Oredev 2012: Day 2 Summary
12 Nov 2012
My second day at Øredev 2012 was amazing! When it was over, I had to sit down and take it all in. There were so many great talks, and I still had to skip many that I wanted to see.
Reginald Braithwaite – The Rebellion Imperative
After stating that wealth breeds ineffectiveness, that the powerful enforce stasis and that we are not in the business, we are the business, Reginald quoted some passages from “Marketing Warfare”, which defines the four sustainable positions in a market (there are other positions, but they are not sustainable):
- The leader
- The rival
- The innovator
- The 99%
These positions have different strategies:
- The leader defends - when Microsoft talks about their SharePoint Enterprise Services Database Storage Platform, they are defending their position, simply by saying “yeah, we’ve got that too…stay with us”.
- The rival attacks - when Apple is considered to be the cool, Google makes a point of Android being for the people who do not have to be cool.
- The innovator disturbs - by coming up with new ways, new models, new markets.
- The 99% rebels - and must watch out for the big trap of trying to play the role of the innovator. Only one (the one who succeeds) can be the innovator. Many can rebel.
Since the theme was rebellion, Reginald gave us some rebel advice. For instance, rebels must not be emotionally attached to whatever they are doing, and be ready to drop things that do not work. As an example he mentioned Audion, who abandoned their mp3 software when iTunes was released.
I found this session to be really fun and thoughtful.
Henrik Kniberg – Lean from the Trenches
Henrik’s talk was a case study of introducing lean processes at the Swedish Police. It was very interesting, but is kind of hard to summarize, so I’ll keep it short.
Henrik’s advice when you start out in a new project and don’t know what to do, how things work etc. is tp visualize what’s going on. By asking people and putting it all up on a board, you quickly get a clue of what’s going on. Also, by discussing the project, organization etc. with others in front of that board, they may learn new things as well…or point out things that are just plain wrong.
Henrik and his team introduced incremental delivery. Optimizing for flow instead of resource utilization (using a traffic jam analogy), features were only accepted if they had customer value, if they were estimated (they only used small, medium and large as estimation values) and if there were room for acceptance tests.
Henrik talked about so many things. It was all very interesting, but I think that it’s better that you watch the video, since I wouldn’t do it justice by extracting small bits from it. Instead, I’ll wrap it up with Henrik’s most critical factors for project success: co-location, incremental delivery and user involvement.
Johan Lindfors – Windows Phone Development Best Practices
I actually attended this session earlier this year, at DevSum, so I have already written about. The reason I decided to watch it again, is that Johan surely will have updated the presentation for Windows 8 and Windows Phone 8.
And he sure had. First, he talked about how strategies for estimating development time, how to make your app stand out from iOS and Android apps (live tiles, pins etc.) and how to handle beta testers. For instance, by measuring and awarding the most active beta testers, you can get more out of them instead of ending up with “testers” who just want your try latest app.
Just like last time, Johan had divided the talk into the following sections:
- Data management
- The Marketplace
I am super-impressed that he actually did cover them all.
When managing performance, focus on the perceived performance. Even if a task takes a long time, make it appear fast by doing stuff in background threads and keeping your UI smooth. Use a splash screen when starting your app and make sure it starts fast. If not, you can make the static splash shift to an animated one that shows what is happening. Johan also implored keeping the fill rate (can be done with a MemoryDiagnosticHelper addon) below 3, removing event subscribers (otherwise, they will not be disposed) and force garbage collection, since it is not forced by default.
For communication, make sure that your app use the static IsNetworkAvailable property and adapt according to it. Johan talked about async patterns, timeouts and retries and how to use the simulation dashboard (new) to simulate connection speed, lock the screen etc.
For data management, Johan talked about settings, state and how to persist. Like me, he almost never uses a database, but instead serialize objects and store them in the isolated storage. I find this to be a very effective and nice way to handle data on the phone.
For design, Johan demonstrated the grid (included by default in WP8) and how to adjust your views so they align with the grid. He showed how to mock data in design mode using XAML sample data, a fake data provider and by simply faking the entire view model.
When it comes to animations, I find this to be a lot less fun in WP7 than it is in iOS. However, I will once again checkout the resources Johan demonstrated, namely Silverlight Toolkit for Windows Phone, WP7 Contrib (better animations than in the former) and Metro in Motion.
Finally, to sum up the rest - Johan mentioned the same MVVM frameworks as he did before and recommended us to use the SmartObservableCollection, since it can disable events when large collections are updated. For Localization, he told us to remember to localize the Marketplace material and showed us how to localize an app without reloading it. For testing, he told us to use the Portable Class Library. For security, he advised us to obfuscate our code before sending an app to the Marketplace, since apps can be “unzipped”.
Phew, that was a lot, and I still did not cover everything. It was a really great session - a must-watch if you’re into WP8 development!
Steve Sanderson – Build Web Apps Faster
Steve Sanderson always entertain me extremely when he live coded. This time, the focus was on how to build web applications faster.
Steve talked about how, though we have so many more great tools, technologies and methods for developing systems today than we had a few years ago…well, it just becomes too much. The last two systems I’ve built in .NET have been focusing more on the system architecture and what’s going on behind the scene, than to actually create a working web app.
Maybe there is another way?
For many of my open source and web app presentation sites, I have begun throwing the above away and just create simple one-pagers, sometimes just in HTML without any backend whatsoever. However, going all that way has felt a bit…extreme, so I was happy to watch Steve demonstrate something in-between.
