Sunday, August 06, 2006

Is Untested Code the Dark Matter of Software?

According to Cedric Beust's blog, untested code is the dark matter of software. I like this analogy. Just what is dark matter? Wikipedia states:
In cosmology, dark matter refers to matter particles, of unknown composition, that do not emit or reflect enough electromagnetic radiation (light) to be detected directly, but whose presence may be inferred from gravitational effects on visible matter such as stars and galaxies.
I find it interesting that just like untested software, dark matter has observational evidences such as gravitational effects on galactic motion but we really do not fully know what it is. Our calculations detect its presence, but due to obvious limitations, we lack the current ability to conduct extensive empirical investigations. Similarly, in software development, (not because we can't do extensive test but because we won't) we will run limited tests on a module and observe a small scope of expected behaviors. In essence, like the observable effects of dark matter on stellar objects, we infer that the software is coded correctly because of its observable outcomes. However, if not properly tested with the array of possible circumstances the module may not function as planned.

I understand that the limitations for software programming are different than cosmological research. Shear distance and inconclusive understandings of gravitation hinder a fuller understanding of dark matter. With software development, customer expectations, unmitigated risks, and changing requirements can hinder proper developer testing and confidence. However, if we really want to, unlike the astrophysicist, we have the ability to test and empirically understand most of the potential effects of the software we write.

What do you think would become of an astronomer who had the opportunity to more closely research dark matter and did not because of time pressures at best, and laziness at worst? Now that matter is dark.

No comments: