DevSum 2012 - Day 2 Summary

May 28, 2012 · Follow on Twitter and Mastodon conferences

This is a short summary of the second day of the DevSum 2012 conference in Stockholm, Sweden.

Keynote - Aral Balkan - A Happy Grain of Sand

Boy, what a great session!

Aral started his talk with a coordinated introduction. As someone who lives for creating things that make people happy, he presented some examples from his daily life that makes him…sad:

  • A hotel toilet, where a flush button outside of the natural grouping, makes it hard to understand how to flush. Proximity implies relation and out of sight = out of mind.

  • A hotel toilet (another one), where the flush button and toilet fulfill their purposes, but where the lid hides the flush button once it opens. An experience is as strong as its weakest part.

Various examples of elevator button panels followed, including the pun Schindler’s lift. He also went into a long description of his washing machine, which together with the dishwasher most often have horrible user interfaces.

According to Aral, we do not need more things, we need things that work better. He compared the Swedish Arlanda Express ticket machine (which is terrible) with Oslo’s solution, which is beautiful and just consist of a card swipe and a small digital screen.

Sometimes the best UI is no UI.

This session was full of great examples of good and bad design. I will not stack up on great quotes, but rather advice you to see a presentation with Aral if you get the chance.

Scott Allen - Modern Javascript

This was a highly interesting session, where Scott talked about great JavaScript features and discussed scope, constructor functions, prototypes, closures, getters/setters, modules etc.

The talk was almost entirely code-oriented and thus hard to summarize, but make sure to watch it if you happen to find the video.

Steve Sanderson - Building mobile web applications with Node.js

This talk focused on Node.js and how to get started, then on to mastering it. Steve started from nil and used Node to create various types of applications and APIs, using extensions like Express to simplify it, handling static content, templating etc.

Steve improved a mobile web application with responsive design, viewport settings etc. and showed how to use the Opera Mobile Emulator to simulate various device types, including limiting the bandwidth of the device and using cache manifest files to improve caching.

Although a bit overwhelming, it was very interesting. Steve took many of the things that I have been meaning to look at and presented them in a way that makes me want to use them now, now, NOW!

Johan Lindfors - Windows Phone Best Practices – part 1 / 2

This was the first part of two, in which Johan and Michael talked about Windows Phone development.

Johan covered the following areas:

  • Design / development
  • MVVM
  • Testability
  • Localization
  • Security
  • Design


Regarding design, Metro apps can (mostly) be defined as minimalistic, text-based and living. So, how can you make your apps behave like the native apps that flow smoothly and look stunning?

Johan first mentioned the Windows Phone design grid, which is 24, 25, 12, 25, 12 … 25, 24 grid on which the Metro UI is built. Use it and you will find that your apps will “feel right”.

Johan then talked about design data, which replaces real data with mock data during the design phase. With design data, developers and designers can see data in the UI when developing and designing.

Before animations, he recommended the Windows Phone Commands VS Extension, which lets you download and run apps on the simulator. No roms are needed - apps can be downloaded directly from the marketplace.


Johan mentioned three toolkits that simplify working with animations on Windows:

  • Silverlight toolkit for WP (resource demanding)
  • WP7 contrib (based on adding base classes to your views)
  • Metro in motion, which is a series of blog posts.

Johan used MiM in this talk, and it looked great.


Johan compared several frameworks for working with Windows Phone and MVVM:

  • the built in framework in Visual Studio
  • MVVM Light
  • MVVM Excalibur
  • Prism
  • DYI (Do It Yourself - requires more code, but gets you exactly what you need).

Johan then suggested keeping the framework simple, using the singleton pattern instead of dependency injection, how to handle communication between the view and the view model etc. He also told us that the SmartObservableCollection might help us out.


Unit testing on mobile devices can sometimes be tricky. Johan talked a bit about how to solve this using portable class libraries, and also adviced us to embed the required tools to make CI work.


Johan added some Swedish and English texts in resource files, then showed us how to translate strings, making the application translate all translatable textual content as soon as a user selects another language. With localization, a general advice is to design text boxes 40% larger than they would be with English texts, in order to support longer texts.


Johan talked about security, including how you can download .XAP files from the marketplace, change the extension to .zip, unzip them and get access to whatever the file contains. Feels like a call for source code obfuscation.


A very(!) thorough session, that I cannot help would have been even more awesome if I had actually gotten around to developing for Windows Phone. Then, a session like this would have been extremely helpful. Now, it just rocked ;)

Michael Björn - Windows Phone Best Practices – part 2 / 2

Michael continued the Windows Phone talk and focused on prestanda, communication, background agents, push notifications and synchronization.


Michael showed us how Windows Phone’s two UI threads - a UI thread and a compository thread - can be used to achieve non-blocking animations. He demoed this with different progress bars - the built in one, an external one and another native one that animates in the very top of the screen.

He also demonstrated moving stuff around and how textures can cause the phone to redraw small or large parts of the screen, depending on how you manage it. Some other general advices were that:

  • If you have a long-running operation, use Thread.Sleep(1) in the operation to avoid blocking the compository thread.
  • In order to avoid memory leaks when working with textures, disable timers that run in views.
  • In order to reduce startup time (long startup times can get your app refused), consider how you load images, coordinates etc.


When it comes to picking between WebClient and HttpWebRequest, Michael strongly advocated the latter. Since the WebClient class runs on the UI thread, using it may come back to haunt you later on. HWR takes a bit more setup, but is worth it.

Also, instead of having many calls with little data, it is better to bundle the calls into one. Otherwise, the phone will open and close the connection for each call, which makes the antenna toggle a lot and drain battery life in no time.

Background agents

Since WP lacks the possibility to create multi-threading apps, you must resort to background agents. However, the two types of available agents (periodic tasks and resource intensive tasks) have great limitations and are non-reliable. You should never resort to background processes and use other means to achieve critical tasks.

Push notifications

There are not that many WP apps out there today that use push notifications, but it is quite easy to setup. Since push notifications can affect your live tiles, using them can make your app feel a lot more alive, even when it’s not being used.


Today, the WP framework only has support for one synchronization framework - the Sync Framework. Michael did not talk much about this, but mentioned, that when stuff is synced for your app, the content will be placed in the isolated storage.

Morten Nielsen - From Continuous Integration to Continuous delivery

This talk focused heavily on the specific challenges for Morten’s company, where the release processes caused updating customer installations to take a day or so. Per customer.

With CI and CD, they managed to drastically reduce this time…which is the kind of effects you tend to get by automating tasks :)

To sum it up, using Hudson, Morten and his team now have builds that are:

  • Fast to create and apply
  • Deterministic
  • Logged / audited
  • Unsupervised

My advice to Morten is to focus less on very company (especially team) specific examples. I expected a CI/CD session and got a case study. Nice, but not what I was after.

Discussions & More

Please share any ideas, feedback or comments you may have in the Disqus section below, or by replying on Twitter or Mastodon..

If you found this text interesting, make sure to follow me on Twitter and Mastodon for more content like this, and to be notified when new content is published.

If you like & want to support my work, please consider sponsoring me on GitHub Sponsors.