0 comments suggest edit

I know, I know, you’d like to see the FizzBuzz discussion die a quick death, but trust me, this is an interesting point, or at least mildly amusing.

Sorry to revive the dead horse, but a comment in my blog brought up a very good point. In fact, I’m kicking myself for not noticing this myself, having been a math major and I love pointing out this type of minutiae.

In the original Fizz Buzz test, the functional spec asks the programmer to print the numbers from 1 to 100.

But as a commenter points out

Why can’t spec writers write? Unless you mean integers, there are an infinite number of real numbers ’from 1 to 100’

Exactly! There are an infinite range of numbers between 1 and 100. The specification is technically not clear enough. Writing a program to spec exactly would… well be impossible.

This is exactly why I said the following in a another comment

I still need to gather requirements! What platform must this FizzBuzz program support? Any performance requirements? Does the output need to be available over the web?…

Unfortunately, I missed the most important question I should have asked.

I assume you mean all intergers from 1 to 100 inclusive, is that correct?

I know what you’re thinking. In cases like this, developers should be able to intuit what the client means. If a developer asks Do you mean Integers or Real Numbers?, that developer is being a smart ass.

But my point is still valid. If a client says, I want a CRM system, you may know exactly what a CRM system is, but it may be totally different from what they think a CRM system is.

This really highlights the difficulty of writing good requirements and a good spec. You don’t know the background of the person you’re handing off the document to.

What makes perfect sense in your mind might mean something different to the reader.

Perhaps it’s situations like this that lead 37Signals to advocate getting rid of functional specs altogether.

Whether you go that extreme or not is not so important as keeping the lines of communication open with your client. Never accept a requirement and functional spec at face value. Specs are always a poor approximation of what the client really wants. All specs are broken to one degree or another (though that doesn’t mean they are all useless). Ask for clarification. Keep the dialog going.

This is also one reason why Big Design Up Front (BDUF) can really nail you in the butt. These subtle things are missed all the time. Having an iterative process where you’re not on the hook for requirements gathered months ago gathering dust helps mitigate the risk of incomplete and inaccurate requirements.

Even by thousands of software developers reading blogs.

0 comments suggest edit

Update: I have an even better startlet for stopping and starting services in my comments.

If you’re running Vista, run, don’t walk, and go download and install Start++ (thanks to Omar Shahine for turning me on to this). Make it the first thing you do. Many thanks to Brandon Paddock who developed this nice little tool. He describes the tool in this post.

I have a message for Start++ from the Start menu. “You complete me!”.

Ok, terribly corny jokes aside, it’s the little things that save me lots of time in the long run. For example, starting and stopping SQL server is kind of annoying for me on Vista. Here’s my the typical workflow.

  1. Hit Windows Key, type in cmd
  2. type net stop mssql
  3. Doh! System error 1060 occurred. Right, I need to be an administrator.
  4. Grab the trackball
  5. Click on the Start menu
  6. Right click Command Prompt
  7. Click Run as administrator.
  8. Now type net stop mssql.

Is your hand hurting by now? Because mine is.

Of course, I’m an idiot. Or, I was an idiot. Now, I’ve mapped the Start++ keywords startsql and stopsql to automatically run the commands I need with elevated privileges.

Click for larger image.


Notice you can check the Run elevated checkbox for any command. Yes, I get the UAC prompt (Yes, I still have that sucker on), but that’s not such a big deal to me. Now my workflow is reduced to:

  1. Hit Windows Key, type in stopsql
  2. Hit the Left Arrow Key and Enter when the UAC prompt comes up.


For your convenience, I’ve exported the startsql and stopsql “startlets” and put them on my company’s tools site here. I figure this one alone saves me a few seconds every half hour.

If you are using a named instance of SQL Server, you will need to change the argument in the Arguments column like so:

/C "net start mssql$NameOfInstance"

I have a few hundred or so startlets I can think of adding. Happy shortcutting!

