Agentless Backup is Not a Myth

D'oh! There is a (probably apocryphal) entry in a naval officer's fitness report that reads: "This officer never makes the same mistake twice. However, he appears to be attempting to make them all once."

As developers we should try to avoid collecting the full set, but making some mistakes are inevitable; the trick is to learn from the ones you do make, and also from the mistakes of others.

Much of my early experience was gained by listening to the coffee room (water cooler, whatever) chat of more experienced developers; some of the best came from Dr John Rushforth. John had established the first computing lab at the University of Dundee in 1965 so his development experience went back to machine code and paper tape. As well as being a great developer, John was also one of the kindest and most generous people I have ever met.

One day (back in the time of mainframes) he received a phone call from a user who reported that the characters on his new, dumb terminal were appearing fine, were making perfect sense and were perfectly readable: the only problem was that they were upside down.

John was completely mystified. Dumb terminals in those days were really dumb, even the character maps were held centrally; so the fault had to lie with the mainframe. Yet John was looking at a similar terminal and all was fine. He took the user through various processes (opening a file, using an editor and so on) but nothing made the slightest difference.

He was stumped. What could it be? Both he and the user were connected to the same mainframe using the same kind of dumb terminal then the characters simply had to be the same way up relative to the screen.

Suddenly the light went on. John asked the user to describe the top of the terminal.

"Err... it's just a flat, gray metal surface, with four little bits of rubber, one in each corner," came the reply.

Some terminals in those days had very simple cases which had four identical sides, apart, of course, from the rubber feet on the bottom surface. As soon as the user had turned the terminal the right way up, the problem was solved.

March of characters

John told this story, not to mock the user, but to illustrate the wider point about how this kind of confusion arises. Users often supply only the information they consider important. We, in turn, apply a massive amount of context to the information they provide. In other words, we make up a great deal of the missing information.

Mostly that works fine; indeed it's a very efficient way of communicating. We don't want a half-hour monologue from the user starting with their keyboard along the lines of: "The W character is to the right of Q. Next comes E, followed closely by R..."

In this case, the user was actually looking at characters appearing in the bottom right-hand corner of the screen and marching leftwards as he typed - but he thought the most important information was: "They're upside down". John's mental picture was of upside down characters in the top left, marching rightwards - but he didn't ask the user to confirm that.

John reckoned he learned a valuable lesson from that: when a problem proves intractable, it's time to wonder whether we are making incorrect assumptions. He shared his bad, we all benefited.®

Regcast training : Hyper-V 3.0, VM high availability and disaster recovery