Eventually I ended up removing persistence.xml and orm.xml from the jar file and adding the JPA annotated classes in the jar file to my Webapp persistence.xml. It worked, but Eclipse JPA claims it can’t resolve the class and shows an error on persistence.xml.

The Webapp works fine so in the meantime I won’t bother about the Eclipse error, but obviously one day I will be Googling for a solution.

HOWTO: IPCop-OpenVPN: http://thinkhole.org/wp/2006/03/28/ipcop-openvpn-howto/

HOWTO:Roadwarrior:http://www.zerina.de/zerina/?q=documentation/howto-roadwarrior

Creating the server certificate took ages in my IPCop machine, bearing in mind it’s a Pentium MMX processor 233MHz :-). The certificate generation doesn’t use that much memory, but it’s CPU intensive.

I first thought it would be better to connect using the Vista wizard, but as usual username & password authentication was easy to setup, however the client certificate setup was a fiddly one. I opted to save my time, abandon Vista Wizard and install OpenVPN client.

After installing the OpenVPN client for Windows (the recent openvpn-2.1_rc15 include OpenVPN GUI as well) remember to review the README file as it includes important information for Vista users.

Due to elevation restrictions in Vista, OpenVPN client will fail to add routes to the routing table if not run as an Administrator. Simply right click on the OpenVPN GUI icon and select run as Administrator. Check the status log for any errors.

I managed to test my VPN connection using my mobile broadband after disabling my LAN wireless connection.

I have always been a fan of Red Hat Linux, starting from RHL8,9 and I was there when Red Hat decided to split it into the Fedora project. And I was there when Fedora Core 1 came out, I tried it and tried Fedora Core 3, I adopted Fedora 5 as my server’s OS and I went to tightly secure Fedora 8 as part of my uni security course. I always loved Red Hat and loved their Linux. In-fact I have never tried any other linux distro other than RHL and Fedora (Although I did install IPCop, a linux firewall distro).I downloaded the Fedora 10 DVD image and as usual for some reason it was corrupt and ended up wasting a couple of hours on the installation, which I then abandoned and decided to download the netintall image which worked perfectly, but I had to leave it overnight to download the required packages.

Amazingly there was a large number of Educational programs and kids games, plus hundreds of other packages of useful stuff. My daughter were so happy to play the games, so I left them on the console and I used vnc from my laptop to have my own X desktop and finish off the installation of other stuff. It just makes realize how powerful Linux is. Anyway as usual with any of my Linux installations I had to add the static routes to manage my sub-nets, I ran the commands then put them in /etc/rc.d/rc.local

route add -net 192.168.3.0 netmask 255.255.255.0 gw 192.168.1.5 dev eth0

route add -net 192.168.3.0 netmask 255.255.255.0 gw 192.168.1.5 dev eth1

All in all, performance was good, the installation was easy, all the hardware was recognised automatically, the only problem I encountered was my fault due to my corrupt image, but other than that it an easy install. Definitely Linux has come a long way and specially Fedora and at version 10 it looks awesome.

Using Amazon’s pricing calculator I tried to enter the same specifications offered by a PoundHost A2 dedicated server and to my shock Amazon EC2 monthly pricing was three times that of PoundHost A2 dedicated server. So what I don’t understand is how would a startup be able to afford using a more expensive service such as EC2?, may be it’s the buzz word ‘Cloud’ and lots of cash coming out of VC funding rounds!

Personally I think Amazon is still got a great deal to do in regards to pricing if they want to compete with the dedicated server hosting market.

A few months ago I was excited by the idea when it was announced by Google, and applied to go on the Google app waiting list. Finally when I got an email to join I had a quick look and I lost interest as the service is based on Python and I don’t do Python I know Java quite well and I waited for a Java stack Cloud. Recently I stumbled upon a TechCrunch article about Stax, a new cloud computing service for Java.

Straight away I watched the screen-cast and applied for an account, the following day I got an invitation code and created my first application and database. I’ve selected to create an Struts application, and lucky enough it turned out to be a Struts 2.0 application. The service is only Beta, but looks very promising, the next step I’m going to try to create a fully fledged Struts2, Spring, Hibernate with JPA and Ext application and see how it all hangs together.

