I would like to see the ability to persist an entire object graph in one DB call. Currently it is way to chatty with the DB which makes it really slow to persist.

System.Data.Mapping.StorageMappingItemCollection doesn't provide useful public API.

If you have a class like

public class CustomerAddress
{
public virtual Customer Customer { get; set; }
public virtual Address Address { get; set; }
public virtual string AddressType { get; set; }
public virtual DateTime ModifiedDate { get; set; }
}

you should be able to create a composite id like this

HasKey(a => new { a.Customer, a.Address });

without defining CustomerId and AddressId properties

Allow mapping of database views without requiring the Key Attribute

When mapping entitis to Views, it´s common that no Key can be set, bescause the view does not provide it and you can´t map it. The idea is:

* Make it possible to map to Views (or tables) in a read only fashion that does not require the Key Attribute

Extensible first and second level cache.

It'd be great if entity framework had an integrated tool such as the Profiler from Hibernating Rhinos (http://hibernatingrhinos.com/products/efprof).

I am aware that you can view the SQL generated for simple select statements by hovering over the DbSet member during debug but it would be nice to have a tool like this built into Visual Studio or that simply outputs the generated SQL to the debug console.

This should be configurable / able to be disabled.

using the include() with a 1 to many relationship, the data are loaded in the list without any specific order, forcing often to reload directly specifying the sort order

Give developers the power to customize and extend the Entity Framework!!!!

Please include caching in EF. I have used Jarek Kowalski’s provider but it didn't cope with Stored Procedures or

Like datatable.Select(Filter) , we need to query what's already loaded in the datacontext using LINQ , something like a disconnected mode , this would be really gret feature , its so difficult to do right now

Allow specifying an interceptor (Castle, LinFu, Sprint.NET, etc) for modifying materialized entities. This allows injecting interfaces such as INotifyPropertyChanged, IEditableObject, etc.

[Column("OwnerId")]
public virtual User Owner { get; set; }
some interest was expressed by the team on this so it may be in...

Provide a method on the Database class to remove all data from the database without knowing anything about the underlying physical database schema, relationships and constraints.

This is critical in order to effectively test EF related code. While one way is to recreate the database every test, it has security and performance implications.

If you perform business logic in the entitys OnPropertyChanged event handler, eg. calculating some related properties or changing a state property when certain properties are changed, your code is also run when the entity is fetched from the database. This is not a real change of the entity, it is just being loaded. It would be nice if the EntityObject class had a property bool IsMaterializing that was true so you could break out of the OnPropertyChanged handler.

Provide an Entity Framework Factory similar to the ADO.Net Provider which allows easy swapping between different Entity Framework models such as SQL Server, Oracle, MySQL, DB2 etc. simply by providing the appropriate database provider information in a manner similar to the current ADO.Net Database Provider Factory.

There should be a way (perhaps by implementing an interface), to add fluent code separately to each entity instead of adding it all in the OnModelCreating method.
The OnModelCreating base method should call all the entities that implement that interface and call OnModelCreating(context) of it.
In this way the bunch of fluent code in the main context, especially if it's a large one, would tidy up in the individual entities.
Of course, this is already simply doable, but I think this should go out the box.

It never works when your conditionals are based on a int column so where Type=1 etc. Only seems to work for when some column Is Null. It complains about mapping when you try and do this. Seems impossible to work around