Sketches are quick, dirty, and cheap and that’s exactly how you want to start out. Draw stuff. Scrawl stuff. Boxes, circles, lines. Get your ideas out of your head and onto paper. The goal at this point should be to convert concepts into rough interface designs. This step is all about experimentation. There are no wrong answers. — From Idea to Implementation — Getting Real
In the last episode I talked about high level goals for the project and I created a framework of rules for myself in order to help me make decisions. In this episode I’m going to follow one of the rules I created for myself:
11. Each feature must begin with a story discussing the feature and its impact, followed by sketches of APIs and GUIs, followed by test coverage, followed by implementation.
One of the questions we ended with last time was, “How do I know that what I’m building is what people really want?” The answer is: ask them.
I’m going to follow a principle from Extreme Programming that says you should have a customer on site with you as you build. If you can’t get a real customer, one of you has to play the role of Customer. That means I’m going to play the customer for my own application.
Is this legal? Yes and no. Under ideal circumstances the customer should be someone who is not a developer because features are a business decision, not a programming decision. Really, it should be the client paying for the engagement or, in the case of retail software, someone from the business side of your organization who is familiar with the domain and the needs of your customers. By playing both roles there’s a possibility that competing demands will cloud the judgment either of the Developer role or the Customer role.
At the same time, I’m a customer of Jetrecord. At least, I plan to be. So in a sense I’m building software for myself, so perhaps the two conflicts cancel each other out.
In any case I don’t have a choice.
Why Agile (or something like it)?
In my limited experience as a developer I’ve come to discover that programming is simply a conversation you have with your code. Just like writing an article, it may begin on one topic and end up on something completely different. Thinking about programming this way fits my personality and the way I learn new things. I love things that grow and take on a life of their own. I also like to look at the big picture several times during the course of a project. If I keep my head down for too long it’s easy to get lost in the details and pretty soon I’ve veered off course.
Certainly every circumstance is different and you have to adjust for the needs of the project. Agile, Extreme Programming, and the like don’t work for everything. When the choice is up to me, though, and the type of engagement is a good fit, I like the communication, freedom, and feeling of progress Agile affords.
And, of course, let’s remember that I’m not following Agile methodologies to the letter. I’m using methods that will help me get the project done reasonably and responsibly. I believe this follows the spirit of Agile, if nothing else.
The Basic Features
The customer’s job is to define the features of the application. Here they are in broad terms:
Log flights
Visualize flights and hours
Export flights
I’m not going to worry about advanced and/or paying features yet. We’ll save those for later. If I don’t get these basic features right, there’s no point moving forward. These are the features that I, the Customer, believe bring the most value to the application at this point in the project.
Creating Stories
An Agile feature story is a “promise for conversation” according to Alistair Cockburn. That means, it’s important to get it down on paper so that we, the customer and developer, can talk about it at length later on. I’ll make sketches and write tests based on the stories I create. As for detail, I’m going to shoot for making each story completable in the span of one to two weeks, including tests.
Logbook Feature Stories
Users can log flights via a simple form
Users can log flights via a table/spreadsheet form, similar to paper logbooks
Users can log flights by importing CSV files from other logbook software
Users can log flights via authenticated API
Users can see tallies of their flights and hours, separated into categories of experience and aircraft
Users can export flights via RSS, JSON, XML, and a number of CSV formats based on the same logbook software supported by the import feature
Discussion
That’s enough to start. None of these have enough detail to tell me everything I need to know about the feature but they help me start conversations with the customer. They also tell me that there are some fundamental features that need to be developed in order support the ones listed. I would look at this list and begin discussing the following:
What fields are included in a flight? Are any of them required?
Please attach copies of example pages from a traditional paper logbook.
Which third-party logbook software should be supported? Please attach examples of the CSV output created by the software.
Also, we’ve talked about flights, but flights belong to users. That means we need a system in place for managing users and their needs.
What is required for someone to become a registered user?
How are users identified? By ID, by login, by real name, by nickname?
Do users manage themselves for things like registration, password changes and recoveries? How much intervention is required by admins?
Can users delete their own accounts? And what happens to everything they’ve created on the site when they do?
What about multi-user accounts? How are they managed? What happens to a user if the account they were associated with closes down?
