If you heard my PDC05 Buzzcast but just can’t get enough of my voice, check out Architect MVP Mario Cardinal‘s Blah Blah Architecture podcast. I spent some time with Mario at TechEd talking about the Architecture Resource Center, why we designed it the way we did, why we have a site on MSDN and MS.com, how are taxonomy works, etc. Mario then takes my basic explanation and helps fill in more of the details, such as the relevance of the Architecture Resource Center to folks not using Microsoft technology. We also talked about Architecture Journal as well as why Microsoft invests in architecture.

BTW, I make some of the same points about architecture – as it differs from engineering, that it’s the space between business and technology, and the effect of title inflation – that I’ve made here on this blog this past week.

I guess I need some new material. 😄

So as a quick break from all this architecture talk, I’ve got a code smell question. Here’s a scenario, I’m interested in feedback on the best way to solve the issue.

I’m writing some VSTO code for Word using VS05. I want to be able to add and update custom properties on the document. You do this via the CustomDocumentProperties property off the document object. This DocumentProperties collection supports the standard collection type operations such as Add and an indexer. However, it’s a little exception happy. If you attempt to access a property that doesn’t exist, it throws an exception. And if you attempt to add a property that already exists, it throws an exception. So the first time you set a custom property you use the Add method and then after that you use the indexer to access the existing item in the collection and update it’s value.

Of course, the way my code is written, I want to hide this ugliness behind a method so that the rest of my code can simply set custom properties with ease. However, I want to use the same method regardless if the item already exists in the collection. So what’s the best way to implement the method? I can think of two primary ways.

1. Attempt to access the custom property via the indexer. If it throws an exception, trap it and call Add instead.
2. Manually iterate through the existing custom properties. If the property exists, update it directly. If it doesn’t, call Add instead.

Neither of these is particularly fragrant from a code smell perspective, but which is less odorous? The first one is more direct to write, but since this is all COM interop code, the COM exception is pretty generic. Theoretically, if something else caused an exception to be thrown, I’d still assume the custom property was just missing and swallow the exception, potentially causing an error somewhere else. However, writing the code to manually iterate through the collection just seems excessive.

In the end, I went with #2 as I was more worried about swallowing exceptions than manual iterating though the collection. What do you think? Was that the right choice?

So if architecture is the connection between business and technology, where does innovation fit? Microsoft has been pushing an idea of “integrated innovation” for several years now, but that’s primarily around technology innovation:

[T]he mission [of] Integrated Innovaton is about ensuring that the value of the Microsoft platform is greater than the sum of its components. It’s the coordination of software products, the way entire systems can be made to work together better. It’s a strategy to add customer-driven features and functionality to achieve specific business results while reducing cost and complexity
[‘Integrated Innovation’ Provides Partners with Roadmap to Success]

But what about business innovation? How often do we talk about that? Not enough IMO. Especially since the business innovation is much more likely to make or break a company than technology innovation. For example, how did Dell became the dominant company in the PC business? To quote from the Amazon review of Michael Dell’s book: “What makes Dell Computer unique is not what it sells, but rather how it sells it”. Usually, people note Dell’s direct-selling model as the catalyst for their success. Direct-selling might have been absent in the PC industry before Dell came along, but it’s not like direct-selling is a particularly innovative business model. However, what made the direct-selling of highly-configurable computers a reality was the innovative approach Dell took to managing it’s supply chain. Quoting from an Accenture case study on Dell’s supply chain:

Explains Dick Hunter, vice president, supply-chain management: “We now schedule every line in every factory around the world every two hours, and we only bring into the factory two hours’ worth of materials. We typically run a factory with about five or six hours’ worth of inventory on hand, including work in progress. This has decreased the cycle time at our factories and reduced warehouse space—space that has been replaced by more manufacturing lines.”

Not surprisingly, the project has produced more than just enhanced supply chain efficiencies and accelerated, highly reliable order fulfillment. At any given time, there is less than four days of inventory in the entire Dell operation, while many competitors routinely carry 30 days or more. In addition, automation has helped Dell react more quickly to correct potentially out-of-balance situations, made it much easier to prevent components from becoming obsolete and improved response times across the supply chain by providing a global view of supply and demand at any specific Dell location at any time.

Certainly, there is technology innovation involved in the management of Dell’s supply chain (and w/ RFID on the horizon, significantly more technology innovation in this space is still to come) but the primary innovation here was business oriented. I wonder where the next business innovations are going to be?