Technorati Tags: Tips, Vista, Start++, Productivity

0 comments suggest edit

Thought I should take advantage of my latest bout of insomnia and do a slight redesign of my website. My goal was to clean it up a bit so it looks less crowded and cluttered.

I also removed the Flickr images because the script was slow to load and caused my site to load slowly. I didn’t think it added a whole lot anyways. I may look into creating a server-side flickr control later.

Here’s a screenshot, in case you’re reading this in an aggregator.

Thumbnail of my

Let me know what you think of the design. My next step will be to focus more on usability. If there’s anything that annoys you about my site (here’s where Jeff will chime in) do let me know.

0 comments suggest edit

Stack of
PaperIn Jeff Atwood’s infamous FizzBuzz post, he quotes Dan Kegel who mentions.

Less trivially, I’ve interviewed many candidates who can’t use recursion to solve a real problem.

A programmer who doesn’t know how to use recursion isn’t necessarily such a bad thing, assuming the programmer is handy with the Stack data structure. Any recursive algorithm can be replaced with a non-recursive algorithm by using a Stack.

As an aside, I would expect any developer who knew how to use a stack in this way would probably have no problem with recursion.

After all, what is a recursive method really doing under the hood but implicitely making use of the call stack?

I’ll demonstrate a method that removes recursion by explicitely using an instance of the Stack class, and I’ll do so using a common task that any ASP.NET developer might find familiar. I should point out that I’m not recommending that you should or shouldn’t do this with methods that use recursion. I’m merely pointing out that you can do this.

In ASP.NET, a Web Page is itself a control (i.e. the Page class inherits from Control), that contains other controls. And those controls can possibly contain yet other controls, thus creating a tree structure of controls.

So how do you find a control with a specific ID that could be nested at any level of the control hierarchy?

Well the recursive version is pretty straightforward and similar to other methods I’ve written before.

public Control FindControlRecursively(Control root, string id)
  Control current = root;

  if (current.ID == id)
    return current;

  foreach (Control control in current.Controls)
    Control found = FindControlRecursively(control, id);
    if (found != null)
      return found;
  return null;

The recursion occures when we call FindControlRecursively within this method. Essentially what is happening (and this is a simplification) when we call that method is that our current execution point is pushed onto the call stack and the runtime starts executing the code for the internal method call. When that method finally returns, we pop our place from the stack and continue executing.

Rather than try to explain, let me just show you the non-recursive version of this method using a Stack.

public Control FindControlSansRecursion(Control root
    , string id)
  //seed it.
  Stack<Control> stack = new Stack<Control>();
  while(stack.Count > 0)
    Control current = stack.Pop();
    if (current.ID == id)
      return current;
    foreach (Control control in current.Controls)
  //didn’t find it.
  return null;

One thing to keep in mind is that both of these implementations assume that we won’t run into a circular reference problem in which a child control contains an ancestor node.

For the System.Web.UI.Control class we safe in making this assumption. If you try and create a circular reference, a StackOverflowException is thrown. The following code demonstrates this point.

Control control = new Control();
control.Controls.Add(new Control());
// This line will throw a StackOverflowException.

If the hierarchical structure you are using does allow circular references, you’ll have to keep track of which nodes you’ve already seen so that you don’t get caught in any infinitel loops.

0 comments suggest edit

We’ve been having an internal debate within the Subtext mailing list over the merits of SourceForge vs Google Code Project Hosting vs Codeplex. Much of the discussion hinges around the benefits of Subversion for Open Source projects when compared to Team Foundation System (TFS).

Before I begin, I do not mean for this to devolve into a religious argument. This is merely my critique from the perspective of running an Open Source project. I personally think both are fine products and both probably work equally well in the corporate environment.

TFS Advantages

  • Easy of use. For developers with a background in using Visual Source Safe or Sourcegear Vault, the interface into TFS will be familiar. Subversion requires more of a learning curve for these developers, though this is mitigated by my suspicion that a large percentage of Open Source developers tend to use CVS and SVN already. 
  • Work Item integration is sweet. I’ve been contributing some code to the Subsonic project and I actually love the work item integration in VS.NET. It’s pretty nice to be able to review and close work items while working on the code.
  • Shelving is great. Certainly nothing stops you from doing something like this in Subversion by using conventions, but I like the syntactic workflow sugar this provides.

