On another note, D is a very cool language. Thanks for your work there.
I suspect with physical books, knowing typical patron categorizations and interest groupings could aid browsing (and accessing all related materials quickly).
Public libraries are also somewhat constrained by desiring a single organizing system so employees can transfer to different branches with less retraining (long term or as substitutes). However, even within a U.S. state's county, patrons' interests can vary.
Also, tagging seems insufficient as different tags have different strengths. Perhaps something more like a word cloud might be useful?
The Melvil (Dewey is proprietary, I think) system is far from perfect, but any weighing of the tradeoffs will disadvantage some use case. Knowing your users helps as does training them and having experts (librarians).
(I am a public library page (shelver), so I regularly encounter the quirks of Melvil/Dewey and what I view as cataloging shortcomings where the subject of secondary significance is used, like a televison show's cookbook being cataloged under television shows, correct but not likely to circulare as well, (and a few errors such as a volcano book cataloged as 551.31 instead of 551.21).)