User Feature Stories
Users can self-register and manage every aspect of their account, including paying for services and deleting the account, without intervention by an admin. Since email is the only method used to communicate with members, email is required to signup. Users must verify their email in order to begin using their accounts. Users are identified on the web site by their nickname. Users should receive email notification upon successful signup.
Users can login from the web site or from third party applications.
When users delete their accounts, all of their flights are deleted, followed by any aircraft used (assuming they’re the only pilot of the aircraft), followed by any routes used (assuming they’re the only pilot of that route), followed by their user account and any associated files. Users should receive email notification that their account has been deleted.
Accounts with more than one user have certain admin users who manage the account, including payment details. Admins can manage permissions for other members and invite and kick members.
When multi-user accounts are deleted by an admin, associated users will have the option to continue paying for features at the same level, upgrade to higher paying options, downgrade to lower paying or free options with features removed, or have their accounts deleted.
Discussion
It looks like we need a fairly complicated user management system in addition to being able to log flights. Ideally, with a larger team, we could work on both at the same time. However, since it’s only me, and since I’m also the customer, I’ve decided that the user management system is the first thing I’ll do after setting up my development and production environments.
Sketches
Again, these are sketches. Not blueprints or wireframes. Not schematics or Photoshop comps. They have just enough detail to let me know what I’m building so I can begin creating HTML mockups, acceptance tests, and unit tests. Ideally, as the developer, I’m drawing these while seated next to the customer so we can make changes quickly and throw things away if need be.
Flight Forms
User Forms
User API
Inventory
At this point I should be able to take the sketches and start building the application using whatever language I choose. I had a great time using the Ruby on Rails framework to build version 1 and I’ve decided to use it again for version 2. It’s well-suited to handle an application of this type and the Ruby language is, for lack of a better word, fun.
These are the things I need to get going (note: gems are installed with RubyGems):
Optional: Ruby on Rails comes with a simple web server that we can use for development purposes. However, I’m going to set up and practice using the server that I will use in production, even though the OS will be different:
I’m not going to cover installation. I assume you can follow the instructions given on their respective web sites and that you know how to use Google. Assuming you’re going to follow along as I write code, if you’ve never worked with Rails before I recommend installing everything and working through any of the simple tutorials you can find on the web before continuing on.
Also, please note, from here on out I’m not going to talk about the topic of programming per se. If you’ve never programmed before, the rest of the series might be interesting to watch but otherwise might go over your head. Let me say, though, that there are hundreds of web-based tutorials that can teach you how to program. There are even some that use Ruby.
Coming Up in the Next Episode
I’m going to set up my Rails project and my Git source code repository. Then I’m going to make sure I can do an initial release. Until next time, cheers and happy flying.
Material You May Find Useful Related to This Episode
We were out at our family cabin in Bolinas, and he was at the kitchen table close to tears, surrounded by binder paper and pencils and unopened books on birds, immobilized by the hugeness of the task ahead. Then my father sat down beside him, put his arm around my brother’s shoulder, and said, “Bird by bird, buddy. Just take it bird by bird.” –Anne Lamott
I don’t think any of us ever realizes the hugeness of the task until we sit down with our materials and attempt to make sense of it all, even though the whole making-order-out-of-chaos thing is highly overrated in my opinion. Nevertheless, that’s what I’m going to do here in this series. I’m going to take a seemingly simple task–building an online logbook for pilots–expose the underbelly, cut it apart, and then put it all together into a fantastic, cohesive product that everyone will love, including me and you and every pilot you know.
Why? Because I believe we are created to do two things: experience pain and build things. Sometimes these are one and the same. I hate to admit right from the start that this is a task to which I feel compelled, by which I feel repulsed, and about which I feel nonplussed but apparently we’re in a new age of transparency so I’m starting with me, as Michael Jackson would say. I both love and hate computers and projects. And it always happens that about the time I’ve finished a web project I’ve forgotten all about this crummy beginning stage and am strangely excited to do it all over again.
This is not an exercise (please place your hands inside the yellow circles); it’s the process of building a genuine application that is already live on the web. What you see here is what will be released for version 2. I’m hoping that by sharing something real you’ll be inspired to create your own projects and you’ll see what it’s like for one developer to wade through this stuff.
My other hope is that we can learn from each other. I am not an expert, as will become more apparent as the series goes on. Please contribute your knowledge in the comments. Can something be improved? Do you have a suggestion or a critique? I want to hear it. Please keep it constructive, though. Shuriken are not allowed.