John asked  me over lunch if the people who read this blog would think I’m an architect or an engineer. Personally, using the definitions I’ve laid out this week, my heart’s in engineering but I’m getting more interested in architecture. Maybe I’m wrong, but it feels to me that pure engineering problems are giving away to more architectural problems as Moore’s law and it’s corollaries in network speed and storage space keep pushing out the limits of computing power. Jim Gray & Charles Levine wrote an funny article pointing out that Jim’s two year old TabletPC with a 1.6 GHz Centrino processor can handle over 8,000 transactions per second. To put that in perspective, in 1976, Bank of America’s DebitCredit system reached 100 transactions per second. It took a decade to build a system that handle over 200 transactions per second. Now, most of us are walking around with machine that could easily handle 40 times that performance.

As Moore’s law continues to solve technical challenges, I think it is creating new business challenges. And you know me…I like a challenge.

Michael cornered me on Tuesday to record a PDC05 Buzzcast with Mike Burner about the Architecture Symposium. At PDC03, the Architecture Symposium was one of the more popular and successful aspects of the overall confernece (though it was marred by a major room change issue that caused literally hundreds of attendees to be forced to watch from the hallway outside as the room was literally overflowing) and we’re looking to do something engaging again this year.

Like last time, the Architecture Symposium will be held the last day of the conference, Friday from 8:30 until noon (with breaks of course). After lunch, we’ll have a panel discussion featuring Gregor Hohpe, David Ing, Tony Redmond, Steve Swartz.

Here’s the full symposium abstract:

You’ve had a tantalizing week of cool technology, but now you need to transition back to your real job: making all of the pieces work together. The PDC Architecture Symposium will zoom you through the solutions lifecycle – from requirements to modeling to requirements to iterative development to requirements to operational feedback (which you might look at as another set of requirements) – showing you how traditional best practices and recent innovations can be used together to build robust solutions that accelerate business value creation.

Topics include:

The Architecture of Connected Systems
In the beginning there are the models – from the thing you scrawl on a napkin at lunch to that enormously complex diagram that your network architect carries around in a cardboard tube. What models are worth creating, and how do they relate to each other? Who are the key stakeholders for each, and how can you help them talk to each other? This session explores how to decompose value chains into your key models – your process and work flows, the information at the heart of your processes, and the access, deployment, and other operational models that you need to stay trustworthy and compliant. We will then map these models into a collection of services, orchestrations, and policies that define a highly integrated solution.

Building Connected Systems
With so much complexity and so many stakeholders, how do you build the right thing on time? This session explores the techniques to iterate agilely through a connected system project, including the patterns and practices for building solutions that combine messaging, workflow, structured information, and human interaction across platforms and across organizational boundaries. How can we give the right access to everyone in the value chain, respecting the very real boundaries around information and process control? How do we keep our models current, and use them to communicate with all of the stakeholders throughout the development lifecycle?

Managing the Connected Systems Lifecycle
As each iteration of your connected system is deployed and used, new requirements and system refinements emerge. How do we design in the operational hooks that give us the insight to learn from our deployed solutions? How do we re-factor and version our services and orchestrations to improve service reuse, scalability and operational efficiency? The key theme is driving collaboration between development and operations groups from the earliest design phase through the ongoing maintenance of the system.

As Michael said on the buzzcast, you just can’t miss the Architecture Symposium. See you there.

So I said yesterday that I’m talking about architecture, not architects. However, today I am going to discuss the word architect and the dramatic misuse of the word I see pretty regularly. Or at least, I see now because Paul Preiss of IASA mentioned it when we were hanging out at TechEd.

“Architect” is a noun, not a verb.

Usually, when I hear the term “architect” used as a verb, it’s being used as a synonym for design. In fact, the term “architecture” is also often used as a synonym for design. For example, Arnon wrote this:

Architecture is design (but not all design is architecture)
[Arnon Rotem-Gal-Oz, What’s “Software Architecture”?]

Like Fowler’s description I talked about on Monday, I don’t like this one much either. Architecture isn’t just “good design” which is what Arnon’s description makes it sound like. This begins to get into the title inflation that Alan Cooper wrote about a few years ago. Speaking of Alan, here’s his definition of architect.

The panoply of software construction includes three vital roles: the programmer, the engineer, and the architect. The architect is responsible for determining who the user is, what he or she is trying to accomplish, and what behavior the software must exhibit to satisfy these human goals. The engineer’s responsibilities are comparable but focused on technology. A good engineer can and should ignore human issues, confident that the architect will cover the human side.

That definition of architecture (i.e. what Alan’s idea of an architect does) dovetails pretty nicely with mine. The architect and architecture is the link between the users (i.e. the business) and the software (i.e. IT). The only thing I would change is that as we get deeper into service orientation, interop and connected systems we have lots of process that don’t have direct user involvement. So the “human goals” Alan mentions may not be end user goals so much as business / organization goals. Either way, they certainly aren’t IT goals.

As Dave Welsh said and my dad pointed out in my comments, Business is from Mars, IT is from Venus. But that doesn’t mean they can’t get together. In fact, they have to get together. You show me a system with no business drivers or impact and I’ll show you a failed architecture.