Category Archives: Uncategorized

Bifrost; Getting back to it…

Its been a while since I wrote anything about Bifrost. In fact the last post I did was about me not maintaining it anymore. The thing is; its been an empty year for me personally since February when I announced it. I didn’t realize it until I was at a partner who wanted to dive deep on SOLID, DDD, CQRS, EventSourcing and more and we only had a couple of days to prototype something. We talked it over and we decided that doing Bifrost would get us there quicker… what a relief… I’m so glad we did that. All of a sudden it all became very clear to me; I need to continue the work – its just too much fun. I had a hunch, but didn’t see it all that clear. A few months back I started pulling things from Bifrost into a new project called Cratis and making it more focused. Never kind of thinking that it should go back into Bifrost.

So, what am I doing about it. Well, first of all; I took down the post announcing the stop in maintenance. It didn’t make sense to have it there when coming to this realization that I need to push on. The second thing I did – in order to get back into the mood and understanding Bifrost (even though I wrote most of it) again, was to start writing the proper documentation that it deserves. This now sits here. The next thing that will happen is that development will be picked up again.

From the top of my head, this is what needs to be done:

  1. Add a support for running on Azure in a distributed manner – with a working sample
  2. Clean up. Remove platforms not being used.
  3. Simplify code. Make it more focused.
  4. Modernise it. Make it run on .NET Core
  5. Rewrite JavaScript to be ES2015+
  6. Break it apart into many small GitHub projects that can be maintained individually

In between there might be some features that sneaks in. But the majority of new development will have to happen after these things have happened.

Alongside with it all; more documentation, more samples, more videos – just simply more. 🙂

Really looking forward to getting back into this and see what 2017 have in store for Bifrost work.

The Code Lab

I’ve been wanting to try out a new format, or at least a new format for me; live interactive webcast. I got inspired when I saw some ex-colleagues of mine started something called “Kodepanelet” and figured I had to do something similar. The goal is to have a one hour show on a regular basis where anyone can chime in on social media and ask technical questions and I’ll try to be as agile as possible answering, or probably preferably show. Some things I can answer live – other things might need preparation for a later show, or a follow up in a blogpost or similar. This part is still a bit fuzzy. To be honest, since this is all new and I’m trying something out – the format will more than likely change over time.

The concept is called “The Code Lab”;


I will be using the Facebook live video streaming system for this and you’ll find the Facebook page for the concept here.


First airing will be on Friday the 4th of November @ 13:00 CET (1PM), here.


What can you expect?

First of all, this is supposed to be for software developers. And I can’t do more than I know or can easily acquire knowledge on. My background stretches from games development including graphics – low level programming to line of business applications and architecting highly available durable and scalable systems. I also have a knack for process and how to build teams. I consider myself pretty open-minded, and as a result I’ve always been all over the place when it comes to platforms. These days I find that I am probably most experienced in macOS and Windows, but an apprentice in Linux – trying to gain more experience there. I try to be polyglot in languages, but have most experience in C/C++, C# and JavaScript. Thrive in the backend as well as the frontend and loves my patterns and eat my SOLIDs for breakfast. I focus a lot of my time on cloud, and specifically Azure. Containers, MicroServices and in general decoupled software is something I am really passionate about.</>

How can you help?

Content is always king – I’m hoping people find this interesting and want to jump in with good questions. To get the topics right, there is a survey on Facebook

Health warning

If you’re out to troll me, I’ll try to ignore you. 🙂

New and interesting challenges

First of July I’m starting @ Microsoft Norway in the Developer Experience (DX) team. My official, on the paper, title is Technical Evangelist – which might sound scary enough. But in reality it’s a multifaceted position ranging from sure enough evangelism in the sense of talking about Microsoft products and promotoing these, to advisory for clients, community work, blogging and more. My particular role will be geared towards the cloud; Azure, but I will be sure to keep in touch with the entire stack – as I really love keep on top of things. I spend a lot of time in front of the computer both work and non-work related, so I’ll be sure to keep up.

I’ve worked closely with Microsoft since 2001, since 2008 I’ve been a Microsoft MVP and at times its felt like I’ve been an employee without having the privileges an employee has. So when my new boss said to me a couple of weeks ago; “Welcome home..” – that is the feeling I actually do have.

The timing for me is perfect, both on a personal level, professional level and seeing all the cool things that really excites me that is going on at Microsoft these days. I’m bursting with joy for how Microsoft has turned around and really looking forward to engage more in the coming years with this.

