Microsoft MVC5 + ASP.NET Identity with Telerik OpenAccess Awesomeness

Microsoft’s new membership system ASP.NET Identity is (IMHO) absolutely fantastic! The system itself comes installed as the default provider in Visual Studio 2013 for ASP.NET templates and can be downloaded for use within Visual Studio 2012.

ASP.NET Identity is the next step in the evolution to what was previously known as the SimpleMembership and Membership providers. Previous versions have been extremely brittle; this new approach simplifies and segments the membership provider in such a way that it makes it incredibly easy to plug-in and remove the components as you require and more importantly simplifies development when implementing a separation of concerns in a Service Oriented Architecture. Anyone who has attempted to achieve this with WCF and the previous membership providers will know the difficulties and lack of clear process through which to accomplish this. Another major benefit is that the framework implements all the .NET Async goodness by default. While you could always create your own Async wrappers in the older providers it makes things easier to see this functionality implemented for you.

The problems with the original ASP.NET Membership provider were:

  1. Designed for SQL Server and you can’t change it. You can add profile information, but the additional data is packed into a different table, which makes it difficult to access by any means except through the Profile Provider API.

  2. The provider system enables you to change the backing data store, but the system is designed around assumptions appropriate for a relational database.

  3. Since the log-in/log-out functionality is based on Forms Authentication, the membership system can’t use OWIN.

As the Introduction to ASP.NET identity article goes on to say

Simple Membership did make it easier to customize user profile information, but it still shares the other problems with ASP.NET Membership, and it has some limitations such as:

  • It was hard to persist membership system data in a non-relational store.
  • You can’t use it with OWIN.
  • It doesn’t work well with existing ASP.NET Membership providers, and it’s not extensible.

So, how is ASP.NET Identity different? Well…..

1. Storage backend is completely separate

Complete and total separation of the storage mechanism used within the provider. For example implement your own storage back-end but use something other than the entity framework? Or you want to use some type of NoSQL implementation (such as Azure Tables) or even just an Xml file? Well all you need to do¹ is implement the IUser and the IUserStore interfaces, inject it into your UserManager and you’ve got your own custom storage provider.

Example.

protected void CreateUserManager()
{
    //Create a new user manager based on your User model, inject your UserStore and finally your ORM context
    UserManager<AspNetUser> Manager = new UserManager<AspNetUser>(new OpenAccessUserStore(new EntitiesModel()));

    //Now inject the validator
    Manager.UserValidator = new UserValidator<AspNetUser>(Manager)
    {
        AllowOnlyAlphanumericUserNames = false
    };

    //And set a property so you can use the UserManager when required
    this.UserManager = Manager;
}

//Public property to access your UserManager instance.
public UserManager<AspNetUser> UserManager { get; private set; }

Note that all you need to do to create a basic IUserStore and IUser is implement this.

public interface IUserStore<TUser> : System.IDisposable where TUser : Microsoft.AspNet.Identity.IUser
{
    System.Threading.Tasks.Task CreateAsync(TUser user);
    System.Threading.Tasks.Task UpdateAsync(TUser user);
    System.Threading.Tasks.Task DeleteAsync(TUser user);
    System.Threading.Tasks.Task<TUser> FindByIdAsync(string userId);
    System.Threading.Tasks.Task<TUser> FindByNameAsync(string userName);
}

AND THIS

public interface IUser
{
    string Id { get; }
    string UserName { get; set; }
}

2. General decoupling

Previously the provider implementation was extremely closely coupled between its various ‘bits’. The result would often cause confusion about which methods to invoke and under what circumstances (at least for me). Sometimes all you’d want to do is authenticate, sometimes you want to authenticate and sign in and sometimes you want to go straight to the database and just retrieve some bit of User info. Because of this change of implementation it’s been significantly easier to build a project using this provider and integrate it with WCF cleanly. In part this is because it’s now much easier to say, “Give me all claims for this user.” then “Authenticate the following claims.”. They are completely different subsystems and have no direct inter-dependencies.

