1. A set of initial solutions to try is created
2. These are then tried against a specific task (in this case can the vehicle move across the terrain) this is called the fitness function.
3. Some vehicles will move better over the terrain than others these will be given a higher fitness rating
4. The GA then selects the solutions with the best rating. To stop future solutions becoming too similar and converging a few random of the not so successful solutions from the previous group are also "bred"
5. These solutions are then bred together in an attempt to produce better solutions

nbsp;

Pretty cool eh?

For more information Wikipedia as usual is our friend:
http://en.wikipedia.org/wiki/Genetic_algorithm

• .Net Services (Message broker, security and Workflow hosting)
• SQL Services (an online database accessed using REST)
• Live Services (a number of different services from interfacing with messanger to sending alerts)
• Sharepoint Services
• Microsoft CRM

Currently .net and to some degree SQL services are available.

To play around with Azure you need to register so sign up at: http://www.microsoft.com/azure/register.mspx.

Microsoft will then send you a key (mine took about a week to arrive) allowing you to access the online facilities.

Until you get this key you can still play around to an extentby downloading the SDK: http://www.microsoft.com/downloads/details.aspx?FamilyId=BB893FB0-AD04-4FE8-BB04-0C5E4278D3E9amp;displaylang=en.

I suggest you also download the tools for visual studio from http://www.microsoft.com/downloads/details.aspx?FamilyId=63D0D248-1B08-4F7D-ABDE-62EB75CB1E69amp;displaylang=en. As this is beta software you probably want to be running it in virtual pc (although I didnt and everything seems fine still).

The first thing you should do is watch Steve Marx's presentation on developing an application in Azure at: http://channel9.msdn.com/pdc2008/ES01/
Steve is an excellent preseter, great to watch and will take you through the basics of Windows Azure.

After installing the Visual Studio tools you can create Web Cloud service ASP.net projects. Cloud projects are very similar to standard ASP.net projects but contain another project that has some additional files that describe where the service is hosted and configuration settings (.csdef and .cscfg). 

If you look at the project name in Visual Studio you will see it is called something like CloudService1_webRole. Role is a term within Azure that seems to describe the type of service you are running. If you right click on the service project you can deploy it to the cloud (which they havent given me access to yet).

As it would be inconvenient to deploy your app every time you wanted to test it Microsoft provide an application called the development fabric which simulates how it will function in the cloud (below). If you open the cscfg in studio you will see its an XML file that contains the following:

lt;?xml version="1.0"?gt;
lt;ServiceConfiguration serviceName="CloudService3" xmlns="http://schemas.microsoft.com/ServiceHosting/2008/10/ServiceConfiguration"gt;
lt;Role name="WebRole"gt;
lt;Instances count="1"/gt;
lt;ConfigurationSettingsgt;
lt;/ConfigurationSettingsgt;
lt;/Rolegt;
lt;/ServiceConfigurationgt;

If you alter the instances count element and run the app you will see more instances created in development fabric. This simulates adding capacity to your Azure application e.g. if you are expecting many visitors to hit the site.

 


Once you have your key and have activated it you can log into the Azure services platform. This will take you to an interface like the screen shot below. At present its a bit rough around the edges but you can see the direction they are going.



Currently within .net services you have access to 3 components:

• Service Bus (acts as a broker/intermediatry for messages)
• Workflow (hosts workflow)
• Access control service (authentication and authorization)

The examples provided in the SDK take you through the basics of how these work. There is an example of a chat application that sends messages via the service broker. What is impressive is the speed of it, no noticable delay at all. There is also another example application to demonstrate the security features )(a calculator app). This threw an security exception when I tried to run it about token being valid 10 seconds after it was requested?!

Security is defined by an input claim e.g. a specific username which is then matched to Output claims (e.g. in the included example give the ability to multiply).

 




Some of the examples are a bit cryptic at present and I would suggest brush up on your WCF and Workflow as these concepts will certainly be used. 

It all looks really promising so far and I cant wait to learn more about it.

• Is the current infrastructure really quick and reliable enough
• Can Microsoft really cope with huge demand (e.g the PDC site is struggling although to be fair there is a lot of video streaming going on)
• Some clients will object to data being hosted out of their organisation or country

Having said this it does look interesting from the point of view of:

• Let Microsoft manage the infrastructure
• Can start off small and easily scale application
• Cheaper? - we will have to wait and see!

• Decorated functions with the WebMethod attribute (although I have had problems using this with nested controls).
• Handcoding webservice calls (This is a pain to create and process the return data)
• JSON and JQuery

Perhaps the best option if you just want a single value back is to use a format called JSON in conjuction with JQuery.

