[PLUG-TALK] The Worst Typo I Ever Made

Keith Lofstrom keithl at kl-ic.com
Wed Jun 10 22:55:57 UTC 2020


On Wed, Jun 10, 2020 at 10:54:21AM -0700, Dick Steffens wrote:
> Interesting YouTube my wife found and shared with me:
> 
> https://www.youtube.com/watch?v=X6NJkWbM1xk
> 
> While I haven't done that one, I probably did something similar years ago.

The video describes using a very powerful SQL command to
make a "slight change" to hundreds of hours of volunteer
work, deleting it all.

Backups.  LOTS of backups.  I've used dirvish disk-to-disk
backups for decades - indeed, when the original author
JW Shultz died, I "maintained" that open source project for
almost two decades (I'm handing over project management to
Mike Beattie in New Zealand now). 

I won't publically discuss how I've arranged my backups
(the bad guys will read this), but a general rule of thumb
is that if you don't have two copies, you will soon have
zero copies.   For important stuff - like any file made by
anyone else, lovingly crafted for me - I prefer FOUR copies
on multiple kinds of media.  I don't always remember to do
that.  After this fiasco, the Youtube presenter might.

What horrified me about the video and the loss of all that
volunteer work was that the SQL command line tool does not:
(1) detect the destruction of large amounts of data, or
(2) tell you what it is going to do before it does it.

Automatically deleting data is almost always a bad idea.
Programs should let you know.

Most of the bytes were probably still somewhere on his hard
disk, but finding and recovering them is a vast amount of
finicky work.  Me, I'd PULL THE PLUG /before/ I thought
about what to do next.  Better a corrupted and single-user-
mode-repairable hard disk image than a program or OS
automatically writing over those "unused" disk sectors.

Pilots and astronauts spend thousands of hours using
simulations to practice dealing with worst case surprises.
Many experience hundreds of horrible simulation "deaths", 
and rapidly cope when actual un-simulated surprises occur.
Wouldn't it be great if similar "data lethal" simulations
were available to train program designers and users?

Keith

-- 
Keith Lofstrom          keithl at keithl.com



More information about the PLUG-talk mailing list