I recently hit an odd accident that turned into a cautionary tale about test data.

Someone was testing one part of a new ecommerce system, at precisely the same time someone else was testing another part. No-one noticed until DHL returned a package that was addressed to the Death Star.

They still charged us shipping, but atleast we didn't have to swallow product cost too. But it did make me look at "random test data" quite differently.

Snap - I use [email protected] with the Pentagon's address.

In the interest of client-friendliness I also use Disney addresses, with names such as Ms. Sleeping Beauty. So with my test data stuff would get delivered by DHL to fictional people in these places with people signing parcels.

I use real addresses so I can check postage rates are correct. It would be really useful for me if there was a standardised list of test addresses used across industry and in the media. Doubly so if you had test code that worked on such a known list of test cases, adding address records for 'customers' without having to enter one's own test data.

Why would you put garbage in your production databases?

It wasn't supposed to be production. From what I gather, someone made sure the queue was empty before witching their component to the live system temporarily. Someone else managed to get a test order in, in the seconds between the check and the switch.