JSON stands for Javascript object notation and is a very lightweight format to return a simple object. For example a stock quote result might look something like this:

{Symbol:"MSFT",Value:"4.4"}

I think we can all agree thats a lot less data than your average SOAP call and its pretty readable making JSON less heavy on bandwidth and probably quicker to process than XML (needs confirming!). Libraries such as JQuery understand JSON is an object which makes it very easy to work with e.g:

alert(objMyResult.Value);

Which is much preferable to processing XML from SOAP results etc...

But how do you use this with ASP.net?

Currently ASP.net and JSON support is pretty ropey (however nothing to stop you returning JSON formatted strings). In the MVC framework there is some nice inbuilt functionality for dealing with JSON posts in controllers. I decided to add the ability for users to comment on my blog posts and decided to implement this with MVC. The client side code is below and can be seen in action by commenting on any of the blog posts. Note you will need JQuery to use this code and currently the Json result class is only included in MVC projects (System.Web.MVC).

Client side code:

function submit()

{

nbsp;

var strName = $("#txtName").val();
var strEmail = $("#txtEmail").val();
var strComment = $("#txtComment").val();
var strSpamCheck = $("#txtSpamCheck").val();
var rand1=window.rand1;
var rand2=window.rand2;

$.ajax({
type:

nbsp;

"POST",
dataType: "json",
url: "lt;%=Application["URLRoot"].ToString() + "BlogEntryComment/Submit?id=" + objBlogEntry.BlogEntryID.ToString() %gt;",
data: { Email: strEmail, Name: strName, Comment: strComment, SpamCheck: strSpamCheck, Rand1: rand1, Rand2: rand2},
success: function(result) {
$("#CommentEntry").fadeOut("slow");
setTimeout(redirect,2000);
},
error: function(error) {
alert('error ');
}

});
}

nbsp;

Then server side I use this code in my controller class for processing BlogEntryComments:

If you want to adapt this for you own use couple of things to note:

• Parameter names for the server side function must match exactly what you declare in the client side Javascript
• Make sure you are posting it to the right controller and URL - this can be confusing if your running on debug with the inbuilt webserver. I found it best to attach to process

• Intellisense in views :)
• New Scripts directory with JQuery
• Html helper enhancements

http://weblogs.asp.net/scottgu/archive/2008/10/16/asp-net-mvc-beta-released.aspx

A number of things come to mind:

• Why were 2 loan accounts created when they had several unique identifiers e.g. my NI number, student number etc
• The letters with the login details really need to match up to a username maybe display the last 3 digits etc
• Why give the user an option with something that can be automated like capitalizing the data in the textboxes?
• Or failing this at the very least tell the user they need to do this on the page
• Why not remove spaces from the phone number automatically?

It doesnt take too much imagination to see the problems with the above. The bozos who designed this system would do well to read a book called Dont Make me think by Steve Krug (http://www.amazon.co.uk/Dont-Make-Think-Usability-Circle-Com/dp/0789723107/ref=sr_1_2?ie=UTF8amp;s=booksamp;qid=1224192586amp;sr=8-2)

Steve considers when people make decisions they make not necessairly the smartest but the easiest decision possible (with no problems).

This has implications for user interface design the ideal being the user should make the decisions without having to think about them. I think you can see this with a well designed site like Amazon or StackOverflow where you just feel drawn to the right button etc. You can help or hinder this process by adhearing to well known standards and ui elements.

After covering basic concepts and web standards Steve takes a number of websites, shows the current design and alters them (some times very slightly to improve the usability). Every developer should be made to read this book. It took me about 2-3 hours to read cover from cover and it was an entertaining read. Every developer should read this book.

• Its not the most straight forward thing but here is the basic concepts
• Your image is essentially a byte array
• In a (RGB) image 3 values from 0-255 make up the image (you could also have a 4th for gamma I think)
• Depending on the strenth of these values it makes a color e.g. 255 0 0 would be red
• For some reason when you read these values they come out back to front e.g. BGR 0 0 255 for Red (?)
• As .net is managed code and we are using pointers we need to mark it as unsafe and click the allow unsafe code
• To manipulate these values you load up the image, Fix its place in memory then get a pointer to the start of it (scan0)
• Each line of the image from the top is the width of the image * 3 (in 24 bit image)
• This number must be divisible by 4 otherwise it will have a padding bit

Belows the code I used to convert an imagr to greyscale. I have commented out brightness adjustment (just add to each value) and some other stuff I was playing with.

