DataDirect pools connections based on unique connection strings (as do most DB providers), but shouldn’t there be a way to use the same connection string for multiple users?  As it turns out, DataDirect and Oracle provide just that mechanism.  DataDirect calls it reauthentication.  This lets you open a DB connection using a common proxy user, set the CurrentUser and CurrentPassword properties on the connection, and then open the connection.  Key here is that all users need the Oracle CONNECT THROUGH permission set.

With this setup, there is only one connection pool for the application, allowing much more efficient reuse of connections across users instead of needing a new connection every time a user logs in.

There is a small performance penalty to be paid with this configuration, however.  Each context switch of a connection from one user to another incurs a small hit, but the overall performance of the application in our case was vastly improved.

I haven’t seen a more effective way of learning about secure coding practices for the web.  It’s certainly not as interesting to read about Cross-Site Scripting than it is to try and get an alert(1) window to pop up on another website.

I definitely recommend going through at least the first few exercises if you have any responsibility for a public-facing website or just an interest in web security.

It turned out that the timeouts were directly caused by a patch addressing mutex issues that needed to be applied to Oracle.  However, many connection settings in the application contributed to the problem, and in the process of debugging the issue I learned a lot about connections that I’ll try to share here.

In this system, all application users are also database users.  Many consider this to be a security best practice in Oracle, but I’ll leave discussion of that for a later date.  The problem arises because DataDirect does connection pooling based on unique connection strings (every unique connection string gets its own pool).  In this setup, ever user was getting his own pool to each database in the application, which led to rapidly increased memory usage under load. 

One issue we identified that was contributing to the high memory usage was that the Load Balance Timeout connection string parameter was set to 1 hour.  This meant that, combined with the 40 minute App Pool recycle, that user connections were never being gracefully expired from the pool.  Every user had their own connection pool that lived until the App Pool recycled and killed every pool at once. 

Changing this timeout parameter to 10 minutes allowed us to change the App Pool recycle to a daily event.  Users did not see any impact, because their connections only expired after 10 minutes of inactivity, and the penalty for re-establishing a connection was rare and comparatively minor.

Unfortunately, the instructions in the Hanselman post didn’t work for me.  I got a Hyper-V VHD of Server 2K8R2, followed the bcdedit instructions, and rebooted.

After rebooting, I got a Boot Configuration Data error saying “The configuration for an element within the object is invalid in the boot configuration data store.”  I’ve seen this error crop up in the comments of every blog post (and on Jim O’Neil’s blog) about Boot to VHD, but no one seems to ever provide a solution. 

The thing that finally helped me find the error of my ways was a post by Michael Flanakin, where he offhandedly says “NOTE: You'll need to be sure you can boot using the necessary device in BIOS settings (obviously).”  He’s talking about the install media, but I realized that I had my VHD on a second spindle (as Scott Hanselman recommends for BitLocker systems), and the second hard drive was not enabled for booting in the BIOS.  A quick trip through the BIOS to add the secondary HDD and I could finally boot to my VHD image!

Cost savings in general refers to benefits of a project that the customer can add back into his budget and spend.  These are costs that have already been planned, but are now unnecessary.

Cost avoidance refers to everything else, typically more “soft savings.”  Often cost avoidance involves slowing the rate of cost increases or obtaining “value-added” services.

The best resource I found detailing the differences between the two concepts was from the eSourcingWiki.

• We moved the file download portion of our ASP.NET 3.5 WebForms application to MVC

Symptoms

• When IE6 users tried to open a file (text, Word doc, PDF, etc) through the new MVC controller action, the appropriate application would launch and fail with a “file not found” error
• If the users clicked “Save” instead of “Open” in the file download dialog, they were able to open the downloaded file without issue
• IE7, IE8, and Firefox users were unaffected

Solution

• The issue lies in the way IE6 handles the NoStore caching instruction on the OutputCache attribute. 
• Removing the no-store instruction from the file download controller action solved the problem.
• What I think happens here is that IE7 and IE8 ignore the no-store instruction when downloading a file (as opposed to loading html content, for example).  IE6 takes the no-store instruction literally and will not store the document in the cache at all.  The problem is that when you click open on the file download dialog, the browser needs to store the file in temporary storage (the cache) before loading it. 

I found that project due to the lack of plug-in support in the Zune desktop application.  I listen to a lot of music at work, and I like to listen to entire albums.  What I wanted was a way to randomly select an album that I should listen to next.  I don’t want an album where I only have one or two songs, I want to select an entire album.

Enter PowerShell.

Step 1: The Integrated Scripting Environment

To get started, open up the PowerShell ISE by typing “ise” in the Windows 7 Start menu.  Right click the PowerShell ISE and select run as Administrator (this will save some headaches later).

We’ll see a three-paned window, so let’s step through the three panes one-by-one.

Top: Editor

The top pane is the script editor.  This is where we can enter a multi-line PowerShell script and save is as a .ps1 file.  It has syntax highlighting, as you would expect from an editor.

Middle: Output

This is where the output of your script will be displayed.  If you ran your script in a PowerShell console, the text that appears in this box would appear just below the line you entered.

Bottom: Prompt

The bottom pane is a PowerShell prompt, just like you would see in a PowerShell console.  You can execute single PowerShell commands here without including them in your script.

Step 2: Directory listing

Let’s start with a simple script to list out the subdirectories of a directory.  We’ve wrapped that in a function, because we’ll be using recursion later.

