the studied users shared an average of about 350 files, and that only 13% shared no files.
riding rate derived from the March 2005 dataset used in Deviant Behavior. The vast discrepancy in these
results may result from some fundamental, but as yet unidentified, change in the programs themselves.
Nevertheless, the different data-collection methods used in Deviant Behavior and Characterizing could
explain some or even most of the differences in user's sharing behavior. As Characterizing notes, its data-
collection method would work only if a particular user 1) was connected to the network for a relatively
long time; 2) was not firewalled; and 3) had not disabled the browse-host feature. In practice, this method
worked only 18.5% of the time.
of Gnutella users generally but the behavior of the two disparate subgroups of users who would be likely
to be running an unfirewalled, browse-host enabled filesharing program for relatively long periods. One
subgroup might consist of highly unsophisticated users who were using browse-host-enabled filesharing
programs without a firewall. The other subgroup might consist of sophisticated "true-believers" in
filesharing who had both the expertise and the motivation needed to configure their firewall in order to give
a filesharing program unrestricted access to the Internet. See, e.g., BearShare, Gnutella Good Citizen Tips
at http://www.bearshare.com/help/citizen.htm (last visited June 19, 2006) ("You don't need to get rid of
your firewall completely, you just need to "drill a hole" in it for BearShare. It won't decrease your security
because BearShare doesn't contain any security holes.") Both groups would be very likely to be sharing
files, and in significant numbers, though probably for very different reasons.
suggest that more remains to be learned about the effects that program design and legal enforcement have
upon users' propensity to share files.
parties other than the distributor of the filesharing program. So used, the term "decentralized" has a legal
rather than technical meaning: Napster, Inc., could thus have converted its "centralized" filesharing
network into a "decentralized" filesharing network just by giving the computers that housed its search-
index servers to third parties. See Edward Felten, "Centralized" Sites Not So Centralized After All,
F
computers of users who have high-speed Internet access. See supra note. 66. Reports also indicate that
these programs now, whenever possible, thwart the actual peer-to-peer file transfers that once occurred
over the Napster, Inc. network: By default, these programs will redirect a user's request to download a file
from another "peer" user to a specialized, high-speed, terabyte-sized fileserver that exists solely to store
and transfer files "shared" over filesharing networks. Programs use this fileserver-based architecture by
default because "downloads ... are faster": "[E]nd-users typically experience a net acceleration effect of
2x--4x." Joltid, Benefits and Recent Statistics, http://www.joltid.com/index.php/peercache/
benefits_and_recent_statistics (last visited March 1, 2005) (available at http://web.archive.org
/web/20041027021141/http://www.joltid.com/index.php/peercache/ benefits _and_recent_statistics). For
example, the owner of the FastTrack protocol and the KaZaA filesharing program warns users that
disabling use of these fileservers and actually downloading files from peers "will most likely slow down
downloads dramatically." Id. at http://www.joltid.com/index.php/peercache/faq/enduser (last visited March