How does this impact other things. Well, I’ve had a company on and off since 1997 called Dolittle. At first as a way of picking up freelance work and the last few years more focused. I am closing down the company. Keeping the brand though, with domains and all, but putting it to sleep as a company. For other things, the opensource projects I’m involved in; Balder, Bifrost, Forseti and more. I’ll keep on working on them when I have a chance. Maybe not as focused as before, as I’ve had the pleasure and luxury of being able to maintain the projects at work for the last few years. They’ve been with me for years and are babies to me, so it would be hard to let them go. Besides, they serve a great purpose for me in keeping up with the whats going on the world of development. I am a developer after all and if I am to do a good job talking about it, I need to maintain my knowledge.

I’m super excited and honored to become a member of the team. Really looking forward to it.

Improving Angular experience with some convention magic

Disclaimer: I’m a line of business app developer – not a Web site developer. My view on things is colored by my experience with building large enterprise applications that has larger teams of of developers working together and needs to keep velocity and code quality at a very high standard through years of development.

I’ve been on a short assignment with a client and they wanted to establish ways for them to work with their Web development. They’re a .net shop with little experience in Web – only thick client technologies like WPF. They had very few requirements, but they wanted to go for things that are fairly established and had a guy that had done some AngularJS. For me it was the first time I did Angular in any structured way – previously I’ve just dabbled with it, basically with the intention of supporting it for Bifrost. The first thing that struck me with Angular is the explicitness of just about everything and how everything needs to be configured with code. Obviously, I’m not going to claim to be an expert in Angular, and I’d love to be corrected if I’m wrong. But from my little experience I started itching, and I had to scratch the itch.

Routes and conventions

In Angular, with the Angular Route extension, routing is one of the things you can do. A fairly simple and consistent API. One typically ends with a pattern as follows:

var application = angular.module("MyApp", ["", "ngRoute"]);
application.config(["$routeProvider"], function($routeProvider) {
$routeProvider.when("/some/route", { templateUrl: "/some/path/to/view.html", controller: "MyController" });
application.controller("MyController", function() { /* Put your controller code here ... */ });

On one side one tells what to do when a route occurs – pointing it to a view and a controller that represents it. Then we need to configure the controller by its name and then point it to a function that represents the controller.
My claim is that looking at your app, you either have formalized a pattern or you have a pattern by accident for how these things are put together. This is a great time to formalize it by creating something that represents it and can automate it as a convention.

For instance, you probably will see that for most parts of your app there is a relationship between the routes and where things are placed on disk. In my experience, routing is more often than not something the end users really don’t care about. I do have a feeling that we put too much thought and effort into something like this, while the end users will just copy / paste / click links and don’t care how they look. With this in mind and if there is more in reality a correlation between routes and disk location, we can automate this whole thing.

Another aspect of Angular that is really useful are the directives, but again as with routes and controllers, one has to set this up very explicitly. This is something that could easily be automated. For instance, you could have a folder in your frontend project called Directives and every folder within it represents a directive by the folder name, within this a directive could then be represented by View.html for the view a Controller.js for the controller and a Link.js for the link part.

Proxy generation FTW
Something we’ve had great success with from our Bifrost development is proxy generation. With backend code written in a different language than the frontend, its just great to augment the frontend with code generated to make the transition between the two less painful. But regardless of the divine of having 2 languages in your system, generating code for automation can really be a lifesaver. With a fixed convention, developers on your team gets less options. You might be arguing that is a bad thing, but I argue its a very good thing. If you’re the only one on your team or you’re two guys – you can probably cope with full flexibility. But applying a regime makes it easier to do things right – or according to the regime, at least. And that should be a good thing. Another benefit of defining regimes is that in some cases when automating things and generating code you get the opportunity for doing cross cutting concerns. With a regime in place for routing for instance and pointing it by convention to a controller that matches the view, you could inject a man in the middle controller that could do a lot of interesting cross cutting concerns, for instance logging, error handling, security or other more domain specific concerns.

Code please
We are going to build support for Angular into Bifrost, and with it we will provide configurable conventions for routing and directives.

As part doing a Visual Studio 2015 Deep Dive for Microsoft recently I created a sample for Visual Studio 2015 showing off how to do all this, but now just using JavaScript and the new Grunt and NodeJS support coming with Visual Studio 2015. You can find the entire sample that uses SignalRKarmaJasmine over at GitHub here.

Bifrost and Proxy generation

One of the things we consider to be one of the most successful things we’ve added to Bifrost is the bridge between the client and the server in Web solutions. Earlier this year we realized that we wanted to be much more consistent between the code written in our “backend” and our “frontend”, bridging the gap between the two.  And out of this realization came generation of proxy objects for artifacts written in C# that we want to have exposed in our JavaScript code. If you’re a node.js developer you’re probably asking yourself; WHY..   Well, we don’t have the luxury to be writing it all in JavaScript right now, but it would be interesting leveraging what we know now and build a similar platform on top of node.js, or for the Ruby world for that matter – but thats for a different post.  One aspect of our motivation for doing this was also that we find types to be very helpful; and yes – JavaScript is a dynamic language but its not typeless, so we wanted the same usefulness that the types have been playing for our backend code in the frontend as well. The types represent a certain level of metadata and we leverage the types all through our system.

