I have several tables in my application where I add some basic auditing columns, such as CreateDate, CreatedBy, LastUpdatedDate, and LastUpdatedBy.  These columns are pretty self explanatory, and I prefer to use the SQL GETDATE() function as the default value for the date columns.  This way I don’t have to worry about setting these dates when a new row is inserted, because SQL will handle it for me.

The first thing you have to do is tell EF4 that these columns are computed by SQL server.  The way you do this is by opening the Entity Data Model (EDM) and finding the matching columns in each entity.  Once you select a column, check out the Properties pane for the column.  You should see something like this:

Notice that I’ve highlighted the StoreGeneratedPattern property in this image.  This property tells EF4 how to handle the default value of this column.  More specifically, toggling this property to Computed will tell EF4 that you want the database to handle initializing this property for you.  If you don’t set this properly, EF4 will attempt to insert the default value for whatever data type the column is.  In this case, in .NET 4.0, the DateTime default is DateTime.MinValue, which is “1/1/1 00:00:00.000.”  If you try to insert this into a regular SQL datetime field, you’ll generate an exception because SQL’s regular datetime type cannot handle a date that far in the past.

One would think that simply setting StoreGeneratedPattern to Computed would fix this problem (and it actually should), however there’s a nice bug in the EDM designer that prevents this from working as intended.  Basically, EF4 EDM’s are broken into a few areas.  You have the SSDL, which is how the EDM describes your database objects and their relationships (think storage models).  Then there is the CSDL which is how your entities themselves are defined (think classes).  Finally there’s the mapping layer which acts as the man in between the SSDL and CSDL.  It’s in charge of mapping the storage models in the SSDL with the entity models in the CSDL.

So what about this bug?  Well, the problem is that when you change the StoreGeneratedPattern property in the EDM designer, only the CSDL is updated.  The SSDL is left in the dark about this change, and because of that, you will continue to experience the out of range exceptions generated above.  There is a way, however, to fix this problem if you aren’t afraid of cracking open an text editor.

First you need to open your *.edmx file in any text editor.  I prefer to use one with XML syntax highlighting just because it makes things easier to find, and it helps me avoid typos.  Right at the top of the file, you should see the SSDL defined.  First you will see the main tables defined, followed by any relationships.  Finally you will start to see your tables defined as storage models like this:

<EntityType Name="tblAnswerDefinitions">
  <Key>
    <PropertyRef Name="AnswerDefinitionID" />
  </Key>
  <Property Name="AnswerDefinitionID" Type="int" Nullable="false" StoreGeneratedPattern="Identity" />
  <Property Name="QuestionID" Type="int" Nullable="false" />
  <Property Name="AnswerTypeID" Type="int" Nullable="false" />
  <Property Name="Value" Type="varchar" Nullable="false" MaxLength="50" />
  <Property Name="Points" Type="int" Nullable="false" />
  <Property Name="IsDeleted" Type="bit" Nullable="false" />
  <Property Name="CreatedBy" Type="int" Nullable="false" />
  <Property Name="CreatedDate" Type="datetime" Nullable="false" />
  <Property Name="LastUpdatedBy" Type="int" Nullable="false" />
  <Property Name="LastUpdatedDate" Type="datetime" Nullable="false" />
</EntityType>

Notice that the CreatedDate and LastUpdatedDate columns don’t have the StoreGeneratedPattern property set, even though we’ve defined them in the EDM designer!  The fix is to simply change the lines and add the property.  When you’re done, it will look like this:

<EntityType Name="tblAnswerDefinitions">
  <Key>
    <PropertyRef Name="AnswerDefinitionID" />
  </Key>
  <Property Name="AnswerDefinitionID" Type="int" Nullable="false" StoreGeneratedPattern="Identity" />
  <Property Name="QuestionID" Type="int" Nullable="false" />
  <Property Name="AnswerTypeID" Type="int" Nullable="false" />
  <Property Name="Value" Type="varchar" Nullable="false" MaxLength="50" />
  <Property Name="Points" Type="int" Nullable="false" />
  <Property Name="IsDeleted" Type="bit" Nullable="false" />
  <Property Name="CreatedBy" Type="int" Nullable="false" />
  <Property Name="CreatedDate" Type="datetime" Nullable="false" StoreGeneratedPattern="Computed" />
  <Property Name="LastUpdatedBy" Type="int" Nullable="false" />
  <Property Name="LastUpdatedDate" Type="datetime" Nullable="false" StoreGeneratedPattern="Computed" />
</EntityType>

The final step is to save the EDMX file, and then open it in Visual Studio.  Make a minor change (like moving the position of one of the entities in the design view), and then save it.  Once the file is saved, the code generator will re-run, and the problem should be solved.  One other pretty annoying thing I’ve noticed is that if you ever have to set another field’s StoreGeneratedPattern property, you actually have to repeat this entire process for all fields, because the designer will wipe out your manual changes.  There is a Microsoft Connect ticket open on this issue, but it has been closed as fixed.  It’s most definitely not fixed, so it could use your attention!  Cheers!

