jeroen said:

such things as Dewey or LoC carry historical load we don't need.

agreed. dewey, especially, has a lot of socio-political baggage. i just read yesterday that there are something like... well here it is:

http://www.hyperorg.com/backissues/joho-sep03-04.html#dewey

Of the hundred numbers set aside for topics concerning religion, 88 — numbers 201-287 — are reserved for Christianity. Jews and Moslems get just one each. But those single-digit religions are still doing better than Buddhists (294.3) who share a decimal point with the Sikhs (294.6) and Jains (294.4), looking up enviously at Christian "Parish government & administration" which gets its own whole number (254).

people who are interested in cataloging issues should read more of this site. it's written by one of the cluetrain manifesto people, and it's really quite good. having said all that, a benefit of numbers from existing systems is that they have already been assigned. so the work becomes getting them, not assigning them, and the former is an order of magnitude easier.

We should go for a nice, multi-facetted, multi-entry classification system for our books, but developing adding that would be a tremendous lot of work...

http://onlinebooks.library.upenn.edu/webbin/bparchive?year=2006&

i'm not sure it would have to be all _that_ difficult. certainly a "first pass" could be done rather simply, and might prove to be a huge head-start. and if the system was built as a wiki, incremental improvement from there might be fairly easy to achieve as well... again, the person doing the website above is working on "latent semantic indexing", so he might be a good person to read for ideas... along these lines too, you might wanna look at this: post=2006-04-11,1 -bowerbird