Preservation of correspondence

Warren Wolfe wizard at voyager.net
Mon Feb 12 01:53:45 CST 2007


On Sun, 2007-02-11 at 17:43 -0600, Jules Richardson wrote:


> Can't tell you the answer to that one, but it's an interesting question and 
> one I've considered for various equipment that has built-in error correction; 
> it's no good having that error correction if it silently masks errors 
> happening when the user could have otherwise done something about the faulty part.
> 
> Memory (at least in ECC form) and hard disks are two obvious candidates for 
> masking faults from the user...


    This takes me back to my IMSAI 8080...  When I got it running
properly, I ended up abandoning games like "Kill the Bit" quickly, and
didn't use the LED output port for much.  I noticed that in the BIOS
routine for disk access, a sector read would be tried over if it failed,
up to a total of ten times.  If a read failed ten times, a disk read
error would be returned.  I decided that I'd like to know earlier than
that about any media problems I might have.  So, I used a byte of
memory, and every time I got an error, before it called the routine
again, I would increment the byte counter (paying no attention to
overflow) and send it out to the LED display.  If a program was running
without complaint, but the LEDs were cycling regularly, it was time to
examine the disk, and perhaps copy off the data.  I was smug about that,
and about the routine that would thoroughly test one byte of memory with
each call to the console status routine, where the computer spent its
idle time.  Might as well be testing memory, one location after
another...  Memory was notoriously flaky, and catching it sooner rather
than later was a good thing.



            Peace,

            Warren E. Wolfe
            wizard at voyager.net





More information about the cctech mailing list