Seiter's Blog

It was about 50 degrees outside and breezy in Half Moon Bay, California when I overheard three people discussing robots over lunch.

"So we'll buy the robot for about 50k, then we'll need about 20k more for setup and configuration. That covers pretty much everything except software, which will cost infinity."

They all laughed and one of them, who looked like the guy who paid the bills, rolled his eyes and nodded as if this joke weren't all that far from the truth.

And of course, it isn't. I've written about software for many years now with varying degrees of accuracy. As I've gotten older and experienced more projects, it has become entirely evident that software has a lot of big problems. There are many paradigms out there which promise all kinds of benefits, some of which can be realized. Unfortunately at the end of the day, complex systems are complex. And while that doesn't sound like much, it's altogether meaningful when architects around the world fall prey to silver bullet solutions, shortcuts and other methodologies that often cause more harm than good or lead customers to unreasonable expectations about results.

So what are we to do about this? How can we bring "infinity" to something more reasonable? Here are my suggestions, which are the product of experience and of nothing else.

1. Complex systems are complex: The first mistake any architect can make is to assume he/she fully understands any complex system. I'm not talking about Hello World, or your senior project. I'm talking about multi-component systems. I'm talking about moving parts and integrated systems. Many architects will go up to a white board and draw a bunch of boxes which correspond to subsystems. Since they can see all of the boxes, they assume they understand the system as a whole. I walk up to the board and say, "Draw the lines." Please draw the lines between systems and while you're at it, explain how every line works. And while you're at that, tell me what happens to the system when a line doesn't work as expected. Tell me what happens when 75% through the project, someone draws a new line. Great architects not only recognize that they don't fully understand the complex system, they embrace it. And in this way, they learn to prepare for the unknown.

2. Hire great engineers: Time after time, this is the moral of the story. Again and again, this single factor can determine the success or failure of major components of a system. A good engineer is not one with an incredibly lengthy title. A good engineer is not the hot-off-the-press graduate and it's not the guy with 15 years of experience. A great engineer only concerns himself with the task of solving problems and if he/she has any of the above as well, fantastic. When you interview your potential engineers, ask them to describe something they are proud about having done. "Completed XYZ project" and "Met XYZ deadline" are fine answers but they aren't great answers. "I figured out how to integrate a 15 year old cold-fusion app with a modern web-service using SOAP and a few hundred lines of code I wrote." Bingo.

3. Don't blame your customer: Isn't it easy to blame the customer? A couple useless words fly out of your mouth and you feel better about the fact that you're changing something about your system yet again. Unfortunately, this does nothing to help solve the problem. Customer problems are complex and it would take me ten pages to talk about them in their entirety. I'm also completely unqualified to do so. What I can say is that I've seen this mentality in many organizations and in many engineers. I've seen it in project managers and I've seen it in myself. The truth is, this mentality does nothing to help you accomplish your goal. There are entire books devoted to managing customer expectations and project managers tend to be better at this than engineers do. The best thing you can do as an engineer is to take changes in stride and try to engineer your software to be flexible. Requirements are driven by customers and customers are human. Accept this fact and you'll find yourself in a much happier, healthier engineering state. And your customer will notice.

Related: To Engineer or Not to Engineer

Some people were surprised to know that I love Clojure apparently. Take a look at my quick sort and compare it to my python and java variant and it will become obvious why this language rocks socks.

(defn quicksort [list]
  (if (< (count list) 2) list ;if the list has 0 or 1 items, return
    (let [
          pivot (first list) ; our pivot will be the first item
          rl (rest list) ; compare against everything else
          lt #(< % pivot) ; our less than predicate
          gte #(>= % pivot)] ; our greater than equal predicate
    (lazy-cat
      (quicksort (filter lt rl)) ; recurse on filtered less
      [pivot]
      (quicksort (filter gte rl)) ; recurse on filtered greater than equal
    ))))

Make sure we give this bad boy a quick test:

(def mylist (shuffle (range 1 25))) ; make some shufled list of items
(println mylist) ; take a look before sorting
(time (quicksort mylist)) ; time our three runs
(time (quicksort mylist)) ; the first should take the longest
(time (quicksort mylist)) ; magic!

And the results, of course:

[1 2 13 20 19 21 3 22 17 14 16 7 18 10 11 5 12 4 6 8 15 24 9 23]
"Elapsed time: 1.975864 msecs"
"Elapsed time: 0.109321 msecs"
"Elapsed time: 0.107011 msecs"
(1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24)

When we observe the world, we do so via a model generated by our minds. Our senses absorb data in various forms and pass that information on to our brain, which then processes this information and organizes it in some meaningful way. The concept of a mental model is not a new one. The idea that there is a disconnect between what we sense and what really exists has been around for ages. In Steven Hawking's new book "The Grand Design", he talks about these models and what they mean for Scientific Theory, which brings me to the subject of this post, which is somewhat different from the subjects Steven describes in said book.