I found a few links which helped me put this together the codeproject was very helpful (although not the most readable):

http://www.bobpowell.net/lockingbits.htm
http://www2.sys-con.com/ITSG/virtualcd/Dotnet/archives/0104/brown/index.html
http://www.codeproject.com/KB/GDI-plus/csharpgraphicfilters11.aspx

In relation to standard ASP.net

• Model is your data
• View is your aspx and ascx controls
• Controller is your business objects, managers etc

MVC doesnt make use of standard view state so some scenerios may be more difficult than classic ASP.net. The application uses the url and http verbs to map to the views e.g /products/GetList is automatically mapped to GetList view

The main advantage of using MVC are that:

• As each unit is independent they are easy to change
• Seperation suited to unit tests
• Better search engine optimization due to url parsing

After the MVC session we watched a talk on optimizing ASP.net performance from the front end.

The main conclusions were:

• Dont make unnecessary http requests e.g. consider combining js and css files
• Place js and css files at bottom of home page so they are preloaded and due to page position dont interrupt user experience
• Careful with use of update panel as it posts back whole page, use [webmethod] attribute instead
• Implement http compression
• Implement compression in asp.net side
• Silverlight can be used to preload stuff and appears quicker than javascript
• Use caching - duh!

The other session I atteneded was given by Mike Taulty on how Silverlight can interact with the page.

This was very impressive:

• Silvelight can refer to, add, modify amp; hook up to events on the external html page
• The html page can talk to silverlight
• Call web services
• Listen on sockets

After meeting up with Howard just outside the exhibition hall the day began with an excellent key note speech given by Bill Buxton ( usability and industrial design expert) Bill has written an interesting looking book called Sketching user experiences which I purchased.

Bill argued that in many places the design phase occurs after the product development and that it was a mistake to let non experts be too involved in design (e.g. the FD on a project being involved to the level of screen design). He argued that good design was dependent on many aspects citing Apples Ipod. Prior to this Apple had been in decline. At this time a number of key personnel Jonathan Ive had been at Apple some time but the Ipod wouldnt become a success until a combination of events such as:

• Jobs empowering the designers
• Lawyers negotiated with record labels $99 per track
• Ives and his team creating the design

After this we attnded 2 sessions by Scott Gu on Silverlight. I have spent some time with Silverlight lately and would highly recommend the video tutorials. Scott said that Silverlight 2 was due to be released shortly (somewhat non committal!). Other items that looked interesting were IE 8 which includes a javascript profiler and useful looking development tools, Visual studio intellisense support for javascript frameworks such as prototype, about 50 new silverlight controls including graphing components coming.

We then saw a session on Sql data services (astoria). Astoria is basically a REST api to query sql server. I am not sure what I would use this for but it may be useful for example in a silverlight situation to avoid writing a web service interface for every crud query.

The final session was on Visual Studio IDE tips given by Sara Ford. This was very interesting and contained several time saving tips. I suggest you look at her blog which has a tip a day http://blogs.msdn.com/saraford/default.aspx

After we left this session I managed to talk to Scott Gu. One of the great things about Remix was how accessible all the presenters were. To have the chance to talk to Microsoft's VP was pretty cool.



I asked Scott about whether we should be using Web sites or Web application projects (a contentious issue in our company). Scott suggested web application projects where the dll is built is the better way to go. This was due to better performance and the ability to produce unit tests as a dll is built.

After the main sessions I had a few beers with Howard and had a chance to look at Microsoft Surface. There was an excellent demonstration given which had a virtual earth map. You could rotate it at any angle and even zoom into and walk inside buildings. At $13000 per unit they dont come cheap through.

I am looking forward to the MVC session tomorrow (finding out what exactly it is!)

• Failure to lock resources accessed over multiple threads. This ends up with unexpected behaviour when 2 threads are altering the same variable. Various methods to solve inc sync lock, refactoring code, using patterns etc
• Not killing threads properly - Use finally statment etc. Never assume the thread will be killed
• Some interesting ways to manage multiple threads - make use of thread pool!
• Deadlocks - not one thats come up yet but something to be aware of

So always practice safe threading or refrain..

• Its £199, ouch!
• Books arent that much cheaper if at all (are you really asking me after paying 200 quid to pay 12.99 for a book?)
• It needs to be plugged in to download new books
• Books are not like music where you want to keep lots of songs. At most people are reading 2 or 3 books at a time and who cares if you can store 160 books your not going to flick through them unless they are some kind of reference.
• Not touch screen - why?