Subversion Advantages

  • Anonymous access. Users who want to look at the code, view the change history of the code, and update their local code to the latest version can do so form the convenience of their favorite Subversion client. This is much more cumbersome with TFS.
  • Patch Submission. This goes hand in hand with anonymous access. Users without commit access can have Subversion generate patch files consisting of their changes and submit these files. This makes it really easy for the casual contributor to quickly submit a patch as well as makes it easy for the Open Source development team to apply contributions to the source. This is a huge benefit to the project. Unfortunately with CodePlex, you either give commit access or you don’t. If you don’t, it’s a pain for users to submit patches and a pain for the project team to apply patches. Just ask Rob Conery what happens if you give commit access too freely.
  • Offline Support. Regardless of what Jeff says, offline mode does matter for many applications. For example, sometimes I have to connect to an obnoxious VPN that destroys my general internet connectivity. It’s nice to be able to connect, get latest, disconnect, work, connect, commit changes, disconnect. Try that with TFS.

Again, as source control systems, I believe they are both great systems. But for the needs of an open source project, I feel that Subversion has advantages. As far as I understand, TFS was designed as an enterprise source control system. However, the needs of the enterprise are often different from the needs of an Open Source team.

Subversion, itself open source, was used during its own development (when it became stable enough). So it is well suited to open source development.

If Codeplex supported Subversion, I would probably want to move Subtext over in a heartbeat. If you feel the same way I do, please vote for the work item entitled Subversion Support (SVN).

It looks like a lot of people would like to see this as well as it is the top vote getter on the Codeplex work item site.

And before you rail on me asking, Why Microsoft would ever consider such a move? Isn’t Codeplex a showcase for TFS and Microsoft Technology Open Source projects?

A member of the Codeplex team informed me that Codeplex is the home for any Open Source project - on any and all platforms. In fact, they do now host a few non-Microsoft projects. Of course their dependency on TFS does naturally limit the types of projects that would host there.

0 comments suggest edit

Micah Dylan (CEO of VelocIT) writes about a tech support call he had with Comcast, the local cable company.

To sum (though you should go read it) Micah can’t login to the Comcast website and the tech support guy tells him to call Microsoft claiming it’s a problem with the browser!

Comcast Really?! You expect your customer to call Microsoft and singlehandedly convince them to update their browser so it works with your website, rather than following the lead of millions of other websites and making the website work with the browser? Really?

0 comments suggest edit

UPDATE: This post was meant to be mostly tongue in cheek. I love programmers. Some of my best friends are programmers. Also, to be fair to the FizzBuzzers, the spec for this question is flawed, as are all specs as I point out in Why Can’t Spec Writers Write Specs.

You gotta love CodingHorror.com. Once again, Jeff Atwood writes a great post entitled Why Can’t Programmers.. Program?But in this case, **the best part is in the comments. Let me explain.

Jeff writes about a simple “Fizz Buzz” test given by Reginald Braithwaite. Here’s the functional spec.

Write a program that prints the numbers from 1 to 100. But for multiples of three print “Fizz” instead of the number and for the multiples of five print “Buzz”. For numbers which are multiples of both three and five print “FizzBuzz”.

Mind you, this is not a trick question. It is as easy as it sounds. Or it should be. According to Reginald Braithwaite, the majority of comp sci graduates can’t do it.

Naturally, the very first comment to Jeff’s post solves this simple problem. But read further and be amazed at how many developers get it wrong in such a public forum!

The problem may not be that these programmers can’t program. It’s that they can’t read!

I think reading comprehension must be at an all-time low among graduating programmers. Perhaps it’s time to bring back more word problems in our curriculum. Of course we’ve all seen this lack of reading comprehension in comments to our blog posts, haven’t we? As in when someone completely argues against a misinterpretation of your post.