The Scientific Method is based on the notion that one can perform observations to prove or disprove a concept. This notion is rooted in the belief that there is a level of purity involved in the process of measurement that is not involved in other forms of inference. The implication is, of course, that observation is the highest form of posit validation. The idea of a model doesn't apply strictly to human observation either, rather, even a device that performs a measurement does so with some set of assumptions. In that case, these assumptions form the foundation of the model from which measurements are made.



But observation only gets us so far. In fact, when you apply this paradigm to electrons, for example, you end up getting some very strange results. The Heisenberg Uncertainty Principle, for example, states that as one increases ones certainty with respect to the position of a particle, one decreases ones certainty with respect to the momentum of the particle. This principle is manifest entirely in the process of measurement. I'm willing to argue that this principle arises out of the application of a classical process (The Scientific Method) to a non-classical (or quantum, if you'd like) concept. (Particles)

Another outcome of this line of thought is demonstrated by a thought experiment called Schrödinger's cat. The result of this thought experiment is that one must accept that at a given time prior to observation of the state of the cat, it is both alive and dead. Statistics can be built around this that properly predict the results of experiments of this sort. These statistics are resoundingly proven to be accurate. But is this really an accurate picture of the universe?



In order to determine the answer to that question, we are left with the tools provided by The Scientific Method. But these tools all require us to perform a measurement. I'm willing to argue that this is the real problem with the current state of physics. Physicists around the world today are struggling to find a "Theory of Everything", a unified theory of physics. I argue that they are looking in the wrong places. The problems that arise when trying to unify Quantum Theory with Relative Theory are caused by the fact that The Scientific Method is a process formed within the classical model of the universe.

So, is there another process we can use to verify posits about the nature of reality which is framed in a non-classical model? Can we determine the laws of nature without using the process of observation and measurement? That is, in my opinion, what the real question on everyone's mind should be.

It's thin, sleek and powerful. But is it really Incredible?

Let me start by saying that I am a lover of all things electronic. I've been scoffed at for the ownership of a ZUNE and I've been scoffed at for the ownership of an iTouch. But this morning at 7AM, I picked up one of the first HTC Droid INCREDIBLE handsets from the Verizon Wireless store in town and I doubt I'll get scoffed at for carrying this piece of fine art around. But 12 hours in, the honeymoon stage is already over and it is obvious that while Google and HTC have done an excellent job, there is still some work to be done.



If you don't know already, the HTC Incredible is the latest in a slew of super-powered Android phones to hit the market. It sports a 1GHZ Snapdragon processor, 8GB of internal memory, a ridiculously vibrant touch-screen and a whopping 8 megapixel camera. It comes with Android 2.1 and HTC's Sense UI on top to smooth out the rough edges. All of this is railed on Verizon's CDMA network which is arguably the best in the nation.

All of that sounds good on paper but how well does it all come together? The answer is a complex one. The INCREDIBLE has a terrible habit of performing feats of pure phone genius one moment and then doing things in the next that simply don't make sense. This process isn't more evident than in the voice recognition features of Android 2.1.

Press a button, say something like "Papa Johns pizza, Syracuse New York" and viola, the INCREDIBLE gives you exactly what you're looking for. The voice recognition is so good in fact, that I kept increasing the complexity of my verbiage just to see how far I could take it. The software finally failed when I threw, "terrible transvestites titillate triple tempura" at it. In all honesty though, it is without a doubt the best speech recognition software I've ever had the pleasure of using.

Unfortunately, after finding Papa Johns spot on, you have to click no less than five buttons to get directions to said location. It's cumbersome, unnecessary and simply annoying. This is a theme that repeats itself throughout the software. The hardware isn't without it's problems either. Sure the back of the phone looks cool; unfortunately the camera lens is the highest level item on the surface and doesn't come with a lens cap, so any time you set your phone on a surface face-up, the lens rubs against it. Only time will tell if the material is affected by this or if it can stand up to it just fine but it's a simple design flaw nonetheless.



Hardware: The HTC Incredible is simply put, a hardware work of art. It's fast, sleek and put together very well. I'd love to see a lens cap though.

Software: Android + Google Cloud + App Marketplace = Awesome. Better control of the software stack would allow a more seamless user experience though and certain features simply feel clunky. This is contrasted by apps that look beautiful, run great and put other apps of their type on other devices to shame. It is difficult as a user to reconcile these two facts.

Overall: The HTC Incredible is a fantastic phone with a fantastic carrier. While it is hard to find many flaws with it, there are some, just as there are with any device. Users looking for a powerful smart-phone on Big Red have an easy decision to make. Those of you who are less concerned about carrier might want to hold your cash a few months and see what the EVO turns out like as well as other offerings coming out this year from Microsoft, Dell and others. For more detailed reviews, head over to any of the links below.


http://www.engadget.com/2010/04/19/droid-incredible-review/
http://gizmodo.com/5520071/htc-droid-incredible-review-more-like-impressive
http://www.infosyncworld.com/reviews/cell-phones/htc-droid-incredible/10909.html