I can see the advantage of using such Cloud service for development and testing, but I’m not sure of the advantages of using it for a production environment. Obviously many factors will need to be taken into consideration such as cost, security, performance, etc… compared to to dedicated server hosting or any other solution.

Trac Requirements (quoted from Trac website):

To install Trac, the following software packages must be installed:

• Python, version >= 2.3
  • if using mod_python together with xml-related things, use python-2.5. expat is namespaced there and does not cause apache to crash any more(see here for details).
  • For RPM-based systems you might also need the python-devel and python-xml packages.
  • See instructions in TracOnWindows/Python2.5
• setuptools, version >= 0.6
• Genshi, version >= 0.5 (was version >= 0.4.1 on previous 0.11 release candidates)
• You also need a database system and the corresponding python drivers for it. The database can be either SQLite, PostgreSQL or MySQL (experimental).
• Optional if some plugins require it: ClearSilver

Verify Python is installed:  

$ yum list python

and if need be

$ yum update python

I had version  2.4.3-9.FC5 installed.

Install  setuptool

# wget http://pypi.python.org/packages/2.4/s/setuptools/setuptools-0.6c9-py2.4.egg#md5=260a2be2e5388d66bdaee06abec6342a

# chmod +x setuptools-0.6c9-py2.4.egg

#./setuptools-0.6c9-py2.4.egg

# easy_install
error: No urls, filenames, or requirements specified (see --help)

all good, now over to the next step

Install Genshi

I don’t even know what Genshi is anyway installing it was so easy:

[root@future ~]# easy_install Genshi
Searching for Genshi
Reading http://pypi.python.org/simple/Genshi/
Reading http://genshi.edgewall.org/
Reading http://genshi.edgewall.org/wiki/Download
Best match: Genshi 0.5.1
Downloading http://ftp.edgewall.com/pub/genshi/Genshi-0.5.1-py2.4-linux-i686.egg
Processing Genshi-0.5.1-py2.4-linux-i686.egg
Moving Genshi-0.5.1-py2.4-linux-i686.egg to /usr/lib/python2.4/site-packages
Adding Genshi 0.5.1 to easy-install.pth file

Installed /usr/lib/python2.4/site-packages/Genshi-0.5.1-py2.4-linux-i686.egg
Processing dependencies for Genshi
Finished processing dependencies for Genshi
[root@future ~]#

Cool, job done, now next step

Installing the database system and python drivers

[root@future ~]# yum list sqlite
sqlite.i386                              3.3.3-1.2              installed

[root@future ~]# yum list python-sqlite2

Available Packages
python-sqlite2.i386                      1:2.3.3-1.fc5          extras

[root@future ~]# yum install python-sqlite2

Installed: python-sqlite2.i386 1:2.3.3-1.fc5
Complete!

Sweet, now what is next?

Optional Requirements

Webserver check

[root@future ~]# yum list mod_python

Installed Packages
mod_python.i386                          3.2.8-3                installed

Subversion check

[root@future ~]# yum list subversion
Installed Packages
subversion.i386                          1.3.0-4.2              installed
Available Packages

subversion.i386                          1.3.2-2.1              updates
[root@future ~]# yum update subversion

Updated: subversion.i386 0:1.3.2-2.1
Complete!

Install Trac

[root@future ~]# easy_install http://svn.edgewall.org/repos/trac/tags/trac-0.11.2.1
Downloading http://svn.edgewall.org/repos/trac/tags/trac-0.11.2.1
Doing subversion checkout from http://svn.edgewall.org/repos/trac/tags/trac-0.11.2.1 to /tmp/easy_install-OpAx-y/trac-0.11.2.1
Processing trac-0.11.2.1
Running setup.py -q bdist_egg --dist-dir /tmp/easy_install-OpAx-y/trac-0.11.2.1/egg-dist-tmp-ssiwz5
Adding Trac 0.11.2.1 to easy-install.pth file
Installing trac-admin script to /usr/bin
Installing tracd script to /usr/bin