Today, however, I came across perhaps one of the most ridiculous design flaws I’ve ever seen.  Anyone that uses Visual Studio will tell you that the right-click context menus in the solution explorer are critical.  As developers we’re constantly right-clicking stuff in solution explorer and accessing the “Properties” command or perhaps some source control commands.  Much to my surprise, Microsoft decided to implement scrolling in these context menus for reasons unknown to me.  Check this out:

Now before you reply telling me about a lack of screen real estate, let me ensure you that this isn’t the problem.  It has some sort of weird dependency on the item you click being positioned somewhere near the center of the solution explorer.  For example: if I simply expand a few more folders and force the selected item closer to the bottom of the solution explorer, the context menu renders fine:

This issue was reported in February.  Remember when Microsoft delayed the release of VS2010 to fix some speed and usability issues?  Why wasn’t this fixed?  GRRRRR!  Please vote on the Microsoft Connect issue to get this fixed as soon as possible.  Asking us to wait until SP1 is asinine.

In the spirit of truly immersing myself in Rails development, I’ve come to realize that even though it’s 100% possible to write Rails applications on Windows, the experience is less than ideal.  I’m sure some of you out there are very successful Rails developers using Windows as your primary OS and more power to you!  But this is my journey, and I feel like I really should use the best-of-breed development environment for the task at hand, and Windows just isn’t going to cut it.  So… I decided that OSX was the best for me.  It seems like the vast majority of the Rails community uses OSX for development, and I don’t mind following in their well-established footsteps!

“But wait!,” I hear you say, “Aren’t you a .NET developer, and don’t you have production websites running on ASP.NET MVC?”  Ah, yes, my fair reader, you are correct!  But in today’s world of virtualization, especially on an OSX host, there’s no reason you can’t run your entire .NET development environment inside a beefy VM, and my main machine is definitely beefy enough to handle it!  See for yourself:

On my main machine, I’m actually running Snow Leopard now.  I installed OSX 10.6 from a retail DVD (thanks Apple Store!) and I’m loving every minute of it.  My virtual machine is a Windows 7 32-bit install with the “Windows Classic” theme because it honestly looks cleaner to me.  VMWare Fusion makes all of this possible, and I couldn’t be happier.  My performance is rock solid, so far.  And the best part is, now I can work in the best .NET environment through sweet virtualization, and the best Rails environment because let’s face it:  TextMate on OSX is sexy as hell:

If there’s interest in a more specific outline of how I got all this set up, I could definitely make a “How To” post or maybe a screencast that details everything.

Earlier this year I attended MIX09, and one of the key takeaways was the Microsoft Web Platform Installer.  Microsoft has realized that for the “average Joe,” setting up a new Windows-based web server with all the necessary configurations and supporting frameworks can be rather cumbersome.  This is even more obvious when you begin to work with IIS7, given the fact that the administration system has changed drastically.  The idea behind this specialized installer is to perform all of the work for you.  Sounds neat, huh?  Well after screwing around with my VPS for hours, I can tell you that things aren’t exactly as easy as they seem.  To make a very long story quite a bit shorter, I’ve found through my troubles that things need to be installed in a very specific order:

1. Grab the Web Platform Installer.
2. Open the “Web Platform” tab and ensure that you only install components from the “Web Server” and “Frameworks and Runtimes” sections.  You might be tempted to add in SQL Server 2008 Express, but don’t.  Trust me when I tell you this.
3. Once everything is finished (which will take forever), now you can attempt to install the SQL Server 2008 Express SP1 package, but don’t bother with the Express Management Studio because it will fail no matter what.  If you must have this package, I recommend purchasing a license for Microsoft SQL Server Developer edition and installing only the client tools on the VPS.  No matter which way I tried, I could never get the Express Management Studio to install properly.
4. Pray to the Microsoft gods that everything goes without a hitch.
5. If you’re lucky, you have a working Windows-based web server.

Good luck to the brave souls that wish to venture into this territory.  I had nothing but issues with it and I wonder if I wouldn’t have been more successful if I’d have just done it manually from the start.  Perhaps these issues are just because of the VPS environment that I’m installing in, but I can’t be the only one out there that uses a virtual server for their web deployments!

