Current jruby/extract.rb in trunk uses a regex on _FILE_ to determine the path to the jruby-complete.jar, and then opens it with java.util.zip.ZipFile, but this only works when jruby-complete.jar is run from a file: url, ie not when loaded over any sort of http connection. New method uses ClassLoader.getResource to determine the url of the JAR file, and then uses java.net.URL.openConnection.jar_file to open a JarFile descendant of ZipFile. It only copies jruby-complete.jar to jruby.home/lib if it was found at a file: url. Patch modifies trunk/jruby/src/jruby/extract.rb