It never fails to amaze me at how some companies I’ve worked with (or at) praise developers who are fast, even if they do the wrong thing because they failed to understand the spec or include so many bugs as to have made the spec irrelevant.

At least he’s fast!


So what’s worse? A programmer who can’t write code, or a programmer who can, but can’t follow a spec?

code, tech 0 comments suggest edit

Forget flame wars. They’re so USENET. The new way of geek fighting is much more subtle.

Subtlety of a blow

Recently Scott Hanselman wrote a post on the SEO implications of URL standardization which Jeff Atwood followed up with his post on URL Rewriting to prevent Duplicate URLs.

Jeff makes it known that he prefers that his blog be referenced with the “www.” in the front. I think that’s an unnecessary three characters that is quite dated. Yes, we know your blog is on the World Wide Web. The World Wide Web is no longer a nice new shiny bauble. You can feel free to drop the www.

Besides, your domain name will sort better without it.

And for those who say: But if you hit CTRL+ENTER in the address bar, it’ll surround what you type with “www.” and “.com” I saw bah! If I want to go to Jeff’s blog, I can type c-o-d and hit tab and I’m there. A modern browser is pretty handy for that with that newfangled auto complete.

Anyways, in this recent post on the reading habits of programmers, I linked to a post of Jeff’s without the “www.”, which did not go unnoticed by Jeff nor LazyCoder Scott (thanks for pointing out the jab Scott, it might’ve gone under his radar damnit! A nice subversive jab at the man.)

Responding with a geek equivalent of “Oh no you di’int”, Jeff’s very next post, he returns the favor. Touché my friend! Touché. Well I’ve hidden a little something for you in this post. Oh yeah. Bring it on.

Aren’t we just so pathetic?

P.S. Jeff now flinches if you threaten to Digg his blog.

0 comments suggest edit

I finally got around to uploading pics of our short Seattle vacation to Flickr. Our first full day there, Saturday, benefitted from beautiful weather. We took a nice boat ride on Lake Washington and skipped by BillG᾿s immense house.

Bill Gate᾿s

The pic does not do it justice. There are a few more here. We must᾿ve gotten too close for comfort because a security guard came to the water᾿s edge and mad dogged us.

We also had the opportunity to check out the new sculpture park. I thought it was quite interesting. One of my favorite sculptures was this chrome tree, seen here behind a real tree.


I also had the opportunity to visit some friends at Microsoft. Apparently Microsoft is experiencing explosive growth. My friend Julie who worked on campus a couple years ago is now at an office complex one freeway exit away off campus.

Our friends worked hard to convince us that we should move to Seattle. But I think it may take more convincing, as we᾿ve both contracted a cold and blame the damn Seattle weather. Wink

asp.net 0 comments suggest edit

Scott Watermasysk has a great set of Quick Tips for ASP.NET on his blog. And this is only part one! I’m Looking forward to seeing the subsequent posts of this series.

However, I disagree slightly with his tip to Never use the ASP.NET Label control. I would rephrase it to:

Never use the ASP.NET Label control when a Literal would do.

The problem is not the Label control. The problem is treating the Label control as if it was merely a Literal control.

The Label control has a specific usage, to be a label for a form input. For example, check out this screenshot of a text label and textbox.

Textbox control and

Let’s look at one way to create this simple form using a Literal control (Notice I’m getting the value from a Resource file because we all want to be internationally friendly, right? But it also points out why I use an actual control rather than just typing the label directly).

<asp:Literal id="label" Text="<%$ Resources:UserName %>"   runat="server" /> <asp:TextBox id="textbox" runat="server"/>

When you click on the word Username what happens? Nothing. Wouldn’t it be nice if the Textbox next to the literal was given focus? Here’s how:

<asp:Label id="label" AssociatedControlId="textbox"     Text="Username" runat="server" /> <asp:TextBox id="textbox" runat="server" />

Now, we’ve associated the label to the control. The underlying HTML looks something like (simplified for the sake of discussion):

<label for="textbox">Username</label> <input type="textbox" name="textbox" value="" />

This associates the label Username with the textbox control, so when you click on label, the textbox gets focus.

