Posting to Autocat
Classified catalogs can be very powerful, but they always had to have supplementary indexes. A quick Google Books search provides several examples, e.g. http://books.google.com/books?id=zpAIAAAAQAAJ "Catalogue of the Liverpool Library" (James Smith, 1814), the copy from the Bodleian.
The main descriptions (very light) are arranged in a classified manner, and you browse everything under the main heading, e.g. I am looking on p. 216: Education, with the individual books arranged in an alphabetical arrangement by a type of catchword title, beginning with either the possessive form of the author's last name, when (I suspect) the main title is more or less copied as on the book, as for instance "Chesterfield's letters to his son", but when the catchword reflects the title less perfectly, they change it to: "Cornish on classical learning", which may not match the item of the original very closely at all, but of course, this would take time to check to see if my supposition is correct. On p. 337, you have an actual author index (supplemented by anonymous authors). Finally, there is a list of the classes on p. 46. These supplements are critical.
But that was not the only way to do it, and if you look at the "Catalogue of the San Francisco Mercantile Library" (1854) http://books.google.com/books?id=4pkQAAAAIAAJ, you will see precisely the opposite arrangement: the main descriptions (more substantial) of the books here are arranged under author (main entry), while the "Analytical Catalogue" (i.e. the classified arrangement) starts on p. 137.
With card catalogs, things changed a bit again. In the U.S., libraries tended to arrange their cards in alphabetical arrangements because they were seen to be serving the common people more than an educated elite. Therefore, a regular person may be interested in Dogs but not have a clue where to begin browsing for it in a taxonomic arrangement. So, they put it under "D", while the arrangement of books themselves served as the classified arrangement for the public, with the shelflist--which was mostly off-limits to the public--serving as the old classed catalog. Many European libraries opted for classified catalogs with indexes. A lot of the reason for this was that most were closed stacks anyway. Both methods had their advantages and disadvantages.
While all this may be interesting historically, that's what it all is and where it should remain. These historical considerations should be there to help us imagine new possibilities but not to limit us in any way. I am sure there were many other arrangements of the bibliographic descriptions out there just waiting to be found, and they should be brought to light, so that they could help us imagine even more possibilities.
With today's tools, there are many, many ways of arranging, resorting, redoing everything, so multiple arrangements of the same materials are possible. So, in a correctly configured system, you could arrange the same materials by Dewey arrangement, LC, Bliss, Colon, UDC, any arrangement the patron would want, while new arrangements could probably even be made on the fly. For example, the Mendeley software (free! http://www.mendeley.com/) will take the documents on your topics, do a semantic analysis of them, and do the searches for you, along with social networking and other incredible things. It's not so great yet, I don't believe, but it does have promise.
I have gone back and forth over the value of classification for online resources. Currently, I think it could be fabulously useful. While some out there would probably argue that classification is old hat now, I think that if it is to be useful, it could (and must) be taken to a new level using these powerful tools. How something like Mendeley and DDC, LCC, UDC, Bliss, etc. numbers would work, I don't know, but I think it could improve matters for everybody.