You could store a 0 or 1 value in a database, and have the value change every time the code executes... but that seems like a lot of resources to use up just to be in control over something like that.
Over a wide sample, random values should equal 50/50 odds.
Why do you need 50/50, exactly?
Also, I'd rather have the code cycle through an array of header values, this way you can easily add more headers later.
