I’ve been doing a lot of interviews lately at work, and one of the usual fields we try to cover is do the candidates know how to use the right tools for the task at hand.  A typical question can be, if you’re given a typical day’s logs, how would you find a particular piece of information (say all the ip addresses of all the customers that visited).

The fast answer is you would use a unix/linux tool called ‘grep’, which allows you to search for a particular (regular) expression in a file.  This is a surprisingly useful tool, and combined with other similar programs, allow you to quickly search through, collate, and aggregate all sorts of information.  It usually takes a couple of minutes to look through 10s of GBs of data.

Some candidates, inevitably, would try to write a program from scratch, that essentially does the same thing as the existing tool.  This is not ideal, but if they can do it right, still acceptable (especially since you should be able to write simplified versions of these tools in a couple of hours).

Imagine my surprise then, when I read this article regarding why it could cost up to $15 million to provide copies of Alaskan state employee emails.  Here’s a choice quote:

How did the cost reach $15 million? Let’s look at a typical request. When the Associated Press asked for all state e-mails sent to the governor’s husband, Todd Palin, her office said it would take up to six hours of a programmer’s time to assemble the e-mail of just a single state employee, then another two hours for “security” checks, and finally five hours to search the e-mail for whatever word or topic the requestor is seeking. At $73.87 an hour, that’s $960.31 for a single e-mail account. And there are 16,000 full-time state employees. The cost quoted to the AP: $15,364,960.

Emphasis mine.  Obviously, the programmers that work for the state of Alaska would not have passed our phone screens.

Also worth noting is that the costs are in aggregate – which means you cannot create the process once, and automate to repeat the same process at a cheaper cost (which is what computers are really good at).

Seriously.