chromatic has come up with a lovely new metaphor to replace the idea of technical debt: “technical insurance”. I love this sentence: “Given the premiums for your team, you then have to decide whether or not to buy a technical insurance policy.”
It hits at something that has bugged me about testing enthusiasts — the notion that you should always have tests. Because there is a balance that needs to be found — the cost of the tests versus the cost of not having the tests.
I’ve talking about this with respect to the ABC module before. Properly testing the
abc2ly script would require constructing an OCR system for reading sheet music. That would be very expensive technical insurance indeed.
And what would it be protecting against?
abc2ly is a open source tool for creating sheet music. It makes no money for me. As far as I know, I am the only active user. The only real risk I can see is if I am printing a long musical document, and a change I’ve made breaks something subtle that only affects, say, one thing on page 14, so I don’t notice the problem in a timely fashion. So basically, there’s a potential for wasting some paper.
In other words, in this case technical insurance would be expensive, and have very limited benefits. I’d be a fool to “buy” it.
On the other hand, the Perl 6 spectests are a fantastic technical insurance policy. For the most part they are low cost (easy to write) AND they are shared across multiple implementations of Perl 6. On the flip side, they help protect against subtle flaws in the implementations, which have the potential of causing problems for every person using Perl 6. We’d have to be crazy to do without this policy.