Steve showed how we can create web applications using a static site generator called DocPad (other alternatives are Jekyll, Hyde and Punch). This lets you use site templates, partials etc. then compile it all into a static site that just consists of html, js and css.
Steve used Knockout for model binding and CoffeeScript to inject things into the template. It looked really slick.
Then, he blew me away as he demonstrated how to use the new Windows Azure Mobile Services to avoid having a backend. While we almost always need a backend, do we really need one on a server of our own? Windows Azure Mobile Services can handle data management, authentication, push notifications, monitoring, scalability etc. which most often take a lot of time for us to build.
Steve then went into the Azure Management Console and demonstrated how to create a cloud-based database and how Azure can generate mobile application stubs for us. For instance, iOS apps are already supported, and can be generated with a single click. Android is on its way as well.
I my opinion, this talk was by far the most exciting and inspiring one at Øredev 2012! Steve’s was super slick and Azure Mobile Services seems awesome. This talk is a must see!
Oren Eini, Alistair Jones and Chris Harris – NoSQL FTW
In lack of other interesting sessions, I went to this talk and was (as expected) gravely disappointed. It consisted of three lightning talks about RavenDB, neo4j and CouchDB. I expected demonstrations. Instead, I got a sale pitch.
A bit of cred to Alistair Jones, however, for his cool neo4j demo that in detail showed how to traverse a graph database.
Steve Klabnik – Designing Hypermedia APIs
Starting his talk with general API discussions (flexibility vs. stability, decoupling, law on demeter etc.), Steve moved on to a Hypermedia-specific discussion.
According to Steve, the problem with many apis out there, is that they are based on out of band information, which means information that is not included in the information being exchanged between the server and the client. These kinds of apis require good documentation outside of the api and that developers read up on how to interact with them.
Hypermedia apis, however, provide (or at least should) its clients with all data needed to interact with the api. For instance, the GitHub api returns pagination links in the header. Besides defining media types, accepts etc. the header is a great way to place this kind of additional information.
Steve also demonstrated how to build Hypermedia APIs, using media types etc. It was a good session, but since I have seen most of it already, I should have gone to another session. If you want to know more about Hypermedia apis, though, it’s a good watch.
Shane Morris – Prototypes, prototypes, prototypes
Shane begun by asking us five reasons why we use prototypes…and we nailed them:
- Validate the concept in concrete terms
- Try out ideas with low risk
- Identify issues before it’s too late
- Sell the vision to stakeholders and investors
- Bring the team together with a common thought
He had some fun architectural to prove his point, like how applications just like buildings break at the joints. He discussed the pros and cons of the five points above and talked a bit of different types of prototypes.
During development, we will begin to move from static prototypes (e.g. wireframes) to more dynamic ones (e.g. dummy web sites). Since high fidelity prototypes are expensive, start off with low fidelity ones and make sure that the prototypes are changeable, accessible and can evolve over time.
This talk was very good, but should be watched rather than read.
Jesus Rodriguez – Rocking the Enterprise with the Kinect Experience
This was the kind of presentation I expected last year, when Tim Huckaby gave us more anecdotes and brief demos, than code that demoed with Kinect’s capabilities. Jesus started right away, showing us how to hack the new Kinect for Windows.
He then talked about gamification and how Kinect for Windows is not primarily for games, but rather for natural user interfaces, which is one of the important pillars in the next generation of user interfaces.
You need to design nuis differently than when designing games. For instance, the unexpected is exciting and fun in game context, while in a nui it is frustrating.
Microsoft created Kinect for Windows after some hackers managed hacking the XBOX 360 Kinect to use for development. Still, it is not just a simple port. It has a new, improved sensor, which provides you with:
- Skeletal tracking
- Depth information
- RGB data
- Facial tracking
- Speech processing
Jesus talked about these various capabilities, demonstrating how write code that communicates with them. When you do, just create an instance of the sensor, then enable each capability. Seems pretty straightforward.
For skeletal tracking, Kinect will support 20 joints when standing up and 10 when
sitting down. It can track 6 persons at once, although only two are tracked with
joints. The other will only be represented with a position. Remember to transform
the tracked joints with
TransformSmoothParameters, to get smooth transitions.
The depth analysis component is used heavily by the other depth-related sensors (the infrared data stream, for instance). It measures the distance in mm from the Kinect Player and can be set to work in a default mode or near mode. The mode should depend on how the user is expected to interact with the ui.
Jesus also demonstrated the RGB data stream (it’s not that exciting, it’s “just” a camera) and the speech and grammar recognition. The speech part was a killer! Besides working very well with initial grammar, you can register your own grammar as well. For instance, Jesus shows how to make the Kinect take a photo and upload it to Facebook with a simple spoken command. I wonder how it works in Swedish :/ Jesus then wrapped up the talk by demoing facial tracking and gesture interaction.
Yet another killer talk. If you’re into Kinect development, it is a must see.
Alexander Bard – The Rebels Come Out Online
I was not expecting much from Alexander Bard’s keynote. As a Swede, knowing Alex from the music scene with many so-so bands in his portfolio, I have no experience of him as a speaker.
I was happy to be surprised :)
Alexander’s talk was provocative, but highly clear-sighted and interesting. For instance, he doesn’t agree with the idea that our ideas form our technology. He thinks it is the complete way around, that the technology we surround ourselves with shapes our ideas. He sure backed that up with some really great examples.
I will not spoil anymore of this great keynote, since you simply must watch it (I told you that Thursday was amazing…almost all sessions were magnificent).