T 80186? machine. That was when I learned about the change in the size of the prefetch buffer. The fact that it had worked on 5150 helped the customer to understand that it wasn't "BAD software", but that it needed some tweaks to work with his "non-standard" machine. And fortunately easily remedied (an otherwise unnecessary JMP to flush the buffer) before the 5170 came out.
-- Grumpy Ol' Fred ci...@xenosoft.com
