As the one whose comment triggered this question, I figure I ought to answer: one of the foundational principles of science is that it should be freely and publicly available.
At the same time, there are a lot of human interests that push in the opposite direction. Some examples:
A nation may want to restrict high-technology in order to promote its own interests. For historical examples, consider the British empire's secrecy around timekeeping for navigation, or Bavarian secrecy on methods for making high quality optical glass.
Information may be considered dangerous to release to the general public, such as regarding atomic weapons or the DNA sequences of deadly pathogens.
Commercial companies invest in technology in order to gain advantage over their competitors.
A scientist may want to avoid publishing patentable research until after the patent is filed.
A lot of science that is done is thus never openly published, or openly published only long after it has been completed. The question then is, how should universities relate to this, particularly regarding dissertations?
To the best of my knowledge, in all of the high-ranked U.S. universities, a Ph.D. dissertation is required to be entirely public, as a matter of scientific principle and integrity. This wasn't always the case, particularly during the convergence of scientific research and military funding around World War II. As the country became more uncomfortable with that association, however, the elite universities began to remove classified research from their campuses and require that theses be publishable. In many cases, classified research still goes on in association, but through a separate entity, such as Lincoln Lab for MIT, SRI for Stanford, and LBNL for Berkeley. Likewise, sensitivities have developed around commercial research.
The general principle that is followed then, at least for elite U.S. universities, is that the research leading to a dissertation may involve unpublished or restricted information. The dissertation, however, must be public and substantial enough to stand on its own without depending on other non-public research that may have been done in association.