Joel, of Joel on software wrote an interesting blog (Making Wrong Code Look Wrong) last week that I just found time to read. I would largely agree with him. Really quick summary: he argues for variable prefixes to prevent accidentally doing the wrong thing with variables that are the correct 'type' (example: string), but not the correct 'kind' (example: trusted string vs. untrusted string). 

I would largely agree with Joel. However, I would like to make one critique – where possible, I would prefer 'types' over 'kinds'. Joel starts out with an example of using 'kinds' to prevent cross site scripting (XSS) attacks. It sounds like a good programming practice, but I think it would be even better to create two classes: 'UserString' and 'ResponseString'.

What I like about classes:

  • You can make the compiler do the work of verifying safety. It is pretty easy to use the wrong variable. Compilers do a better job of finding problems then I could.
  • Easier for the guy down the hall. In order for 'kinds' to be practical, they need to be abbreviations. This means that when the guy down the hall reads my code, he probably won't understand all of my 'kinds'. Hopefully the guy down the hall is just reading my code. If I decided to go join a different team, all of my 'kinds' might loose their value forever. This is not the true for classes.
  • Easier to review. Using Joel's example, come security review time, I would need to look through my entire code base to make sure that I did use my kinds correctly. If I used types, I would just have one file to review.

Anyway, that is my two cents.