The second line calls two built-in PowerShell cmdlets, get-childitem and where-object.  Get-childitem, when called on a path, lists the contents of that directory.  The output is then piped into where-object, with a constraint saying we only want directories, not files.  The $_ variable is used similar to the way it is in Perl to refer to the output of the last line.  The end result of this line is an array of subdirectories of D:\Music. 

For me, this is a list of artist names and a Zune directory that contains all the music I’ve downloaded from Zune.  What I want is the albums (subdirectories of the artists) and the albums within the Zune directory (contained with artist directories).

Step 3: Getting Albums

These changes now pull subdirectories of the artist directories (which are albums) and limit those albums to ones that have at least 8 files in them.  Note that the last call to where-object looks for files, not directories (it negates the psiscontainer call).

One thing that always trips me up in PowerShell is the operators.  Sometimes it looks very similar to C#, but the operators are totally different – see the –gt for greater than.

Step 4: Handling the Zune Directory

Now we’re using recursion to handle the Zune directory.  We also need to ignore some of the Zune special directories that aren’t artist directories (note the –or operator, that tripped me up).  We also start using array variables and adding albums to the array using the += operator.

Step 5: Random Album

The last thing left to do is select a random album from the results.  We’ll do this by wrapping the Get-Album function with a new function that uses the Get-Random cmdlet to select an album from our master list.

Step 6: Running the Script

You can execute the script at any point by pressing F5, but after you save the script, you’ll need to change the permissions to allow execution of the script.  By default, PowerShell is fairly locked down in terms of permissions.  To allow execution of your own saved scripts, you’ll need to type this in the PowerShell prompt once:

set-executionpolicy remotesigned

Here’s a listing of the entire script:

function Get-Album ($topDir)
{
    $directories = get-childitem $topDir | 
                    where-object {($_.psiscontainer)}
    $returnvalues = @()
    foreach ($dir in $directories)
    {
        if($dir.Name -match "Channels" -or 
            $dir.Name -match "Books" -or
            $dir.Name -match "Playlists" -or
            $dir.Name -match "Podcasts" -or
            $dir.Name -match "Received Pictures" -or
            $dir.Name -match "Zune Temporary Music") 
            {
                continue
            }
        if($dir.Name -match "Zune" -or
            $dir.Name -match "Subscription")
            {
                $returnvalues += @(Get-Album $dir.FullName)
            }
        else
            {
                $subdirs = Get-Childitem $dir.FullName | 
                            where-object {($_.psiscontainer)}
                foreach ($subdir in $subdirs)
                {
                    $files = Get-Childitem $subdir.FullName | 
                                where-object {!($_.psiscontainer)}
                    if($files.count -gt 7)
                    {
                        $returnvalues += 
                            @($dir.Name + ' - ' + $subdir.Name)
                    }
                }
            }
    }
    return $returnvalues
}

function Get-RandomAlbum ($topDir)
{
    $albums = Get-Album $topDir
    $index = Get-Random -min 0 -max ($albums.count - 1)
    return $albums[$index]
}

Get-RandomAlbum D:\Music

The next step is to wrap this all into a Windows Gadget so it’s on my desktop all the time and I don’t need to launch PowerShell to get a random album suggestion.

• ASP.Net 3.5 application with a SQL Server 2005 backend
• Full Text Search index (to search a document library stored in the database)
• Web frontend to the document search
• Search is handled by a stored procedure

Symptoms

• Running a query first thing in the morning results in a SQL timeout
• Subsequent queries are very fast
• Any query run after a delay times out
• Re-compiling the stored proc every few minutes prevents the timeouts

Solution

We initially suspected that the timeout was due to the changing index somehow invalidating the execution plan for the stored proc.  As it turned out, the real reason was something much stranger.  Apparently the Internet connection for our DB server didn’t allow SQL Server to check the certificate revocation of the full text word breaker (as described in this KB Article).  Unlike the scenario described in the article, the DB server does have an Internet connection, but it needs to go through a firewall.

Even so, the solution in the KB worked for us, we just added this line to our SQL scripts:

sp_fulltext_service 'verify_signature', 0;

Note: File associations can only be set for Full Trust ClickOnce applications

The MSDN article walks you through creating a file association using XML in the application manifest, but most developers will find it easier to use the Project Properties screen:

Unfortunately, this is where the MSDN documentation leaves off, which tells you nothing about how your application knows the location of the file it is supposed to open.  With a ClickOnce application, the file location is not passed in as a command line argument, it is actually in

AppDomain.CurrentDomain.SetupInformation.ActivationArguments.ActivationData[0]

One other interesting point is that the file location in ActivationData is in a uri form such as file://c:/temp/Test%20File.testApp.  In order to use this as a path to IO functions like File.Exists, you’ll need to create a Uri from it, get the absolute path of the Uri, and Unescape the string (to convert %20 to a space):

For me, Expression was just familiar enough to start using without any training, but new enough that I didn’t feel that I was using all the features I needed.

Thanks to my co-worker Marcos, who pointed me to a great series of Expression videos on the Microsoft Expression website.  The videos are titled “Getting Started with Silverlight – For Designers,” but I found the content was very valuable as a developer unfamiliar with design processes.

The videos provide a great introduction to taking full advantage of the design and animation tools inside the Expression Studio.  I’m hoping to use what I learned on my project and also as material for a future blog post!