But the real killer is that when you change pages the entire screen flashes which is really irritating. Why with all the effort that has gone into the rest of it would you do this.

Nice try Sony but I dont think its there yet...

• When estimating time to complete a task estimate that only 60% of your day will be spent actually coding the rest on fixing, meetings, browsing adult sites etc
• Choose your battles - its not worth arguing every point even if you "know" you are right -something I know I can be drawn into :)
• Many times the ideas that are chosen are not necessarily the best - just the best presented and argued (all the more reason to practice speaking/presentation abilities!)
• Its important to be able to speak both tech and business (obvious but have seen many developers forget this)
• Testers should be made to feel part of the team - What you work in a company where they actually employ testers you lucky thing!
• You are not going to change some places, if you dont like it leave!

On the whole not a bad book and if you think you need to wise up to some of the non technical issues well worth a read.

• It allows you to automate complilation (builds) of your application
  You can get it to retreive the latest version from your source control system before building
• You can automate other tasks such as NUnit tests to be performed as part of this build
• You can receive the results of the builds and tests
• And best of all its free (always a plus with tight ass bosses)!

But why would you want to do this?

• In a large scale project with many developers working on it you can find out early on if a change has broken the build
• Integration testing e.g. run it overnight to receive notification of build/unit test failures
• Create many different build configurations and use CruiseControl.net as a way of kicking them off

Perhaps the only problem with CruiseControl.net is the documentation which you are going to have to do a bit of searching around. However CruiseControl.net is a popular application so there are quite a few tutorials.

CruiseControl.net has 2 parts the application itself and a web front end. The web front end is okay however you can modify various bits yourself through XSL files.

Using CruiseControl.net

1) Download NUnit from: http://sourceforge.net/project/downloading.php?groupname=nunitamp;filename=NUnit-2.4.8-net-2.0.msiamp;use_mirror=osdn (not strictly necessary but this tutorial will be integrating NUnit tests in the build)
2) Download CruiseControl.net from: http://sourceforge.net/project/showfiles.php?group_id=71179amp;package_id=83198
3) Check you have IIS amp; .net 3.5 installed
4) Install NUnit amp; CruiseControl.net
5) This will create a short cut on your desktop click this to run it you should see a dos window with a load of messages similar to this:



Create Project to build

Now what we need to do is create a sample project for CruiseControl to build.

I have created a new blank solution called cctestapp and added a console application to it. My solution file is at E:\wwwroot\Testing\cctestapp\cctestapp.sln and the application is at E:\wwwroot\Testing\cctestapp\cctestapp\.

Once you have created your console application add a reference to Nunit.framework.dll (by default this is at: C:\Program Files\NUnit 2.4.8\bin)



Add the following code to your console application:

using System;
using NUnit.Framework;
namespace MonitoringApplication
{
[TestFixture]
public class Program
{
[Test]
public void AlexTest()
{
Assert.AreEqual(1, 1);
}
public static int Main(string[] args)
{
//My program
return 0;
}
}
}

Check the application compiles okay.

Now we need to tell CruiseControl.net we want to build this project. To do this we need to alter an xml configuration file. This XML configuration file contains all the projects that are to be built. The configuration file contains a reference to a program called MsBuild. When you compile a project in visual studio behind the scences MsBuild is used to create the final output.

1) Open ccnet.config in notepad (by default its at C:\Program Files\CruiseControl.NET\server).
2) Modify it so it looks like this config file. This config file automates a build on Mondays and Wednesdays at 18:00 and runs NUnit tests.
3) Remember to alter the various paths to where you have created your project. I think these are fairly self explanatory. e.g assembly, executable, working directory, project file and logger.
4) If you now open up a web browser and go to http://localhost/ccnet/ you should see the main build screen like below:



5) Click the Force button to run the build.
6) You can then click refresh status to see if it built correctly if it did under the build status it should say Success.
7) To look at more details about whether your project built properly click the project name you can then view the build report and see the results of any tests.
8) If the project didnt build properly check your file paths are correct in ccnet.config.


You will probably want to install the CruiseControl tray application this notifies you of failures and allows you to kick of builds etc. This you can install by opening the web application and clicking the download CCTray link on the left.

Further reading

These tutorials helped me to get this running:

http://johnnycoder.com/blog/2008/01/29/getting-started-with-cruisecontrolnet/
http://weblogs.asp.net/jdanforth/pages/How-to-Hook-Up-a-VS.NET-2005-Solution-With-CruiseControl.NET-in-a-Few-Minutes.aspx