If you’d like to contact me by email concerning the series, use harry@harrylove.org.
The Scope of the Series
The process of creating a web application from start to finish involves several subjects and my plan is to explore breadth rather than depth. That is, I’m going to talk about whatever will help me complete the project and I’ll mostly concentrate on the places where subjects overlap. If I don’t explain things in enough detail, I apologize. The truth is I may not know enough about a subject to explore it in depth. I’ll try to point you to the places I’m getting my information. Don’t hesitate to speak up if you’d like me to say more about a particular topic. If I have something more to say, something worth creating a new episode for, I may just do it. Otherwise, of course, I’ll point you to my sources.
Let’s Start with Questions, Goals, and Rules
I’m going to start this project the same way I would approach an essay, article, or presentation, by asking questions and then stating goals and rules. You don’t have to do it this way. This is how I do it because it works for me. I’m going to say that just once. It applies to everything I’m going to do from here on out but I don’t want to bore you to death by repeating myself every episode (there’s an important programming principle in there somewhere).
Questions
What do I want to build?
I want to build a web-based logbook for pilots.
Why do I want to build it?
I love aviation, data, and the connections made when data cross paths.
I am building this because it’s a product that I myself will use.
I have surveyed the available logbook software, both desktop- and web-based, and have not seen any that use the same approach to building logbooks that I’m using to build Jetrecord.
I believe this is a product that can work in concert with existing desktop- and web-based products.
Is this an art project or do I need to consider my audience?
It’s both, but I am building this product for public consumption so I need to consider my audience.
Who are the primary people groups in my audience?
Pilots
Air and ground crew
Affiliates
Aviation buffs
Their family and friends
Do these people need this software? Why should they care?
Pilots must show proof to governing authorities that they have obtained any necessary knowledge and flying experience. Logbooks have been the standard method for recording flying experience for several decades.
The second two groups may have business concerns with the logbooks but will otherwise not need a logbook of their own.
The last group is important to consider because the entry of each flight will be made public, suitable for sharing with family and friends in the same way that family and friends share news, pictures, and video through blogs.
Can I afford to build it?
The operating system, programming language, web framework, web server, and database server I am using are all available under open source licenses.
I have a computer that is well suited for the purpose of building a web application using this software.
The web application is currently hosted at Joyent using the $125 per month Accelerator option and there is room to grow using the current resource limits; there are both horizontal and vertical scaling options available at Joyent should the application outgrow its resources.
My company bank account is free but I may need to switch to a paying account if Jetrecord becomes very popular to the point of making more than a few hundred deposits per month.
The merchant account and credit card processing fees per month have been quoted from a reputable business at less than $50 per month.
The cost of doing business in Longmont, Colorado as a Sole Proprietor is a small annual fee.
I have access to help in preparing my taxes.
Can I afford to maintain it and keep it running for years to come?
If the current ongoing fees are any indication, the answer is yes. Obviously, in the best case scenario, if Jetrecord suddenly jumps from 1,000 users to 100,000 I will need to upgrade the servers a bit. I may also need additional partners or staff to help manage the workload.
Do I have enough domain knowledge to build it?
I have enough knowledge about aviation and logbooks to make a good product.
I have already built version 1 and it has been well received by the pilots who use it, including both general and commercial pilots.
Do I have access to information about the domain?
I have access to the FAA’s regulations concerning logbooks through their web site and also via the FAR/AIM book.
There are many web sites, magazines, and books that focus on issues concerning pilots and aviation.
Do I have access to customers and are they willing to teach me about the domain?
I have access to the current users of the product.
I have spoken with several pilots over the phone who have expressed interest in helping me.
Goals
Succinctly and broadly, what is the desired end state? For example, the goal of basketball is to have more points than your opponent at the end of the allotted time for the game.
The logbook is the key application, regardless of any related features.
It must be as easy as possible to get data into and out of the system both from within the Jetrecord web site and from the developer API.
It must be as useful as possible to as many classifications of pilots as possible, but focus on general and commercial aviation.
Mantra
Pilots Aren’t Robots Yet!
Rules
I’m writing some rules for myself that will serve as a guide for the decisions I make. They are written with this project in mind only.
Rules should help us make decisions; this is the only criteria for adding a rule.
Modify or discard any rules that are not helpful.
Use the best design principles we know how to use and spend time researching concepts we are unfamiliar with.