3. Unit Testing

Clear interface based contracts between the various components of the framework make dependency injection and unit testing much simpler.

4. Claims based

While roles still exist within ASP.NET Identity the emphasis to a claims based authentication seams clear. Claims are useful because they abstract the individual elements of identity and access control into two parts. Unless you’ve got a reason not to, I suggest it makes sense to use claims rather than roles.


Finally I’ve gone ahead and built a generic Visual Studio project template extension which shows how to integrate MVC5 with ASP.NET Identity behind WCF webservices and utilizing Telerik OpenAccess ORM for the storage provider (Because OpenAccess kicks ass…) and deployed it to the Visual Studio Gallery for everyone to use or if you prefer just a straight download of the files.

The project itself demonstrates a number of things including:

  1. How to create a NetTCP WCF service without autogenerated proxy classes.
  2. How to use ASP.NET Identity in a WCF service
  3. How to use a different ORM as an alternative to the entity framework
  4. How to include System.ComponentModel.DataAnnotations in a portable class library for your Views.
    1. Which allows the ability to use them across your MVC, WCF, Windows Phone and Xamarin applications.

Hopefully someone else will find the template and source code useful!


  1. It’s not quite that simple (you might need to implement between 1 and like 10,000,000 interfaces), but almost!
Advertisements

Building a generic security framework

Its a mind blowing fact that despite the countless libraries and frameworks available for the ASP.NET platform there still doesn’t appear to be any options regarding a plug and play security framework.

This is now the second time I have had to build a security framework for my applications and each and every time it seems to prove just as frustrating as the last. For every generic implementation I am able to concoct and two steps forward that I take, I am subsequently pulled 3 steps back when I confront a requirement that simply doesn’t fit anywhere within my implementation.

Despite my issues I do however have some tips.

Use a generic approach

  1. It may initially seem an like overwhelmingly complicated problem to solve generically but it isn’t, just like most things in our field it needs to be solved using a staggered approach.
  2. Start by implementing only the most basic functionality (Edit and View for example) and slowly move to more complicated requirements (If this value is x this control is enabled, if this value is y this control is hidden). Slowly as you solve these problems generically, you will build an overall solution.
  3. Testing a non generic approach is nearly impossible and will cost you much more in the long run.

Use bitwise operations (flag enumerations)

  1. I have tried both an approach with and without flags and I can confidently say that the code in the latter is cleaner and more understandable (even if it does make DBAs go mad)

Create a separation of permissions in the following way

  • Groups
  • Page Level Permissions
  • Control Level Permissions
  1. Groups are your standard security groups as with any other application.
  2. Page level permissions are the permissions that are tied to a single page or a group of pages and will typically only require View, Create, Edit, Delete functionality.
  3. Control level permissions are usually where things get tricky, these can require any number of ‘truths’ to either enable or show the control and the logic to express this can become very convoluted. I approach this part of the problem with the idea of invoking defined methods (that are linked by control names) that perform the required validation returning either true or false.

The above separation will cover just about any scenario you can envisage and if implemented with even the smallest of foresight should be enough for the basis of any generic security framework. Having said that, if anyone wants to become another Telerik, go get started on this project immediately…

Persisting Telerik RadGrid grouping UI after postback

I have recently run into a problem where after utilizing the grouping functionality along with the RadAjaxManager my RadGrid grouping UI state was disappearing after rebinding data.

This is undesirable for a number of reasons but mainly because it is unclear to the end user if an edit performed to an entry in the grid has been successfully achieved. Not to mention frustrating to have to open the group again to verify.

After much searching to see if the RadGrid supported such functionality or not I was extremely confused with some forums and posts saying this functionality was built into the grid and others suggesting it was impossible to do.

Anyway, in the end I was unable to find a built in way to perform this and resorted to the hack demonstrated here.

Pity Telerik doesn’t appear to have built this functionality straight in.