I'm dubious about whether to suggest this, but something that enabled easier high availability for a SS Service would be great. This would maybe be more like a recipe than a feature, but possibly with features to help support it.

E.g.
We run a ServiceStack API Service which depends on redis and postgres, each with replication to secondary nodes which are readonly until we manually execute a failover command. Running multiple ServiceStacks on VPSs (or whatever) is easy, but having them co-ordinate is slightly trickier. So the type of thing I'm suggesting would include

• Digital Ocean recipe for running up multi-region VPSes with ServiceStack


• Endpoints for inter-node communication (for availability checking)


• DNS Setup guide (assuming this will be the simple HA technique)


• Each node having independent, but possibly co-ordinated config


• Support for "routing/proxying" back to master node (depending on DNS setup), so if slave node receives a request, it will execute as if on master, but with extra latency


• Live config reload (for automated failover) (e.g. change postgres connection string to x.x.x.x)

Imagine you have large DTOs that are ushared for GET, POST and PUT.

Some properties might only be used for display purposes, and cannot be updated (i.e. denormalized fields).

You might also allow the setting of a property on POST, but not on PUT. For example, you might not allow the Id of an object to be set on POST (because it will be the primary key identity insert), but if you want to update it, you can pass it in a PUT:

We currently are doing this using a custom attribute and interface on the models and model properties.

[ReadonlyRequest(ApplyTo.Post, null)]
public int? Id { get; set; }

Before each request we build up a list of warnings that we add to the request context. We also remove / set the default on any properties using this attribute.

After the request is processed, we automatically add a list of warnings to the ResponseStatus.

The calling client then knows based on the response, which of the data they have passed will simply be ignored.

When creating immutable objects, it's common to not have a default constructor. ServiceStack.Text supports this presently with the use of FormatterServices.GetUnitializedObject. However, GetUnitializedObject does not do any field initialization. The DataContractSerializer works around this with support of the OnDeserialized attribute; but why would we want to write the initialization code twice? Instead, We should change ServiceStack.Text to fall back to private constructors. Here is some proposed code.

    public static ConstructorInfo GetEmptyConstructor(this Type type)

    {

if NETFX_CORE

        return type.GetTypeInfo().DeclaredConstructors.FirstOrDefault(c => c.GetParameters().Count() == 0);

else

        var ret = type.GetConstructor(Type.EmptyTypes);

        if (ret != null) return ret;


        // attempt to use constructors with all default values

        var ctors = from ctor in type.GetConstructors()

                    let prms = ctor.GetParameters()

                    where prms.All(p => p.IsOptional)

                    orderby prms.Length

                    select ctor;

        ret = ctors.FirstOrDefault();

        if (ret != null) return ret;

        if (JsConfig.UsePrivateConstructor)

        {

            ret = type.GetConstructor(BindingFlags.Instance | BindingFlags.NonPublic, null, Type.EmptyTypes, null);

        }

        return ret;

endif

    }