"Home truths" about computing

by Sean McGrath

March 5, 2007 —

This week, for a complete change of pace, I have gathered together some of the cynical/rueful thoughts that I accumulate on my travels around the IT business. They are in no particular order.

- The number of utterly impossible things that computers are expected to do automatically, doubles every 18 months.

- For every problem, there is an apparent solution that is simple, elegant and wrong. Unfortunately, these tend to get implemented before they are known to be wrong. When this happens, it is common practice to simply change the definition of the problem.

- If it takes one resource N days to complete a task, two resources can complete the same task in 2N days. This involves N days for the task and N more days agreeing which one of them has time in their schedule to spend N days on the task.

- Software cannot be proven correct but it can be proven to be over budget.

- An enterprise application is correct if and only if flaws in it are rectified faster than they can be escalated to the project board.

- There are only two hard problems in computer science: 1) distributed memory cache invalidation and 2) explaining the limits of computer science to those who believe computers can do anything.

- The easiest way to detect subtle flaws in computer programs is to spend time analyzing the problem and then go get a pizza. The answer normally manifests itself afterwards. Nobody knows why this works.

- Every computing design problem is solved by the following three step process: (1) add a layer of abstraction so that everything becomes a simple set of interlinked boxes on a whiteboard (b) remove that layer of abstraction when it becomes obvious that nobody follows your logic in adding it (c) design an ugly but functional solution, telling yourself that one day, you will revisit the design and make it beautiful.

- We will soon reach the inflection point in the history of computing where computers are better than humans at misunderstanding English.

- Intractable problems can be made emotionally tractable as follows: (1) decompose the problem into really small pieces (2) lose sight of the overall problem to be solved (3) spend all of your time fussing over the minutia of the individual little pieces.

- Measure twice, then send a memo suggesting that somebody else should really get in there fast in order to cut once.

- There are only four differences between an Executive Information System and an ordinary information system.a) The executive model costs more (b) it does less (c) it has larger icons and (d) nobody knows if it is being actively used or not.

ITworld.com