I had to move one of the web services we offer in my regular job to its own server. It was a good change, we moved the entire app from a CentOS/cPanel box hosted on the Planet to an instance running Ubuntu on AWS.

CPanel is a really good product and saves you a lot of time but it is a pain in the butt if you are serving more than shared hosting accounts. If you have ever tried to automate tasks that handle virtual domains, email accounts and DNS entries you know what I’m talking about.

Anyway, I must admit that I’m having a blast wearing a sysadmin hat during part of my day. I’m learning a lot. I have this great excuse to read some books that were on my TO READ list for a long time.

One of these books is Pro DNS and Bind by Ron Aitchison. I have had this book on my desk for months but I finally finished reading it. The book gave me a nice introduction to DNS servers, resolvers, zone files, resource records and diagnostic tools. Highly recommended.

The cleanup included my bookshelf. My mission was to dispose of as many books as possible. Boy, books are heavy to carry and physics does not apply when it is part of your luggage.

While running through the titles I started to compare dates of purchase against subjects. I have had this habit since I was a teen, every time I buy a book I sign and date it.

Anyway, my small library was another lead to an obvious issue: my quest to become a programmer is a mess.

Here’s one good example. I wanted to code games when I was a teen. After some research I concluded that C++ and DirectX were the right tools for the job. I decided to start with C++ and ordered “C++ How to Program” by Deitel. Back then, a small number of web pages would advise: “You have to learn C”. My mind discarded that advice as fast as I could read it. Why would I learn C? Isn’t C++ not only C but PLUS PLUS! It is the future. Yeah, I know you are laughing at me.

I read and tried all the examples from “C++ How to Program” before reading or writing any line of pure C code and felt ready to dive into some DirectX books and code the best tic-tac-toe ever!

Turns out my ambitious plan needed a detour. You have to go through some Windows programming before you can dive into DirectX (I don’t know how it works now a days). Gosh, 70 lines of C code just to print Hello World in the Win32 API world.

I can’t just copy and paste code so I ordered “Programming Windows” by Petzold to become familiar with Windows programming. BTW, Programming Windows is a great book — and I found a minor mistake in the text! (Erratum #5).

I read the book, tried the examples and the fact is: Win32 API is C Code.

Because of that I finally ordered the classic “The C programming language” book to fill this awful gap in my skill set. Comparing the dates of purchase I realized that I bought the C book 2 years after the C++ one. Sad.

I now remember that after going back and forth I finally achieved my goal. The best tic-tac-toe game ever written did compile and run without crashing. The euphoria lasted for almost 5 seconds!

Before writing my first line of code I was already humble and comfortable with the idea that it takes a lifetime to be a great developer. BUT, it took me a long time to have a better overview of the skills set you need to build in order to be a software developer.

It’s a matter of proper guidance.

It doesn’t matter if it’s a CS degree or guru souls willing to show you the way.

A few months ago I added to my schedule a Windows Day. Brand new world. I dove into Vista, ASP.NET, C#, IIS, MSSQL and Visual Studio one day per week (and occasionally weekends) for a few months.

I was excited because of the novelty. I read a few books on the subject and had  to play with it.

Here are some random thoughts about my experience.

Right on the first day I got frustrated using the mouse for simple tasks like moving and copying files around so I installed cygwin.  A difference between two worlds.

I learned later that there is a Windows option for the command line: Powershell. Although I didn’t try it the Wikipedia article gives me the impression that learning Powershell is probably a good investment for Windows developers.

Back to random thoughts.

I was then ready to try Visual Studio. It is amazing to try examples from books just dragging and dropping icons that build some correctly written snippet of code. I just doubt that experienced developers would follow that approach.

One problem I had with Visual Studio was trying to set the Emacs key bindings. Leaving the keyboard home row to reach keys like arrows, page down and page up is disturbing. I followed this post advice and installed Emacs Everywhere.

The next step was the database. Oh boy, choosing a MSSQL version is so damn confusing! I’m probably getting older and dumber but I couldn’t understand which option to download from the MS website. Installing MSSQL in Vista was quite painful, too. Had to follow some MS black magic recipes to get it running. Installation on Windows 7 Beta was a breeze though.