Anywho, the principle was simple; use .net reflection for the types we wanted represented in JavaScript and generate pretty much an exact copy of those types in corresponding namespaces in the client. Namespaces, although different between different aspects of the system come together with a convention mechanism built into Bifrost – this also being a post on its own that should be written :), enough with the digressions.

Basically, in the Core library we ended up introducing a CodeGeneration namespace – which holds the JavaScript constructs we needed to be able to generate the proxies we needed.


There are two key elements in this structure; CodeWriter and LanguageElement – the latter looking like this:

public interface ILanguageElement
    ILanguageElement Parent { get; set; }
    void AddChild(ILanguageElement element);
    void Write(ICodeWriter writer);

Almost everything sitting inside the JavaScript namespace are language elements of some kind – to some extent some of them being a bit more than just a simple language element, such as the Observable type we have which is a specialized element for KnockoutJS. Each element has the responsibility of writing themselves out, they know how they should look like – but elements aren’t responsible for doing things like ending an expression, such as semi-colons or similar. They are focused on their little piece of the puzzle and the generator will do the rest and make sure to a certain level that it is legal JavaScript.

The next part os as mentioned the CodeWriter:

public interface ICodeWriter
    void Indent();
    void Unindent();
    void WriteWithIndentation(string format, params object[] args);
    void Write(string format, params object[] args);
    void NewLine();

Very simple interface basically just dealing with indentation, writing and adding new lines.

In addition to the core framework for building the core structure, we’ve added quite a few helper methods in the form of extension methods to much easier generate common scenarios – plus at the same time provide a more fluent interface for putting it all together without having to have .Add() methods all over the place.

So if we dissect the code for generating the proxies for what we call queries in Bifrost (queries run against a datasource, typically a database):

public string Generate()
    var typesByNamespace = _typeDiscoverer.FindMultiple&lt;IReadModel&gt;().GroupBy(t =&gt; t.Namespace);
    var result = new StringBuilder();

    Namespace currentNamespace;
    Namespace globalRead = _codeGenerator.Namespace(Namespaces.READ);

    foreach (var @namespace in typesByNamespace)
        if (_configuration.NamespaceMapper.CanResolveToClient(@namespace.Key))
            currentNamespace = _codeGenerator.Namespace(_configuration.NamespaceMapper.GetClientNamespaceFrom(@namespace.Key));
            currentNamespace = globalRead;

        foreach (var type in @namespace)
            var name = type.Name.ToCamelCase();
                .WithType(t =&gt;
                                .Variant("self", v =>; v.WithThis())
                                .Property("generatedFrom", p => p.WithString(type.FullName))
                                .WithPropertiesFrom(type, typeof(IReadModel)));
            currentNamespace.Content.Assign("readModelOf" + name.ToPascalCase())
                .WithType(t =>
                                .Variant("self", v => v.WithThis())
                                .Property("name", p => p.WithString(name))
                                .Property("generatedFrom", p => p.WithString(type.FullName))
                                .Property("readModelType", p => p.WithLiteral(currentNamespace.Name+"." + name))

        if (currentNamespace != globalRead)

    return result.ToString();

Thats all the code needed to get the proxies for all implementations of an interface called IQueryFor<>, it uses a subsystem in Bifrost called TypeDiscoverer that deals with all types in the running system.

Retrofitting behavior, after the fact..

Another discovery we’ve had is that we’re demanding more and more from our proxies – after they showed up, we grew fond of them right away and just want more info into them. For instance; in Bifrost we have Commands representing the behavior of the system using Bifrost, commands are therefor the main source of interaction with the system for users and we secure these and apply validation to them. Previously we instantiated a command in the client and asked the server for validation metadata for the command and got this applied. With the latest and greatest, all this information is now available on the proxy – which is a very natural place to have it. Validation and security are knockout extensions that can extend observable properties and our commands are full of observable properties. So we introduced a way to extend observable properties on commands with an interface for anyone wanting to add an extension to these properties:

public interface ICanExtendCommandProperty
 void Extend(Type commandType, string propertyName, Observable observable);

These are automatically discovered as with just about anything in Bifrost and hooked up.

