I know, not exactly a new observation. But I was thinking it again this morning, while reading a historical survey on the topic of homosexuality and civilization (see booklist to the right).
The book's a survey, of course, and a secondary source by definition, but it draws on a lot of primary sources: letters, legislation, Church documents, and especially trial records. One point that comes up over and over again is how many gaps there are in the record, because the primary documents upon which the author must draw to make his case are lost or destroyed. (In the latter case, sometimes deliberately so--and even being able to find out that much is telling.)
At my library, our emphasis is on use. Our budget and our physical facilities are simply too small for us to have the kind of large research collection of, say, the big state university up the road. When I'm weeding the collection, of course I'll keep the classics, as well as the heavily-used materials (not always the same thing, you'll note). And of course the main thing is that the information is available somewhere.
But using recorded knowledge is about more than keeping favorite materials in the library collection. It's also about keeping knowledge as part of the current understanding about the world, its circumstances, and the people in it.
Historians probably think this way all the time, but it's a rather new perspective on the preservation issue for me.