After setting up the database I went to the HTTP server. IIS is pretty straightforward. You can setup a working environment with just a few clicks.

Working environment ready and I started to follow the books’ examples. A few ASP.NET web forms later and I was ready to start learning C#.

Talking about computer languages is somehow a taboo so it should suffice to say that C# is verbose.

Well, a few months after my first Windows I’m back to a Linux side project but I’m glad that I took some time to get an overall idea about the ASP.NET world. It’s not for me. Once you go open source you never go back.

If you are only interested in building your client project it doesn’t really matter the platform you pick but it’s a disturbing feeling that you can’t build IIS or poke Visual Studio’s source code. There is this sense of being a powerless and passive consumer always trying to catch up every .NET update.

I don’t know how it works here in the US but years ago banks in Brazil came up with a file format dubbed CNAB to exchange financial transactions. It’s basically a layout where data fields are delimited by their byte positions. Seems pretty straight forward, right? So try implementing one and you will feel the pain.

The first problem is bureaucracy. You need a business account, lots of papers, signatures, phones calls and of course talking to Elaine. She is your bank account manager, really polite and willing to help but obviously completely clueless regarding any technical question. Most of the time she is not allowed to put you in contact with someone who works all day writing Cobol and has all the answers to your questions (I actually empathize with the Cobol guy and pity him whenever he needs to answer the phone).

After everything is set and you are allowed to interact with the bank system via their file format they email you the technical instructions. It’s a word document detailing the layout. Prepare yourself to learn how it works through trial and error and a lucky chance to talk to someone on the phone (yes, a real human being) who actually knows what is the default char for the field delimited between bytes 116 and 118. Is it a blank space or just a zero? Can the file be encoded in UTF-8?

You just need a second bank to send you their layout documentation to prove your inner fear. Each bank implements its own CNAB layout version. This is the time you start praying for a framework to deal with all layout inconsistencies across bank institutions.

All is not lost though. After going through all this you have a new skill set. You are ready to work at any bank institution because you did your homework. You have grasped an understanding of all kinds of bank transactions. You have to do this even if you only want to automate monthly invoices for your customers. It’s a hell of a ride my friend.

Now, I’m not trying to undervalue the CNAB file format. It was probably a good solution 20 years ago. It still works. The problem I see is how banks are outdated compared to companies like Paypal and alike. What are they thinking? Can’t you just build a web interface to deal with your internal file formats? Just give me a Restful Web Service in XML or JSON so I can send monthly invoices or pay the employees.

After installing sloccount* I immediately started to harvest data on all projects I currently have on my hard drive. I was really impressed by the first results. The costs are too high, I thought. A small project with 2,000 lines of PHP code like the Autoresponder has a development cost estimated at more than US$55k.

After reading sloccount’s user manual and pondering about the Awake Autoresponder I started to change my mind. 2k lines of PHP code encompass a lot more work than the number suggests. A few reasons for this:

1. The code was updated every time a customer submitted a bug. As you know, fixing a bug does not necessarily imply more lines of code
2. Out of the box sloccount doesn’t compute HTML, CSS or Javascript
3. Technical support
4. Documentation

It took me a few thousand commits, I don’t know how many emails and several hours of night debugging the script on customer’s servers to reach 2k working lines of code. It may sound like cheap advertisement or self promotion but I must say that US$49.90 for the script is a bargain!

* if you are running Ubuntu it’s easy as

sudo apt-get install sloccount

Anyway, chances are I’ll be devoting fewer hours developing for my current employer  so I decided to improve our documentation base.

Documentation includes how to properly setup a working environment, main concepts of the system, shell scripts that do black magic, coding guidelines, etc. I’m also writing down a lot of ideas to improve the current system we have and formating TODO lists that only make sense in my head into prose.

The problem is that part of the system was first developed and discussed only in non-formal meetings. Yes, productive and get to the point meetings. We didn’t have the chance to properly document everything as dictate theory and common sense.

So, right now I’m writing. A lot.

One of my main goals is to make the life of future developers joining the team easier. What are the first steps? What is all that code supposed to do? What needs to be done? How it can be done?

It forced me to stop and look back. Stop being a GTD code maniac and look back at everything that was done in 3 years.

