I'm pleased that people find my indexes useful and worth discussing, although it's a little disconcerting that nobody thought to ask me the answers. I met a young student recently who had been using them and she said, 'Oh, I presumed you were dead', so that might explain it. I can say categorically that Roud numbers have no meaning in themselves - whenever I encounter a new song it gets the next number. It stands to reason that I encountered the common songs first, but the order in which they appeared is sheer chance. This really is unavoidable in an open system like mine. Laws could give things meaningful numbers because he first gathered all the songs he wanted to include, then sorted and numbered them.
As Steve G and others have commented, my indexes are not designed as classification schemes, but simply a way of finding things.
I'm acutely aware that we really need a subject-index approach and would dearly love to provide one. I have several times got half such a scheme devised but have always had to lay it aside for other more pressing projects, including keeping the indexes growing. But I'm not the only one interested in such an approach, and I flatter myself that such a scheme is now more feasible because my indexes will help the compiler gather things together more easily. In the meantime, a judicious use of keywords in the TITLE, FIRST LINE and SUBJECT fields will often yield good results. If you're interested in sea songs,for example, many of them will include the word 'sea' or 'ship' or 'sailor' somewhere in those fields.
Regards to all
Steve Roud