Make forward progress; no product is perfect.
Follow the spirit of Agile methods for software development; remember rule #2 above
Keep documentation in the source code except where sensitive information may compromise security or business secrets.
Documentation is for developers and is an interface into the system; design it well.
Everything should be added to the source control repository unless there is a good reason not to.
Code that can be extracted into a shared library (gem) or plugin should be.
Favor gems, plugins, other libraries, and framework code to writing our own unless there is a good reason not to.
Each feature must begin with a story discussing the feature and its impact, followed by sketches of APIs and GUIs, followed by test coverage, followed by implementation.
Metrics should be built into the project early and often.
Questions drive the need for answers which drive the need to record data, not the other way around.
Build for data access first and GUI display second.
Build for universal access.
Build for progressive enhancement.
Browsers must prove their worth to be included in our support; mainstream use is not proof in and of itself.
Next I’m going to write a brief outline of the basic application. I’m also going to sketch the application on paper and make a list of parts I need to get started. I hope you’ll join me. You can subscribe to the RSS feed in order to be notified when the next episode is up. I have also created a table of contents page for this series. Until next time, cheers and happy flying.
Material You May Find Useful Related to This Episode
Screencasting! It’s the teaching medium that’s sweeping the nation!
I love screencasts. Certain ones, anyway. Done well they enrich understanding. Done wrong they’re a snore and a half. I’m in the middle of redesigning Jetrecord and the more I redesign the more I’m considering redoing it from the ground up (and doing some screencasts about the process).
Why? Well, I’m looking at all the shortcuts I’ve taken to get to where Jetrecord is today and frankly, it’s embarrassing. Don’t get me wrong. I LOVE Jetrecord and I think it’s both a great idea and a great product. I just think I can do better. I owe it to my customers to make it the best it can be.
I’m also one of those people who loves great foundations; well, I love the idea of great foundations. I love talking about foundations, dreaming about them, planning them, scheming. But when it comes time to build the application I get so focused on delivering a great experience and a great product that I glance right over the possibility that I may change my mind on a few foundational things, that people involved in the project come and go, that features come and go, et cetera. I know it’s impossible to plan for everything but I know there are some things I could do a whole lot better.
When I build an application I put a whole lot of time into building great user experiences (I try, anyway) and just enough time into making sure the application keeps chugging along. I tell myself, “it’s not a problem until it’s a problem,” which is a great way of looking at life for most matters. 80% of the time, we really don’t need to plan for all the contingencies we plan for. Of course, it’s that scary 20% that keeps us awake at night with a death grip on our Blackberries waiting for that 3am email from the error system. (Thanks, Internet!) An example of a 20% problem would be the current financial crisis in America. We were doing fine ignoring the underlying foundational problems for a while but now it looks like we picked the wrong year to stop sniffing glue. So when it’s a problem, it’s a real big, fat, hairy problem, with mutated teeth growing out of a tumor in your stomach.
Granted, most of us don’t encounter these problems. For small, local, mostly internal applications, applications in which we and maybe 20 other people are the only users, the 20% problems can usually be contained. It’s when we move into the realm of needing to support thousands of users in multiple countries on a 24/7 basis that we start encountering the mutant teeth kind of problems.
Jetrecord isn’t quite the super-multinational app yet but I believe it’s going there. It will, I tell you!!! Moreover, I’m at a place in its development where stopping and doing the right thing makes sense. I’m redesigning anyway and my plans already involve a better foundation. If I wait until it actually is supporting 20,000 users in 20 countries, I’ll have quite a job on my hands. I’ve racked up some code debt and it’s time to pay it off–and start saving a little.
But why do a screencast about it?
For one thing, I’ve been a code leech for most of my career. I’ve written a couple free things here and there and released them into the wild but nothing too substantial. Mostly I’ve read other people’s books and articles, attended other people’s conferences, and taken other people’s free code samples. And truly, all of these things have taught me a lot and I wouldn’t be where I am without the generosity of many, many people. But I think it’s time I did something substantial, extraordinary, and excellent for others.
Second, it’s going to force me to write the best code I can. I’m embarrassed during code reviews if there are glaring mistakes so I’m going to try hard to make sure it’s worthwhile code because it’s public. I already know I’m going to have to do a lot more research to find best practices and I’m sure I will make mistakes and I hope someone will be kind enough to suggest improvements. This will benefit me and you as coders and ultimately it will benefit the applications we build and the clients and the users we support. Win-win-win-win-win.
