Re: more on gnutella

[Karen Coyle <karen.coyle@ucop.edu>]

At 04:58 PM 5/26/00 -0400, Eric Lease Morgan wrote:
>I'm not so sure people "need" librarians here, but the precision/recall
>ratio of gnutella searches would definitely be improved if queries were
>applied to more than just file names and/or the protocol supported more
>structured queries. Put another way, gnutella *might* just be another way to
>increase information overload, but I might be blind, again.

When I give lectures to computer science classes about the Net and
information organization, I used to use Archie as an example -- basically a
crude gnutella that allowed you to find files on the Net by their file
name. I likened it to the level of information retrieval in medieval
monasteries -- an unordered title list. This leads to an easy discussion of
how such a thing doesn't scale when you go beyond a small number of items,
and how it requires users to have special knowledge of those items in order
to retrieve them.

Now I'm going to incorporate gnutella and Napster into my talks,
illustrating the same thing (since most university students today have no
idea what Archie was). Essentially, it's a highly automated version of a
nearly useless information retrieval method based on users knowing the
existence of and names of their target. It's a phone book, not an
information retrieval system. If you want an information retrieval system,
you'll need something much more sophisticated. And probably some librarians.


----------------------------------------------
Karen Coyle                    karen.coyle@ucop.edu
  University of California Digital Library
  http://www.kcoyle.net        510/987-0567
----------------------------------------------