🚀 go-pugleaf

In article <2SEP199209361066@rover.uchicago.edu>, frank@rover.uchicago.edu (Frank - Hardware Hacker - Borger) writes:
> 	Mercifully I'm not generally involved at this now, but I once was,
> 	in the early days of RSX on 11s. I remember one series of patches.
> 
> 	Patch I		Reported to fix a problem that odd parity did not work
> 			right on a single line DC interface.
> 
> 	Patch II	Fixed the problem that patch I fixed odd parity, but
> 			broke even parity.
> 
> 	Patch III	Fixed the problem that patch II fixed even parity, but
> 			broke no parity.

Do you REALLY think this is any better without PATCH? A year or two after
designing the code update distribution scheme I mentioned (100% patch-free),
the vendor I mentioned had to invent a whole new type of distribution tape,
available only for one particular release of one of their major products.
Reason: there had been so many fixes to the fix that broke the fix that fixed
the fix that correct the problem (...) that installing the product on one of
the smaller CPU's (say something the speed of a MicroVAX 3600) took ONE WEEK OF
CPU TIME, to check that everything was in order (all pre-requisites applied,
all fixed to depending products also applied, etc). Each individual
sub-component had thousands of fixes; the single most buggy component was the
software update management system itself, which you had to fix before you could
use it to fix other programs.

So they got this great idea: they would get an employee to play the one-week
game on a freshly installed machine, and then they would dump the disks to
tape, and ship these tapes to customers who would just reload the data to the
corresponding directories. That looks like directly patching the software
update management system database if you ask me, but customers were DELIGHTED
:-)

  Eric