8-bit IDE

From: Eric Smith <eric_at_brouhaha.com>
Date: Tue Apr 18 13:44:43 2000

"Richard Erlacher" <edick_at_idcomm.com> wrote:
> Frankly, the reason I'm exploring this is because with the 8-bit mode, I
> don't have to buffer the data at all beyond the on-board data-in and data
> out buffers at the bus interface.

However, it doesn't matter if 8-bit IDE was standardized (at one time,
but not any more), if there aren't any available disk drives that do it.
It seems much more worthwhile to add the extra latch, buffer, and logic
to make 16-bit IDE work, since then you can use readily available
drives.

This reminds me a little bit of the story of the main memory for the
IBM 704 and 705 computers, as given in the book _IBM's Early Computers_
(regrettably out of print). IBM had previously shipped the 701 and 702
computers using electrostatic (Williams tube) memory, but it was found to
be unreliable, so IBM was in the process of replacing it all with new
core memory.

Meanwhile, the product teams for the replacement computers, the 704 and 705,
decided to use electrostatic storage rather than core because it was cheaper.
The decision was overruled by Tom Watson Jr., who said something to the
effect that choosing something that didn't work because it was cheaper than
something that did was the damndest business decision he'd ever heard of.

Of course, if IBM was structured into separate profit centers for new
computer sales vs. repair, you can see why the managers for new computer
sales would make that decision. Pacific Bell does the same sort of stupid
stuff. Installers will routinely steal working and in-use pairs from one
customer for new service for another. It makes their statistics look good,
and repair service is stuck with fixing the original service, which isn't
the installer's problem.

Eric
Received on Tue Apr 18 2000 - 13:44:43 BST