Backward compatibility in astrophysics

Computer science is well known for having all sorts of backward compatibility issues. The reason is that users are more likely to use a new program (or a new release of a program) if their existing code will still run on the new version. It often leads to unnecessary overhead and occasionally to somewhat unpleasant code gymnastics to ensure that older functionality will continue to work in newer versions of a product.

Backward compatibility shows up in other fields, too. One great example of this is terminology. A researcher will have an interesting idea and give it a name, mostly just to be able to discuss it. Then the name starts to creep into the literature on the subject, since it's the best way to describe the idea. Before you know it, the name is everywhere, even though that wasn't the researcher's original idea. In some cases, this works fine. For instance, "quark" is as good a name as any for the subcomponent particle of hadrons. In some cases, though, a name doesn't quite fit. Color, for instance, is a three-valued "charge" that is often confusing to people learning about the strong force and nuclear interactions.

There are (at least) two interesting examples of this phenomenon in astrophysics: spectral types and galaxy classifications.

Once upon a time, in the late 1800's, scientists were scrambling to classify stars by their spectra. One classification scheme (there were plenty) used the letters A-N to describe the strength of the H-alpha line, a prominent spectral line produced by hydrogen. Later on, these classifications were reorganized to reflect the temperature of stellar surfaces, which provides a much more useful framework for astrophysical organization. This leaves us with the wonderful spectral types: OBAFGKMN... not exactly alphabetical order.
From O to N-type stars, surface temperature decreases, and the stars grow redder. The reason that the temperature of the star isn't exactly reflected in the strength of the hydrogen lines is that different temperatures produce different ionization states in hydrogen. Clearly, if a hydrogen atom is ionized, its electron can't change energy levels and thus emit the characteristic light, because it doesn't have an electron. Similarly, only singly-ionized calcium emits the characteristic K spectral line, so the K line doesn't appear in stars too cold to ionize calcium, or stars hot enough to doubly ionize it.

On a different note, galaxies were originally classified according to the Hubble Fork. Edwin Hubble called elliptical galaxies early-type, and spirals (divided into spiral and barred spiral galaxies) late-type. The logic here was that galaxies started as elliptical blobs and evolved into spirals. As it turns out, galactic evolution actually happens the other way around, with spiral galaxies turning into ellipticals in their old age. Also, we have only the fuzziest of ideas about the mechanisms that allowed for the production of so many spiral galaxies. Thin disks should be fairly fragile, so that galaxy mergers produce blobbier structures, but astrophysicists also believe that most existing galaxies formed via mergers. Yet another mystery.