Low Road Software

In How Buildings Learn, Stewart Brand writes:

Take MIT —the Massachusetts Institute of Technology. A university campus is ideal for comparing building effectiveness because you have a wide variety of buildings serving a limited number of uses—dormitories, laboratories, classrooms, and offices, that’s about it. I’m familiar enough with MIT to know which two buildings are regarded with the most affection among the sixty-eight on campus. One, not surprisingly, is a dormitory called Baker House, designed by Alvar Aalto in 1949. ... But the most loved and legendary building of all at MIT is a surprise: a temporary building left over from World War II without even a name, only a number: Building 20. It is a sprawling 250,000-square-foot three-story wood structure—“The only building on campus you can cut with a saw,” says an admirer—constructed hastily in 1943 for the urgent development of radar and almost immediately slated for demolition.

At the time Brand wrote his book, in 1993, Building 20 “was still in use and still slated for demolition.” Five years later, and 55 years after it was built, it was finally demolished to be replaced by the Ray and Maria Stata Center designed by Frank Gehry.

Brand continues:

Every university has similar stories. Temporary is permanent, and permanent is temporary. Grand, final-solution buildings obsolesce and have to be torn down because they were too overspecified to their original purpose to adapt easily to anything else. Temporary buildings are thrown up quickly and roughly to house temporary projects. Those projects move on soon enough, but they are immediately supplanted by other temporary projects—of which, it turns out, there is an endless supply. The projects flourish in the low-supervision environment, free of turf battles because the turf isn’t worth fighting over. “We did some of our best work in the trailers, didn’t we?” I once heard a Nobel-winning physicist remark. Low Road buildings keep being valuable precisely because they are disposable.

There are many examples of software intended as throwaway prototypes, but that became permanent. Quick and dirty code that has been deployed and on which development has continued – patched, hacked.

But software is different. And still the same.

Final-solution, over-specified software sometimes take so long to build that reality obsolesces it. But deployed prototypes are most often inefficient to adapt and evolve. Code is scattered, duplicated. Trivial changes take time.

But there can be a software equivalent of Building 20. Keeping duplication down doesn’t have to take a lot of time. Just agreeing upon this principle and being disciplined enough to respect it.

The picture of Stata Center is taken by Dan Hill and released under a Creative Commons License. See his post on the new building.