Over the next few weeks I plan on spending a lot more time with Rails and ASP.NET MVC.  My end goal is to present an overview of the benefits and faults of each technology, and hopefully I can provide a somewhat objective point of view for people new to MVC.  Stay tuned!

Given that I have a MySQL database laying around, I decided to poke around in the source code to see exactly how MySQL’s PASSWORD() function really works.  It didn’t take me long to find inside the file “libmysql/password.c” this little gem:

/*
    MySQL 4.1.1 password hashing: SHA conversion (see RFC 2289, 3174) twice
    applied to the password string, and then produced octet sequence is
    converted to hex string.
    The result of this function is used as return value from PASSWORD() and
    is stored in the database.
  SYNOPSIS
    make_scrambled_password()
    buf       OUT buffer of size 2*SHA1_HASH_SIZE + 2 to store hex string
    password  IN  NULL-terminated password string
*/

void
make_scrambled_password(char *to, const char *password)
{
  SHA1_CONTEXT sha1_context;
  uint8 hash_stage2[SHA1_HASH_SIZE];

  mysql_sha1_reset(&sha1_context);
  /* stage 1: hash password */
  mysql_sha1_input(&sha1_context, (uint8 *) password, (uint) strlen(password));
  mysql_sha1_result(&sha1_context, (uint8 *) to);
  /* stage 2: hash stage1 output */
  mysql_sha1_reset(&sha1_context);
  mysql_sha1_input(&sha1_context, (uint8 *) to, SHA1_HASH_SIZE);
  /* separate buffer is used to pass 'to' in octet2hex */
  mysql_sha1_result(&sha1_context, hash_stage2);
  /* convert hash_stage2 to hex string */
  *to++= PVERSION41_CHAR;
  octet2hex(to, (const char*) hash_stage2, SHA1_HASH_SIZE);
}

As you can see, the password algorithm uses a double-SHA1 hash to format the password.  The other little tidbit is that any passwords generated post-4.1 will be pre-pended with an asterisk (*).  Given that I have spent a fair amount of time recently working with cryptography in .NET, it didn’t take long to adapt this code to C#.  The following function produces the same output as the MySQL PASSWORD() function.

public string GeneralteMySQLHash(string key)
{
    byte[] keyArray = Encoding.UTF8.GetBytes(key);
    SHA1Managed enc = new SHA1Managed();
    byte[] encodedKey = enc.ComputeHash(enc.ComputeHash(keyArray));
    StringBuilder myBuilder = new StringBuilder(encodedKey.Length);

    foreach (byte b in encodedKey)
        myBuilder.Append(b.ToString("X"));

    return "*" + myBuilder.ToString();
}

I hope this helps someone out there. I know it was kinda of fun to put on my super-slueth hat for a while .

Update 5/29/2009: I found a neat way to do this in T-SQL also. It’s a bit ugly, but it works!

SELECT '*' + SUBSTRING(master.dbo.fn_varbintohexstr(HASHBYTES('SHA1',HASHBYTES('SHA1','VALUE-TO-ENCRYPT-GOES-HERE'))),1,42)

When I logged into my VPS, one of the things I wanted to install right away was SQL Server 2008.  In order for SQL Server 2008 to install, you must first have installed the Microsoft .NET Framework 3.5 SP1.  I figured it was no big deal and decided to let SQL handle my install.  Once the installer got going, it immediately halted and displayed an error message telling me it couldn’t be installed.  After checking the log file, I found the following:

WapUI: [2] DepCheck indicates Microsoft .NET Framework 2.0SP1 (x64) (CBS) is not installed.

Obviously on Windows Server 2008 SP1, the .NET Framework 2.0SP1 is already installed by default.  Just to prove it to myself, I downloaded and ran the installer.  Sure enough, it failed before it even started, telling me that it could not be installed on Windows Vista.  So I began to poke around the VPS.  One of the things that jumped out at me was the fact that the Windows Update service was disabled.  When I checked with my VPS provider’s documentation, they indicated that Windows updates must be disabled because they are slipstreamed into the OS build.  Updating your VPS through Windows Update can (and probably will) break it.  Yikes!

After a bit more searching, I found a solution to the problem.  It turns out that the .NET Framework 3.5 SP1 installer requires the Windows Update service to be running, but you can still set your server to “Never check for updates.”  By temporarily enabling the Windows Update service, I was able to install .NET 3.5 SP1 without a hitch.  I hope this helps anyone else using a VPS, because it sure gave me a little bit of a headache!