So why use the Label server control here? Why not just use the <label /> html tag directly?

Because you might not know at compile time what the client id of the TextBox will actually end up as. What if these two controls were inside a UserControl inside a page? The ID might get a bit funky as its nested in a control within a control.

Technorati Tags: Tips, ASP.NET,

0 comments suggest edit

Logo In Decemember I wrote about a great soccer camp I attended. Well one of the other attendees is a Sports Anchor at Univision and had a camera man in tow taking footage during the camp.

Soccer BallStarting on Saturday, February 17^th^, Univision will be airing a four part series about the Galaxy Camp. The second part will air on Sunday and the third and fourth parts the following weekend.

Of course, this is subject to change without prior notice, but if you tune in, you just might catch a glimpse of me on the tube. Unfortunately, I didn’t feel comfortable enough with my spanish to give an interview like several of the other campers.

In Los Angeles, this would would be on KMEX channel 34.

Disclaimer: I was once a Univision employee.

0 comments suggest edit

I’m posting this more as a reminder for myself. Where exactly are you supposed to place XSD files to give you intellisense for XML files in Visual Studio.NET?

Visual Studio.NET 2003

C:\Program Files\Microsoft Visual Studio .NET 2003\Common7\Packages\schemas\xml

Visual Studio.NET 2005

C:\Program Files\Microsoft Visual Studio 8\Common7\Packages\schemas\xml

This is useful for such things as Nant.xsd and UrlRewriting.xsd.

To get intellisense for NAnt, you have a couple of other steps to do.

0 comments suggest edit

SeattleThe wife and I are headed up to Seattle for President’s day weekend to visit some friends.

Yeah, I know. Why Seattle At This Time Of Year!?

Well with her being all preggers and all we didn’t want to go somewhere too distant (like Tahiti). Not only that, a little cold weather might be a nice change of pace. I mean you can only take so many consecutive days of blue skies and sunshine.

I’ll also be swinging by the Microsoft Campuson Tuesday to have lunch with my compadres Adam Kinney and Erik Porter. My hope is to grovel and beg for them to sneak me through the back door of Mix

  1. It probably won’t work, but I gotta give it a try, no?

I plan to show up a little earlier to see another old friend from college who works at Microsoft and, believe it or not, is not a developer! She’s in Finance doing something or other.

Hopefully I’ll run into some other Microsoftie characters I know!

0 comments suggest edit

Steve Harman just announced the release of Subtext version 1.9.4 Windward Edition. This one comes with a lot of bug fixes, so be sure to upgrade.

Just so you know things work, I add bugs to Subtext, and Steve and the other developers fix the bugs. It’s a rather efficient ecosystem and is working quite well for us. It keeps everyone on their toes.

Steve’s post has the full list of bug fixes and such. The most interesting addition is that we’ve implemented a Google Sitemap, which was submitted as a patch, if I remember correctly.

This release is notable because of the increased number of patch submissions. I greatly appreciate the contributions by all the new contributors along with the stalwarts. Working on Subtext is a joy because of these people.

As for Subtext 2.0, Progress has been slow, but steady. One challenge we’re dealing with is how to cleanly handle the following two multi blog scenarios as summarized by Simo in our mailing list.

​1. The multi blog site is a communiy site: so user registered to a blog, are already registered for all blogs.

​2. The multi blog site is just a host for many different non related blogs: here all users are different, and even if 2 blogs are on the same system, the user shouldn’t know about that.

We pretty much can already handle case 1 which is nice because it means the same user can be the owner of more than one blog, rather than requiring a user account per blog. This is useful for me personally as I host three blogs on a single installation, and right now, the admin is technically three different accounts.

Strictly speaking, this implementation makes implementing the isolation requested in case 2 difficult, because usernames must be unique. So if there’s ever a name conflict between two users attempting to register with the same username, the two blogs will affect each other and not be completely isolated.

