[bsder]

> Why? Because unless you're working at MIT or CERN that's how labs in the real world work. Everything's broken and you don't have any funding but you still have to publish.

That's close, but not quite the real issue.

The lesson is to document everything. Repeatability IS an issue even if no one is going to repeat the experiment--because YOU need that information to make any progress. It allows you to isolate to one thing, change one thing, and see what the result is. And THAT is the core of debugging.

I wish I could send every programmer through a physics lab class. The number of times I have had the following conversation with software people is staggering:

"Well, I did X and it didn't work."

"Oh? So, what was it supposed to do before you did X and what was it supposed to do afterword?"

"I dunno."

"THEN WHY THE HELL DID YOU DO X AT ALL?" <bangs head on table>