Installed /usr/lib/python2.4/site-packages/Trac-0.11.2.1-py2.4.egg
Processing dependencies for Trac==0.11.2.1
Finished processing dependencies for Trac==0.11.2.1
[root@future ~]#

Installing the Trac environment

[root@future trac]# trac-admin /backup1/trac/TestProj initenv

Project environment for 'TestProj' created.
You may now configure the environment by editing the file:
/backup1/trac/TestProj/conf/trac.ini
If you'd like to take this new project environment for a test drive,
try running the Trac standalone web server `tracd`:
tracd --port 8000 /backup1/trac/TestProj

Then point your browser to http://localhost:8000/TestProj.
There you can also browse the documentation for your installed
version of Trac, including information on further setup (such as
deploying Trac to a real web server).

The latest documentation can also always be found on the project
website:

http://trac.edgewall.org/

Congratulations!

[root@future trac]#

All Done.

Introduction

This extension provides an implementation to minimize ExtJS Windows by showing a 48×48 icon on the window container, the window title is shown under the icon. When the user click the minimize button on a minimizable Ext window an icon is shown and the window gets hidden. The icon is draggable so can rearraged. Double clicking ont the icon shows the window and hides the icon. Right clicking on the icon show a context menu to either show or close the window. Closing the window will remove the icon.

This behaviour is ideal for Web Applications that follow the MDI (Multiple Document Interface) approach, where unused windows can be minimized to the main container background. If the window has unfinished work the user can minimize the window and come back to it later, double click on the icon, finish the work save and close the window.

In a Multiple Document Interface application each UI functionality is implemented in its own window that lives in a larger window container. The control of the various windows is usually done through the use of command menus. This type of design is suitable for Web applications that offers direct replacement of tradition  thick clients.

Demo

http://omer.dawelbeit.info/ext/window.html

Download

http://www.extjs.com/forum/showthread.php?t=52923

Sounds great, but I get really frustrated when I write JavaScript code that works fine in Firefox and Safari, but errors in IE. These errors usually result in hours of cursing and hair pulling, which lead me to believe that if Microsoft did not bundle IE free with Windows in the first place it would have never been the first browser anyway. Normal people do not really care, but for developers IE is a great pain. You either develop for IE or other browsers, or go to great length and cost to ensure your code works in at least IE, Firefox, Safari and may be Opera. The cross browser compatibility issue is hindering the adoption of browser based Rich Internet Applications that are based on JavaScript, DHTML and CSS. Many toolkits such as Dojo and ExtJS are compatible with many browsers, but you have to write your own JavaScript code which might not be compatible with all browsers.

I have faced two issues with IE where I pulled my hair for a while, but was glad to find some answers in Google before I ended up with no hair, apparently many other developers experienced the hair pulling over IE behaviour and if you feel the same you are not alone.

First problem: tr.innerHTML, I developed a nice piece of JavaScript that manipulates a table and dynamically create and delete table rows based on available data and user’s actions. Since I developed it in Firefox I used the innerHTML function to populate the table columns in the table rows, a piece of code that worked fine in Firefox, but failed miserably in IE, after some hair pulling and Googling the answer was you can not use innerHTML with TR in IE. You have to use the tr.deleteCell and insertCell functions I had to grow my code considerably to also work in IE.

Second problem: IE error message ‘Expected identifier, string or number’. I got this error message on some code that worked successfully in Firefox. And although the Microsoft Script debugger pointed me at the error location, I just could not see any problem with the code. A couple of hours of frustration and hair pulling I decided to search for the error message in Google and voila I found the answer it is to do with comma in the code below which will work successfully in Firefox, but fail in IE:

// private functions
/**
* @return window if found
* @param {Object} windowName
*/
var getWindow = function(windowName) {
var win = Ext.WindowMgr.get(windows[windowName]);
if(!win)    {
// create a new window
win = new Church.windows.addChurchWindow({
id: ‘add-church’,
});
if (desktopEl) {
win.render(desktopEl);
}
}
return win;
};
I think I left this comma in place by mistake after deleting a second line. Anyway I would not expect more from IE, I just wish everyone will be convinced to move away from IE and use Firefox or Safari, so we developers don’t have to go through the pain of developing for IE. And I also wish that IE would degrade gracefully in case of the issues mentioned above and many others.

