We had a system to censor users' IMs. Customers could configure a forbidden word list, or use a default one that had some seemingly obvious swear words. That was fine, until you realize that a global customer base has at least one person with almost any imaginable word in their names.
The North Koreans even made a phallic missile called No-dong. There is just no way to apply English assumptions about what words are across names. Oh, and the South Koreans have a town called Nodong-dong (http://www.agoda.com/city/nodong-dong-kr.html).
In Norway we have a man named "Jo Å" that lives in a city named Å. I often uses that combinations for testing it-systems and gets a lot of odd results.
Cool link. There are some real gems there. Especially translating dirty words to "clean" ones seams especially dangerous.
Apparently if the sprinter Tyson Gay had been assassinated in 2008 the American Family Association filtered would have translated it to "Tyson Homosexual has been buttbuttinated" (replacing gay->homosexual and ass->butt).
A story from my wife's first marriage, specifically from when she went back with her Harvard PhD husband to his homeland of Turkey:
Harvard diplomas are of course signed by the university president, who for many years was Derek Curtis Bok. "Bok", in Turkish, means "shit". Turkish bureaucracy evidently requires (or used to require) showing your diploma more often than one might think.
The issue is that actual null values are stringified as "Null". It doesn't take much for someone to decide to just convert all string instances of "Null" back to the null value, despite that leading to issues.
Caterina Fake, founder of Flickr, tells a story about one particular airline she just can't fly on. The system lets her buy tickets just fine, but when she shows up to fly, they've never heard of her.
I guess that's one way to solve the test data vs real data problem.
I "almost", AKA thought of it but didn't tell my wife, made my son's middle name Null. It even entered my mind when I was handed the Birth Information card. I knew it was funny but not worth the eventual price I would pay when I got caught.
We have static typing and dynamic typing and duck typing... is there a name for this crap where stuff gets converted willy-nilly without your say-so? Drunk typing?
Does SOAP (or WSDL?) or whatever he is using not store type information with the value? Take JSON, which has null (the actual value) and "Null" (a string) - it isn't ambiguous which you mean.
I feel like if my serialization format couldn't differentiate between those, that'd be a deal-breaker.
We had a similar problem with an old internal IRC bot written in TCL. Someone joined the company with the initials 'nan' (which was never equal to 'nan' because NaN != NaN). The bot never worked for them.
I graduated high school with a kid whose last name was Null. It would occasionally cause issues with the school's computers. He learned to use it to his advantage.
The North Koreans even made a phallic missile called No-dong. There is just no way to apply English assumptions about what words are across names. Oh, and the South Koreans have a town called Nodong-dong (http://www.agoda.com/city/nodong-dong-kr.html).