Lastly, I’ve been doing screencasts since 2001, since before they were called screencasts. In one of my previous jobs we were publishing “help videos” in Camtasia on our web site to show users how to use the site and some of the software products we supported. I consider myself knowledgeable enough to make a good video, or at least, to give it a go. Whether they’re actually good or not is up to you.
Having said all of that, I still haven’t fully settled this in my mind. I want to do it. I know it would be good. However, screencasts take a lot of time to do right. For every minute of production video there’s 30+ minutes of material and mistakes that don’t make it in, not to mention planning, outlines, scripts, research, creating or finding other media like music, artwork, and extra video. Then there’s editing, producing, and finding a place to host it. It’s not unlike being a T.A. in grad school. You’ve got this whole other full-time teaching job to do while you’re trying to complete your own studies. (At least, that was my experience.) It’s both fantastic and super-sucky at the same time. And you don’t get paid!
So, if the series is going to show up, it’s going to show up here on this web site, and probably very soon because I’m already building the foundation for the next Jetrecord (which is going to rock, by the way; if you’re a pilot, you should use it). If I wait too long, I definitely won’t do it.
I attended Startup Weekend in Boulder, Colorado this past weekend and I’d like to encourage everyone and anyone looking to start any kind of company to attend one if they bring the event close to your town.
The format of the weekend is like this: you gather together Friday night, meet awesome people, propose ideas, gather enough support for your idea (in terms of human capital), and then spend the rest of the weekend assembling your idea.
On Friday night there were roughly 100 people in a small lecture classroom on CU’s campus. All of them are smart, motivated, tech-focused people, the kind of people you always hope you get a chance to work with, unless you’re already working for a startup of some kind, in which case, you probably already are. Everyone’s on Twitter and/or Facebook and/or every other social web service. Everyone reads the same people you read or knows someone you should be reading. Everyone knows the same current events on the web you know or knows something you should know. Some of them have flown across the country to attend.
I pitched a few ideas but ultimately decided to work with a small group that wanted to create a poor man’s content failover service for web sites that receive unexpected Digg/Slashdot traffic. We called it Hitsurance. By 11pm Friday night we had our idea, a name registered, and a dedicated server set up with a splash page and email submission form. I went home and crashed with my brain racing away on ideas well into the night. I woke up twice to write some notes. Needless to say, sleep was fitful.
We worked Saturday from 9 to 9 and got about half way there. We had a design, a logo, the basic premise for the demo fleshed out, a blog, SEO research completed, Google Apps for Your Domain acquired, a Twitter account, and some of the backend scripting in place.
All work and no play would be boring, of course, so part of Saturday the attendees spent time socializing, working, eating, and bouncing ideas. We had a short concert from local singer/songwriter Reed Foehl, which was unexpected and strangely appropriate. By the end of Saturday most of us were fried. I went home and crashed but again, didn’t sleep very well.
We returned Sunday and worked from noon to 7, finishing up as much of the demo as we could. We got the service working around 6:30pm with just enough time to eat and do some fine-tuning before the group demo session at 7. Our demo worked as expected and, dare I say it, the site and the signup experience look pretty sharp considering we started from nothing Friday night. As for the future of the service, my co-founders and I are still discussing it. It’s definitely a useful service and a win-win all around. Content producers keep their audiences going and hosting providers stave off potential server crashes, which makes the rest of their customers happy.
You can read about the other groups at the Startup Weekend re-cap. I think the Web2Splash idea is both cool and useful. I’m also interested in seeing the IMDB for Podcasts go forward. It would mesh really well with one of the ideas I proposed. If Jetrecord doesn’t take up all of my time I might contact Andy in the future.
Now how much would you pay?! If the whole experience of building a product over a weekend with good people is not enough, all the groups got to pitch their ideas and chat with Loïc Le Meur, Jeff Pulver, Guy Kawasaki, Eric Litman, and Stowe Boyd, each individually. Where else do you get a guaranteed chance to do that, especially with an idea you just came up with the night before? All for $40?
I had a great time meeting and working with everyone and to boot I got to use a few Perl libraries that I haven’t tried before.
Again, I highly recommend it to anyone thinking about starting their own company, especially one with a web focus. But any type of high tech idea is a good fit for this event.
