Subject: Re: Knowledge classification systems (Was: Re: becoming a better programmer)
From: Erik Naggum <>
Date: 17 Sep 2002 13:39:29 +0000
Newsgroups: comp.lang.lisp
Message-ID: <>

* Frank A. Adrian
| You've mentioned the Dewey Decimal System.  I assume from the above that you
| believe that other classifications would also be OK (Prytz, Library of
| Congress, etc.) as long as the classification was universal and hierarchical
| or do you also prefer numberical, as opposed to alphanumerical?

  I cannot find any useful information on Prytz, and I have not been able to
  figure out the Library of Congress system.  I am comfortable with Dewey's
  and the Universal decimal classification system, probably because I remember
  numbers accurately without effort, while alphanumeric strings are lossy.
  Others I have talked to about this over the years see higher error rates in
  alphanumeric, including an unexpectedly interesting conversation with the
  designer of postal codes in Norway (which introduced them in 1968) and an
  equally unexpectedly entertaining conversation with the designer of the
  Norwegian social security number (introduced in 1964) -- I fear that such
  people are seldom recognized for their achievements.  So I actually favor a
  system that is as simple as possible, possibly using coded representation of
  extraneous features.  Dewey reserves 0 in subclasses and codes such things a
  geography, special subcategorization, etc, with 0 followed by additional
  digits.  Thus the total number is a simple string of digits.  UDC uses a
  more complex syntax which I personally find harder to remember.

| I also find the ideas of a universal classification system useful, but I
| have some things that bother me about the use of these systems when
| searching for materials.
| First, hierarchical classifications are fine for finding books within a
| given subject, but give no indication as to level.

  There are codes available for this in most of the existing systems.

| Second, I wonder about stability of item location within the classification
| system.

  This has become a more significant issue.  If you look for the Library of
  Congress Cataloging-in-Publication Data of most U.S. imprints, you will find
  the Dewey code after the LC classification (a.k.a. call number) and year and
  before the LCCN (control number).  

    Book                                    Ed  Year Dewew CIP      LOC
    Common Lisp the Language                    1984 001.64'24      (dc19)
                                            2nd 1990 005.13'3       (dc20)
    Winston & Horn: Lisp                    2nd 1984 001.64'24      (dc19)
                                            3rd 1989 005.13'3       (dc19)
    James Gleick: Chaos                         1988 003--dc19
    Structure and Interpretation                1985 001.64'24      (dc19)
      of Computer Programs                  2nd 1996 005.13'3--dc20
    Introduction to Algorithms                  1989 005.1--dc20
                                            2nd 2001 005.1--dc21

  The addition of the DDC edition in the Dewey code started with DDC 19, but
  was recommended with DDC 20 and mandatory with DDC 21.  DDC 22 is due out
  next year.  We see that programming languages moved from DDC 19 to DDC 20,
  with at least one book published before DDC 20.  (I included the last two
  items just to show that things also do not move. :)

| I remember finding a 1960's technical report on SNOBOL from Bell Labs back
| in 1972 and thinking that this was already "historical material". 

  But while the material itself is historical, the subject matter is not.  The
  crucial difference between here is in the intent of the publication.

| If I found it today, I'm pretty sure it would be better classified under
| history of computing and not programming.

  That would require an arbitrary decision as to "currentness" of the
  material, which is explicitly excluded from consideration in classification.

| In subclassifications dealing with rapidly changing topics, I could see this
| becoming an issue.

  Strictly speaking, topics do not change, but new subclasses are frequently
  added and guidance on classification is published by the OCLC.  E.g., I was
  happy to find "Semantic Web" classififed there as 025.04 for me when I did
  not find an obvious classification for it.  The world-wide agreement on the
  application of and changes to the Dewey classification is a fascinating
  thing it itself.

| I guess I'm wondering if any single ontological system really works that
| well, or if a linearization of a multi-dimensional or networked system might
| be better.  In any case, I am, of course, thankful for what we currently
| have!

  Many are surprised to find that DDC and UDC are multidimensional.

| Finally, just a note of curiosity.  While looking for information on library
| classification systems, I noticed that Finland uses a Dewey derivative
| called YKL.  Does Norway also have its own derivative for its libraries or
| does it follow Dewey's system?

  Just this week, the Norwegian Library Association published its 5th edition,
  a translation and adaptation of the 21st edition of DDC (i.e., including a
  number of pre-built numbers for typical Norwegian uses and downplaying
  classes that are more fine-grained in the U.S., particulary the 800 class
  (literature)).  We have been lagging behind the U.S. development and only
  now prepare to reclassify public administration (350-354), education (370)
  and life sciences (560-590) according to DDC 21.  The Oslo public library
  system (Deichmanske), being as it was the first library in Norway to adapt
  and apply Dewey's classification in 1898, have been a leading force in the
  application of Dewey in the Nordic countries and have resisted the
  temptation to make the kind of renegade changes that plagued many other
  Norwegian public libraries a few decades ago.

Erik Naggum, Oslo, Norway

Act from reason, and failure makes you rethink and study harder.
Act from faith, and failure makes you blame someone and push harder.