SOHO routers are not designed for heavy use or to handle a large number of UDP connections usually required by VPN connections. This is a great pain for someone working from home and using VPN to connect to their company’s network. So cheap SOHO and ISP supplied routers tend to run very slow and drop the connections sometimes if they get overheated or if they have to handle many VPN or P2P connections. There is a good article on the ’slow running’ of routers on dd-wrt.com website. For me I’ve decided to ditch my router and look for an alternative, reliable and cheap solution.

Using a computer as a router seemed as a good idea at the time, after all SOHO routers have the following features:

• Microprocessors run at around 75MHz
• Have a number of RJ45 interfaces, at least one for WAN and one for LAN access
• Use IP Masquerading (NAT) to allow LAN computers to access the WAN using a single WAN IP address.

Now all the above features are achievable using a computer. I looked around and couldn’t find a 75MHz computer and I regretted throwing away my old 8086 Amstrad. Anyway I had a Pentium 1 233MHz MMX motherboard with a processor lingering around since 1998 so I decided to use it to build a computer and use as a network router. This is what I needed:

• Pentium 1 233MHz MMX motherboard with a processor
• 2 Network cards, one for WAN (Red Interface) and one for LAN (Green Interface)
• An old 5GB hard disk
• 256MB of memory
• An old VGA display adapter
• Some sexy neon lights and bio-hazard logo to spicen up things.
• Keyboard and monitor only needed to setup the OS, after that they can be taken away

For the operating system, using a lightweight Linux distro seemed like a good option, so searching around on Google I found two good Linux distros for firewall routers: SmoothWall and IPCop. Having briefly looked at the manual for both, it seemed like IPCop has most of the features I needed specially the ability to setup custom routing. I run two subnets with the traffic between them hitting the router, so the router needed to forward the traffic between the two subnets as configured, bearing in mind the router is not the gateway between these two subnets, I use a Fedora Core server for this purpose.

IPCop is basically a cut-down Linux destro with an Admin Web interface. There is no graphical interface, but remote login through SSH is available and can be switched on or off as required. There are so many other features available from the Admin Web interface such as Intrusion Detection, Traffic Shaping and Dynamic DNS, to mention only a few. Ever since it was installed, my IPCop Firewall Router is running effortlessly day-in and day-out without any issues so farewell routers.

Java Persistence API (JPA) coined with Java 5 Annotations made Object to Relational Mapping (ORM) a bliss. The problem is most of the detailed documentations available usually consider JPA from the POJOs perspective, but what if we are creating a JPA application that should persist to an existing database. An Entity Relationship diagram can easily be constructed for existing databases using a tool such as Toad Data Modeler. From the ER Model we can work backwards to create the JPA annotations required to presist to the existing database. This article provides a mapping that can be used to create the JPA annotations for an existing database’s ER Model using Hibernate’s implementation of JPA.

I prefer to use the crow’s feet notation in the ER model to represent the cordiality of the relations and the participation conditions as explained in Table 1 below. An example of an ER model that uses this notation is shown in Figure 1 below.

Table 1 – Crow’s feet notation used in the ER- modelling

Figure 1 - Sample ER Model using crow’s feet annotation

The approach that can be used to map relations form the E-R model into JPA annotations is summarised in Table 2 below. This approach can be followed to map the entity classes into the existing data schema using JPA annotations. Good database design should resolve all the Many to Many relationships during the logical database design stage by introducing a third dependent relation, and hence there was no need to use @ManyToMany JPA mapping.

I have also discovered that the Cascade annotation for foreign key updates on the parent table should not be included in the child entity as this will result in the deletion of the parent row whenever a child row is deleted. It was clear that this constraint should be implemented in the database schema and not included in the JPA annotation as it does not have the desired effect.

Table 2 – ER-Model and Relational Mode to JPA and Hibernate mappings