However, if we make the user’s email address their login, we can implement case 2 in spirit. First of all, there won’t be two different users using the same email address, so the naming conflict issue is resolved.

Secondly, if you wish to register for blog 2, why should we make you fill in your information again if we already know who you are? We should simply present a message that says, Hey, we know who you are, if you want to register for this site, click here and we’ll auto register you.

That way, the user is in control over who which blogs get to see their registration information, yet they only have one user account in the system. Certainly there are some improvements we can make later, for example, some blogs may want more information than others to register.

But for now, we only ask for the minimal amount of information and will keep things simple and consistent across the board.

In any case, enough blabbing. If you want to download the latest release, you can find it on our SourceForge site or just use this [DOWNLOAD] link.

0 comments suggest edit

UPDATE: How could I have overlooked the Plugin Keyvan Nayyeri wrote two months ago. Sorry buddy!

You know what would go well with my XFN Highlighter Script like Kimchi goes with Bulkogi? How about a plugin for inserting proper XFN hyperlinks?

Screenshot Scott Dornan (and Keyvan) delivers exactly that with his Insert XFN Hyperlink plugin for Windows Live Writer.

Apparently he was introduced to the concept of XFN after reading a post on my blog.

Yes, you have me to thank for this fine plugin (not to mention Scott who actually did all the work, while dealing with a-holes like me trying to take the credit).

It’s not yet available in the Windows Live Gallery, but will hopefully be there soon.

In the meanwhile, we’ll have to beg for a direct download. Wink

0 comments suggest edit

My feed looks like several items were reposted. This is due to a bug in Subtext when editing an existing post using Windows Live Writer. In any case, we’ll have a fix available in the next release of Subtext soon.

0 comments suggest edit

I received a call from a nice lady asking to speak to someone who could confirm an employment at VelocIT.

Me: Sure, I can do that.

Lady: What’s your name and title?

Me: I’m Phil Haack, the CTO.

Lady: Is there someone else I can speak with?

Me: Uhhh, why? Is there a problem? I am an owner and manager of the company. Certainly I have authority to confirm an employment.

Lady: I can’t have you confirm your own employment.

Me: Oh! This is for me!?

Man, I’d be in big trouble if I were a one-man shop.

0 comments suggest edit

IBM Second Life Offices - From
http://www.sutor.com/newsite/blog-open/?p=1382Micah Dylan, co-founder and CEO of VelocIT (you know, the place I work for my day job), has recently discovered Second Crack Life. He’s been hanging out with some big wig at IBM in charge of open source and open standards. And you know how I loves me that open source.

Apparently IBM is looking at ways to build internal community among a distributed workforce and exploring whether or not Second Life fits into that. Obviously, this holds appeal to VelocIT because we’re a distributed company.

It’s hard not to find Second Life intriguing if you’ve ever read Snow Crash, but I’m tentatively skeptical of its use as a day to day communication tool. I’m just not ready to have a processor intensive 3D environment running in a window while I’m trying to get my coding on.

However, I do see its appeal for having short fun meetings. Jon Galloway and I have hung out in Second Life for a short period before we realized what a time sink it could end up being.

I think Second Life could be a great tool for Open Source teams to hold meetings. In fact (and you heard it here first), when we release Subtext 2.0, we’re going to have our release party in Second Life.

0 comments suggest edit

ConeryOn Tuesday night I met Rob Conery, aka Mr. Subsonic, aka Mr. Commerce Starter Kit. The man is shifty and goes by many aliases. He was in town visiting family and was able to pull away for a night to kick it.

From time to time, I’ve met people in person at various conferences who I only knew via their blogs. Usually, because you have no idea what to expect, those meetings can be full of surprises.

You’re nothing like how I pictured you in my head. I thought you’d be taller, have hair, and be more female.

Since I’ve chatted with Rob on Skype before, there were no surprises.  Except that he’s ginormously tall. I literally almost blurted out, You’re a freakin’ giant! But was able to hold my tongue.