The end result for a command with the validation extension is something like this:

Bifrost.namespace("Bifrost.QuickStart.Features.Employees", {
    registerEmployee : Bifrost.commands.Command.extend(function() {
        var self = this; = &quot;registerEmployee&quot;;
        this.generatedFrom = "Bifrost.QuickStart.Domain.HumanResources.Employees.RegisterEmployee";
        this.socialSecurityNumber = ko.observable().extend({
            validation : {
                "required": {
                    "message":"'{PropertyName}' must not be empty."
        this.firstName = ko.observable();
        this.lastName = ko.observable();

As I started with in this post; this has proven to be one the most helpful things we’ve put into Bifrost – it didn’t come without controversy though. We were met with some skepticism when we first started talking about, even with claims such as “… it would not add any value …”. Our conclusion is very very different; it really has added some true value. It enables us to get from the backend into the frontend much faster, more precise and with higher consistency than before. It has increased the quality of what we’re doing when delivering business value. This again is just something that helps the developers focus on delivering the most important thing; business value!

Touring south of California

In February I will be in south of California doing a couple of talks at different venus; 4 in total. 2 of them has been announced already here and here, when I get the links for the two remaining ones I will update this post with the details, so stay tuned. Thanks to Kim Schmidt from vNext_OC for making this happen and for asking me to drop by.

Basically the talks will be on two different topics.

Below you’ll see the different topics with the synopsis of them. So, if you’re nearby these venues, don’t hesitate to stop by. Also worth mentioning, Charles Petzold is giving talk at one of the user groups I’ll be doing a talk at, be sure to not miss it as well – more details can be found here.


Let’s focus on business value

Creating software is very hard, a lot of practices has been developed over the years to accommodate this and make it easier. Some of these are DDD (Domain Driven Design), SOLID, BDD (Behavior Driven Development) and concrete architectural patterns such as CQRS (Command Query Responsibility Segregation) and MVVM (Model View ViewModel) came as reactions to these practices. Einar will take you on a tour through all of the above mentioned subjects and show you concretely how you can achieve true developer productivity by applying all these. By utilising an open source framework called Bifrost, Einar will show end to end how these practices and patterns can come to life and can really let you as a developer hit the ground running and at the same time capture and deliver true business value without sacrificing code quality. All this and also cloud ready!

It’s primetime, a JavaScript story

It’s pretty fair to say that JavaScript is not a fad; it is by far
the most widespread programming language out there and also the most available runtime we have, ranging from toasters to the web, and even to the backend development through Node.js. Its probably also fair to say that we should really embrace it and start treating it like a first class citizen of our day
to day work. In this talk, Einar will take you on a tour of how you can work
with JavaScript with similar patterns you’re already used to from the rest of
your server code. Writing tests or specifications that proves your code is also
important, Einar will show how to get started with this and how you can achieve more testable JavaScript by applying patterns like MVVM (Model View ViewModel) using KnockoutJS

MCTS 70-507 guide winners

Last week I had a contest on the blog here to win the Silverlight 4 development certification guide book, and the lucky winners are, based off the comments : 


Congratulations to you both! 

Slides + URLs from NNUG talk 29th of February 2012

Yesterday, Pavneet Singh Saund and I held a talk for NNUG Vestfold about modern Web development using the MVVM pattern and compositional UIs.

The slides can be found here.

In addition we talked about a few things, these are the resources of the talk :


I did mention and tried to demo another site I’m working on, which was down at the time – but up now :

Geekrider – it shows navigation of features, using Bifrost.

and lastly, the actual demo app we did can be found here. We did also show the inkFinder that was developed at Komplett that uses Knockout and MVVM, which you can find here.

Removed ratings

Recently I’ve had a couple of strange ratings on posts here. People, in plural, rating simple posts that basically just announce a talk to a 1 out of 5 and not leaving a comment. That just messes with my head. There is no way I can improve posts when there is just a bunch of negative rating but no comments. As a consequence, I decided to remove ratings all together – hopefully forcing people to engage in a conversation instead.

If there is something you don’t like in a post, or the entirety of a post – please don’t hesitate to post a comment. I’m very open to listening to feedback, but I can’t just read into the lack of a star and figure out what I’m doing wrong.


Courses, fall 2011

I’ll be doing a couple of courses this fall for Programutvikling, focusing in on LOB development in general with the subject of data access and Azure.

The first course will provide a backdrop for those wanting to take the MS 70-516 exam, focusing in on data access in your LOB apps. Read more about it here.

It has been scheduled for the following dates :

  • 24th of August
  • 5th of October
  • 23rd of November

The second course is a crash course in Windows Azure and will provide you with the capability of developing with Azure and take the right choices to make your application scale and become reliable in the cloud, you’ll find the details here.

It has been scheduled for the following dates :

  • 19th of October
  • 9th of November