So yes, if you wish to insist that evolution is true then you need to show how you can alter a program at random and make it better and more complex. -me- Microkernel design for the operating system and a plug-in API for the program.
You cannot get away from design. You are pushing it back a step (if what you say is true, which I doubt and would like to see you back it up, but I am sure you will not) but you are essentially verifying my statement. It takes design, not random chance to alter a program.
I wasn't trying to prove that it works purely randomly. More that the program can adapt to a new circumstance which is what evolution is about. You are right that a program cannot be "written randomly." The idea behind the microkernel design is that only core I/O is built into the kernel and all other functionality is written into servers that are dynamically loaded and unloaded as the kernel needs them. BeOS is one of the best examples of a microkernel design. I don't believe that the donh was trying to imply that a program can be written randomly. The closest thing on the programming side that could fit into that category is polymorphism which allows several functions/methods of the same name/return type to take different parameters. C++ example:
int MyClass::buildDate(int hours)
int MyClass::buildDate(int hours, int days)
int MyClass::buildDate(int hours, int days, int weeks)
And polymorphism has to be used intelligently so in a way it is both seemingly random and intentional. You use which ever method you want to use and the compiler intelligently discerns which one you used during compilation. In the end, whoever tried to use software development as an example to support evolution or even to refute it must have been smoking crack because it can go half way on both sides.