Errata

The statement

node.setNext(null);

is unnecessary, I think, since the current head node is being dereferenced and
is therefore subject to garbage collection, so whether it references the next
node is a moot question. In the remove method on the next page the reference
to the next node is not nulled out whether the node being removed is a head
node or not.

If I try to compile the Sorter.java Code there's some errors,
concerning the class ComplexNumber (cannot resolve sysmbol)
The same happens to the method magnitude ...

My problem is that this happens with my own written text AND
the downloaded code from this site.
Is this a problem with my own java installation or an error in
the program code?

The popup menu for the applet version of this program does not "pop-up" in the
latest releases of Netscape (v 4.74) and Internet Explorer (v 5.5) browsers.
It does work in the AppletViewer.

The program does not work correctly.

When run on a Windows 98 PC using Java 1.3, HardcopyWriter$Demo and
HardcopyWriter$PrintFile only print out the line at the top of the page. The
code was downloaded from your website. The printer used is a Hewlett Packard
Deskjet 600.

Sending email through a URL connection: same sort of problems with println()
as mentioned in errata for [161] (in the "confirmed errors" above). But the
Sun JDK 1.1.6 support for "mailto" seems seriously broken on an NT machine.

The example invisibly sends helo machinename mail from: username@machinename
data which fails for many SMTP servers which don't allow data until an RCPT
from: has been given too.

I got to the bottom of this using a heavily edited SendMail example, the
GenericClient example and by modifying the HttpMirror example to mimic an SMTP
server, and by playing with a telnet session to a real ESMTP server (SendMail
8.8.8/8.8.8) running on a Sun.

Editor's note: This sounds like a bug in the JDK, not in the code in the book.

if (searchType == (SEARCHANY | SEARCHALL)) { requireAll(searchArray); }

MUST BE

if (searchType == (SEARCHALL)) { requireAll(searchArray); }

var refineAllString = allString.substring(0,allString.indexOf('|HTTP'));

MUST BE

var refineAllString = allString.substring(1,allString.indexOf('|HTTP'));

In the book is the example of records.js incorrect. It starts with the url. The
source ends with the url.

I use Java 2 on Windows 2000 and Java 1.1.6 on SunOS 5.6.

You can't remove a service properly. The ServerSocket "listen_socket" remains
on the port ad infinitum. There seem to be two errors:

a) half-way down the page in the Listener constructor, the author passes a
dummy value to setSoTimeout (10 minutes) to allow an interrupt() call to
interrupt the accept() call in the Listener.run() method at the bottom of the
page. I don't think that anything will allow an interrupt() call to affect a
blocking IO call like accept().

This means Listeners never return from Listener.run().

The solution is to use setSoTimeout for real, passing it a low value like 1
second (1000). This causes a lot of polling, and may cause performance
problems. You might as well remove the otiose line "this.interrupt();" in the
Listener.pleaseStop() method as well. All of this is due to the Java thread
model not handling IO very well.

b) Even when Listeners die in a timely fashion, the ServerSocket still
survives. I even tried an explicit call to System.gc(), which had no effect:
it seems that you need to close the Sockets directly, the garbage collector
won't do it for you when the last reference goes out of scope.

Here is the fixed Listener code:

public class Listener extends Thread {
ServerSocket listen_socket; // The socket we listen for connections on
int port; // The port we're listening on
Service service; // The service to provide on that port
boolean stop = false; // Whether we've been asked to stop

/**
* The Listener constructor creates a thread for itself in the specified
* threadgroup. It creates a ServerSocket to listen for connections
* on the specified port. It arranges for the ServerSocket to be
* interruptible, so that services can be removed from the server.
*/
public Listener(ThreadGroup group, int port, Service service)
throws IOException {
super(group, "Listener:" + port);
listen_socket = new ServerSocket(port);
// give the socket a non-zero timeout so accept() can be interrupted
/** @todo: set this timeout value to something reasonable for your
system. */
listen_socket.setSoTimeout(6000);
this.port = port;
this.service = service;
}

/** This is the nice way to get a Listener to stop accepting connections */
public void pleaseStop() {
this.stop = true; // set the stop flag
}

/**
* A Listener is a Thread, and this is its body.
* Wait for connection requests, accept them, and pass the socket on
* to the ConnectionManager object of this server.
*/
public void run() {
while(!stop) { // loop until we're asked to stop.
try {
Socket client = listen_socket.accept();
connectionManager.addConnection(client, service);
}
catch (InterruptedIOException e) {}
catch (IOException e) {log(e);}
}
/** @todo: do we really need to close this socket explicitly? */
try {
listen_socket.close();
}
catch (IOException e) {
log(e);
System.err.println("Service removed, but couldn't close server
socket.");
}
}
}

The code sample below the comment "Check through the list of connections, removing
dead ones" traverses the "connections" vector in the forward direction, removing
"dead" connections from the vector as it goes:
for(int i = 0; i < connections.size(); i++)

This does not work properly, as the size of the vector and the indices of the
elements are changed in the body of the loop.
This should be changed to:
for(int i = connections.size() - 1; i >= 0; i--)

Also note that the "Connection" threads don't seem to cooperate well (try doing
something non-trivial in a service), at least with my configuration - Windows 2000
Pro, JDK 1.4.2 - even with the addition of calls to Thread.yield() sprinkled
liberally throughout.
I have been able to solve this by setting the connection threads to low priority.