It’s unbelievable not only the amount of things we’ve got done (lines of code in my case…) but also how much we improved. It was crazy, it is still crazy, but we fixed bugs, released updates and new features almost every day. Ohhh… s t a r t u p s.

Well, this morning I did put into practice something that is most certainly not new for you: the post-commit SVN hook. As an experienced developer you can stop reading the post right now.

The post-commit is now active and all developers (the development team is small) receive an email after a commit is done. The email contains the revision number, author, log message and the diff generated by the commit.

I learned about this procedure reading the book Producing Open Source Software.

The goal is to help everyone involved in the code to have a better idea of what is being done. Who is working where and most importantly, if someone is touching part of the code that is of special interest to you.

After a few commits and just skimming through the code you’ll see the benefits.

You start to police yourself to make small and well documented commits. Good.

It’s also easier to comment about code that was introduced/updated by your peers. The diff is right there in your email box and the changes are still fresh in the author’s head. Just hit reply!

In 2003 I was working in my home-office in a small town in Brazil while trying to get a college physics degree. To pay my bills I was basically building websites in PHP after I left the Perl world. I was part of the staff for a project designed and managed by a grumpy old man in New York. God, I should have sent him a copy of Peopleware (Peopleware: Productive Projects and Teams (Second Edition)). One day he subscribed for an autoresponder service. It was not cheap and the service was not good at all. I thought I could code the same features and add the others we needed in no more than 2 days. I did it in a weekend and started to share it for free. Back then I had no idea what FSF or OSI were. I started the professional version a while later since it was taking crucial time out of my day night to give support and add new features.
Since the release of the first version I receive emails daily from people reporting bugs, asking for new features, giving suggestions or simply wanting to say hello. Well, although it might not help me to sell more copies, I must be sincere about my efforts into this project. Time to work on it is scarce. I switch between my regular job (there are always new projects at a startup) and research mode at night.

Releasing is not easy. I suffered from it and some symptoms I believe are shared amongst a lot of developers. One is frustration. You know you can do better but time is against you. Another problem is loss of focus. I learned how to manage and maintain this project (more on this some other day) in the hard way. I made all the common naive decisions like trying to rewrite it from scratch or adding a new feature to make a new sale.

Not everything wrong though. I’m still working on it, right?

One major key: Releasing.

It taught me a few lessons.

The first version was small and built to do only one task. It solved one problem of mine but it also worked for a few more users.

The main benefit: Feedback.

I not only read and pay attention to all the emails people send me, but also take note of each one. There are two possibilities here.

1. add a ticket to the bugtracking as a feature to be built or a bug to be fixed.

2. take notes in my to-do log for random idead.

After a few tickets and paragraphs I started to notice patterns and features people wanted most. This way I avoid writing fancy features and make better use of my time writing features that people actually want.

The downside of this approach is, unfortunaltely, the elimination of these “fancy features”. One of these features could turn out to be innovative and a killer feature for the app. Right now, I think this is not a concern. I have all  these crazy ideas being developed and written down in my notes. Today these features are expensive for me to code. Meanwhile I have fun building a more mature version of the autoresponder.

To fix the problem I just had to open
/usr/local/lib/php.ini
and move the line
extension=”pdo_mysql.so”
to the end of the file.

Restart Apache and that’s it!

Hope it helps if you are dealing with the same problem.

The idea is that all subdomains point to the same location on your server. You can then fetch user data from a database using the subdomain name as a key. We use a lot of PHP here so our key would be something like $_SERVER['DOMAIN_NAME'].

I first thought of building a small script at the server to manage the subdomains. List, add and delete. Boring.

The server for this system had cPanel installed. It is a good tool but it took me a while to find out I could handle all subdomain errands with CPanel’s XML API. The documentation is very poor and lack details. By the time of this writing you don’t even have a complete list of methods for the subdomains:

CPanel API

You can see here that the only method listed is for deleting Subdomains. Too bad. Lost some precious time digging cPanel code.

Anyway, here’s a simple class that you can use to list, add and delete subdomains if you have cPanel installed on your server.