Jon Galloway joined us after a harrowing drive up the rain soaked freeway from San Diego and we headed for Hurry Curry of Tokyo on Sawtelle for some delicious Japanese style Curry. After a couple of bottles of sake and Sapporo beer, we headed over to The Mór Bar for $2 Tuesdays with Moontribe.

The Mór Bar is a sweet spot to hang out, get a drink, talk and listen to good music. They have a nice arrangment of plush couch/benches in a U-shape arrangement at the front of the bar on a deck, which evokes a Lawrence of Arabia feel.

We settled in to geek out and Rob’s a fun guy to have a conversation with. Just brimming with great ideas and interesting stories. Shortly into it, Rob remarks on my total ADD (Attention Deficit Disorder). What the heck do you expect when you buy me a Maker’s Mark for the first round of drinks? Yeah, I get easily distracted after multiple shots of sake, beer, chased by a strong whiskey. What can I say?

So there you have it, Rob’s Tall, Jon’s Bald, and I have ADD. Any case, Rob, if you read this, I hope you got your Credit Card back. Good times. Good times.

0 comments suggest edit

Jeff Atwood writes a great post on how to become a better programmer by not programming. For the most part, I totally agree with his premise that to become a better programmer, you need to…

Learn about your users. Learn about the industry. Learn about your business.

Understanding the big picture is absolutely essential to being a good developer. Especially when you’re in the 99^th^ percentile. The effort to get to 99.5^th^ is very great and provides diminishing margins of return.

However, most developers are not in the 99^th^ or even 90^th^ percentile (by definition), and thus still have room for improvement in programming ability, along with the important skills just mentioned.

I am not convinced by the idea that developers are either born with it or they are not. Where’s the empirical evidence to suport these types of claims? Can a programmer move from say the 50th to 90th percentile?

Jeff points out that study after study shows there is no relationship between a programmer’s amount of experience and code quality or productivity. I don’t doubt that for a second. I’ve worked with developers who have 10, 15, 20 years in the industry and couldn’t code a virtual rat through a maze consisting of two parallel lines.

But recent research points out that the belief in innate talent is “lacking in hard evidence to substantiate it” as well. I wrote about this topic recently in my post, The Question Of Innate Talent.

So how do I reconcile these seemingly contradictory statements?

Well going back to the Scientific America article, The Expert Mind, we get a clue.

Ericsson argues that what matters is not experience per se but “effortful study,” which entails continually tackling challenges that lie just beyond one’s competence. That is why it is possible for enthusiasts to spend tens of thousands of hours playing chess or golf or a musical instrument without ever advancing beyond the amateur level and why a properly trained student can overtake them in a relatively short time. It is interesting to note that time spent playing chess, even in tournaments, appears to contribute less than such study to a player’s progress; the main training value of such games is to point up weaknesses for future study.

So what we learn from this research is that experience does not matter to a person’s performance. Exactly what the studies cited by Atwood support. However, what does have an impact is effortful study.


This makes a ton of sense to me. Typically, experience in any field, especially software development, often means solving similar problems over and over again using the same techniques as before. There is no way this contributes to being a better programmer.

Sure an experienced developer had to learn new technologies to stay relevant, but if the experienced developer applies the new technologies in the same way over and over again, the developer has not improved.

I’m currently looking at some legacy C# code written in a procedural style. The developer can write C# code, but hasn’t taken the time to learn to recognize when object oriented patterns would help solve a problem more elegantly. Thus his experience has not made him a better programmer.

However, with focused effortful study and training, a programmer can lift him or herself out of mediocrity. It’s not by programming more one gets better, but by programming better. Even better is to program more better (as in more often and better).

Keep in mind though, that this takes nothing away from Atwood’s main point. My point is that most developers (by definition) are not in the 90th percentile, at which point there is diminishing margins of return for effortful study. Most developers still have room for improvement in their coding skills as well as the ever important tangential skills to which Atwood refers.

In fact, I believe that for many developers, the tangential skills will distinguish and increase the value of a developer faster than improving programming skills. But don’t toss out that book on Design Patterns just yet.