Q: You recently started a new job where you have the opportunity to use a host of the “Connected System” technologies within your architecture. When looking across the Microsoft application platform stack, how do you begin to align which capabilities belong in which bucket, and lay out a logical architecture that will make sense for you in the long term?
A: I’m Development Lead. Not a Lead Developer, Solution Architect or Develop Manager, but a mix of all those three, plus that I put on a variety of other “hats” during a normal day at work. I work close with both the Enterprise Architect and the development team. The dev team consists of “normal” developers, a project manager, a functional architect, an information architect, an tester, a designer and a “man-in-the-middle” whose only task is to “break down” the design into XAML.
We’re on a multi-year mission to turn the business around to meet new legislative challenges & new markets. The current IT system is largely centered around a mainframe-based system, that (at least as we like to think today, in 2009) has too many responsibilities. We seek to use “Components of the Shelf” where we can, but we’ve identified a good set of subsystems that needs to be built from scratch. The strategy defined by the top-level management states that we should seek to use primarily Microsoft technology to implement our new IT platform, but we’re definitely trying to be pragmatic about it. Right now, a lot of the ALT.NET projects gains a lot of usage and support, so even though Microsoft brushes up bits like Entity Framework and Workflow Foundation, we haven’t ruled out the possibility to use non-Microsoft components where we need to. A concrete example is in a new Silverlight –based application we’re developing right now; we evaluated some third party control suites, and in the end we landed on RadControls from Telerik.
Back to the question, I think over time, we will see a lot of the current offerings from Microsoft, either it targets developers, IT Pro’s or the rest of the company in general (Accounting, CRM etc. systems) implemented in our organization, if we find the ROI acceptable. Some of the technologies used by the current development projects include; Silverlight 3, WCF, SQL Server 2008 (DB, SSIS, SSAS) and BizTalk. As we move forward, we will definitely be looking into the next-generation Windows Application Server / IIS 7.5 / “Dublin”, as well as WCF/WF 4.0 (one of the tasks we’ve defined in the near future is a light-weight service bus), and codename “Velocity”.
So,the capabilities we’ve applied so far (and planned) in our enterprise architecture is a mix of both thoroughly tested and bleeding edge technology.
Q: WCF offers a wide range of transport bindings that developers can leverage. What are you criteria for choosing an appropriate binding, and which ones do you think are the most over-used and under-used?
A: Well, I normally follow a simple set of “Rules of thumbs”;
- Inter-process: NetNamedPipeBinding
- Homogenous intranet communication: NetTcpBinding
- Heterogeneous intranet communication: WSHttpBinding BasicHttpBinding
- Extranet/Internet communication: WSHttpBinding or BasicHttpBinding
Now, one of the nice thing with WCF is that is possible to expose the same service with multiple endpoints, enabling multi-binding support that is often needed to get all types of consumers to work. But, not all types of binding are orthogonal; the design is often leaky (and the service contract often need to reflect some design issues), like when you need to design a queued service that you’d eventually want to expose with an NetMsmqBinding-enabled endpoint.
Often it boils down to how much effort you’re willing to put in the initial design, and as we all (hopefully) know by now; architectures evolves and new requirements emerge daily.
My first advice to teams that tries to adapt WCF as a technology and service-orientation, is to follow KISS – Keep It Simple, Stupid. There’s often room to improve things later, but if you do it the other way around, you’ll end up with unfinished projects that will be closed down by management.
When it comes to what bindings that are most over- and under-used, it depends. I’ve seen someone that has exposed everything with BasicHttpBinding and no security, in places where they clearly should have at least turned on some kind of encryption and signing.
I’ve also seen highly optimized custom bindings based on WSHttpBinding, with every small little knob adjusted. These services tends to be very hard to consume from other platforms and technologies.
But, the root cause of many problems related to WCF services is not bindings; it is poorly designed services (e.g. service, message, data and fault contracts). Ideally, people should probably do contract-first (WSDL/XSD), but being pragmatic I tend to advice people to design their WCF contracts right (if in fact, they’re using WCF). One of the worst thing I see, is service operations that accepts more than one input parameter. People should follow the “At most one message in – at most one message out” pattern. From a versioning perspective, multiple input arguments are the #1 show stopper. If people use message & data contracts correctly and implements the IExtensibleDataObject, it is much easier in the future to actually version the services.
Q: It looks like you’ll be coming to Los Angeles for the Microsoft Professional Developers Conference this year. Which topics are you most keen to hear about and what information do you hope to return to Norway with?
A: It shouldn’t come as a surprise, but as a Connected Systems MVP, I’m most excited about the technologies from that department (Well, they’ve merged now with the Data Platform people, but I still refer to that part of MSFT as Connected Systems Div.). WCF/WF 4.0 will definitely get a large part of my attention, as well as Codename “Dublin” and Codename “Oslo”. I will also try to watch the ADFSv2 [Formerly known as Codename “Geneva”] sessions. Apart from that, I hope to use a lot of time talking to other people. MSFTies, MVPs and other people. To “fill up” the schedule, I will probably try to attend some of the (for me) more exoteric sessions about Axum, Rx framework, parallelization etc.
Workflow 3.0/3.5 was (in my book) a more or less a complete failure, and I’m excited that it seems like Microsoft has taken the hint from the market again. Hopefully WF 4.0, or WF 3.0 as it really should be called (Microsoft product seems to reach maturity first at version 3.0), will hopefully be a useful technology that we’ll be able to utilize in some of our projects. Some processes are state machines, some places do we need to call out in parallel to multiple services – and be able to compensate if something goes wrong, and other places do we need a rule engine.
Another thing we’d like to investigate more thorough, is the possibility to implement claims-based security in many of our services, so (for example) can federate with our large partners. This will enable “self service” of their own users that access our Line of Business applications via the Internet.
A more long term goal (of mine, so far) is definitely to use the different part of codename “Oslo” – the modeling capabilities, the repository and MGrammar – to create custom DSLs in our business. We try to be early adopters of a lot of the new Microsoft technologies, but we’re not about to try to push things into production without a “Go-Live” license.
Q [stupid question]: This past year you received your first Microsoft MVP designation for your work in Connected Systems. There are a surprising number of technologies that have MVPs, but they could always use a few more such as a Notepad MVP, Vista Start Menu MVP or Microsoft Word “About Box” MVP. Give me a few obscure/silly MVP possibilities that Microsoft could add to the fold.
A: Well, I’ve seen a lot of middle-aged++ people during my career that could easily fit into a “Solitaire MVP” category 🙂 Fun aside, I’m a bit curious why Microsoft have Zune & XBox MVP titles. Last time I checked, the P was for “Professional” I can hardly imagine anyone who gets paid for listening to their Zune or playing on their XBox. Now, I don’t mean to offend the Zune & XBox MVPs, because I know they’re brilliant at what they do, but Microsoft should probably have a different badge to award people that are great at leisure activities, that’s all.
Thanks Lars for a good chat.