> [ Re: using Dewey Decimal system to classify Web pages ]
> This suggestion is a bit off the mark, for a number of reasons. I work
> in a library and appreciate that what you propose, Jon would solve a lot
> of problems. Attempts in the past to get some agreement of what
> should be used for indexing and searching have not been very successful,
> mainly I suspect because the requirement has been vague and unwieldy and
> tried to encompass too much.
> There is a group of librarians in the US somewhere (I can't recall off hand)
> who have put together a list of meta data called the _Dublin Core_. It's
> just too complicated for ordinary, non librarians to handle, and even
> they need a reality check as they seem to want to re-invent marc.
> On the other hand I do not think it too much for a HTML author to
> identify and index/keyword his/her document - pretty much the same as
> printed material. Keep it simple so that authors who only need to learn
> simple HTML can management to have their documents located more easily on
> the WWW.
Here's a thought: how about using the Usenet newsgroup hierarchy
as a classification scheme? I.e., "if this Web page were a Usenet
article, in which newsgroup(s) would it belong?"
The Usenet hierarchy is fairly well-organized, and likely to be more
familiar to the average Web author/user than the Dewey system would be.
The newsgroup hierarchy is also a better classification scheme for
much of the stuff that's on the Web. For example, it would be much easier
to locate Tom Christiansen's "Perl Data Structures Cookbook" 
if it were classified under "comp.lang.perl" than if it were under
"Computers and Data Processing" (or whatever the Dewey category is)
along with the host of other "Computers and Data Processing"-related
material that's out there now.