You’ll probably need to adapt the code to your needs and enforce error checking.

cpanel.tar

When I thought about that time there are two things that I learned and I still take benefits from it:

1) how to drink like a man
2) how to study

Well, for number one the good thing is that you can hang out on Saturday night and still be productive in the next morning. No extreme hangover as an ol’ friend once told me.

On how to study I took the lesson from real guitar players – then Musicians. You need method. Otherwise you’ll play 10 hours a day and still suck. No matter what skills you want to improve you need method. Otherwise you won’t be able to play that killer arpeggio that make teens to listen to the same lead break a thousand times.

Now, I don’t believe that there’s a perfect or magic method to improve your learning experience. It is up to you to discover and implement what works better for you. Two rules are mandatory though:

Rule #1 – Small steps

It’s like Jack the Ripper. Take it through pieces. It’s better for you to read 30 minutes a day than to spend your whole sunday with your face into that algorithm book you promised yourself to master.

This said, you need a schedule and this bring us to

Rule #2: Routine

Routine is the hard part. At least in the beginning. It’s easy to be distracted and procrastinate. A good way to motivate you to go through is to give you some credits when you achieve goals.

Make small goals. Goals that you can achieve. You can start with something like I’ll read some completely unknown technology to me 30 minutes everyday.

Let’s face it. 30 minutes won’t hurt you. In a week you’ll see that it’s a breeze. You’ll raise your bar for the next week.

My routine at this moment

Before building my schedule I had to couple with the idea that I won’t live forever – at least in this incarnation. So there’s a finite number of subjects and time to spend on them.

At this current stage of my life I have the ability to divide my learning time by

1) the mornings before doing “real” code. I go earlier to the office.
2) night time
3) small breaks during the day. Generally a 10 min break to let your subconscious deal with some problem you’re trying to solve.

For the subjects to learn I have priorities. Some subjects are mandatory and some are fun and new – and they improve your mandatory skills anyway.

I’m working as a software engineer in a startup company for almost 2 years. The language I deal most of the time is PHP. So, PHP is in my priority list. It’s very important for me to have a deep understanding of the language that pays my rent. I try to get better at the PHP world by reading books, magazines and even reading the language source code.

For the fun part I have included Python and Game Programming in my schedule. I mainly use Python at small side projects and use C for some stupid and small games. Games are a fun place for me to learn about algorithms and data structures.

I label everything else as Tools. In this section I have included 5 subjects. Right now they are Apache, Mysql, Emacs, Regular Expressions and SVN.

Ok, with the list of WHAT what to learn I insert it into my WHEN agenda.

1) Mornings – PHP. Here I read regular books, security with the language, what’s being done for version 6 and once in a while I try to read the language C source code.

2) Nights – Python or C Game Programming. Here I try to work on small side projects with Python/Django and sometimes coding small pieces of games.

3) Small breaks during the day – Here I set a subject for each day of the week. Apache on Mondays, Mysql on Tuesdays, Emacs on Wednesdays, Regular Expressions on Thrusdays and finally SVN on Fridays.

There are of course lots of other subjects that I’m interested in. Again, since my time is limited all other subjects are in my list for the future.

Having a schedule like this seems tedious at first. It’s not. In the end it’s all about having fun in what you do. If you’re not up to reading an algorithm book on Wednesday read something else or just go out to drink beer!! Salute.

Now my machine is running with 2 monitors on Ubuntu and with all fancy but cool effects from Compiz. We use 19″ monitors here. Hardware in Brazil is so damn expensive. Let’s say thanks to the government for all taxes and making the acquisition of TOOLS we NEED to WORK this HARD.

Anyway, I was curious on how people work with 2 monitors here. Right now, the best approach for me is to have Emacs on the left. On the right monitor I have the browser and the terminal. I leave the terminal maximized since it has transparency in it and it’s such a nice feature. I can type my bash stuff and still read what’s on the browser. Wohooo!! There’s just one feature missing in GNOME. I can’t set a different wallpaper for each screen. KDE has this option though.

Whenever I need to use another program or do some unusual task I go to a different Desktop. Two Desktops are enough for me. I can only do an unusual task at a time. I guess Rands is right when he says I Don’t Multitask.

After some time working this way I’m changing my mind about two monitors. I’m now willing to work with only one monitor. But it must be HUGE!!