Someone asked: "how do you design your RoR web application?"


To be honest, there's no right or wrong way to create a new app. Some people starts from DB tables, some starts with html, and others starts with user cases. Whatever works for you is the best way. Here are some pointers.

Coding

• keep your design modular. If things can be broken into small pieces, do so. make them into plugin, gems. because doing so those building blocks can be reused later. if everything is big and chunky, less likely they'll be reused.
• tests is the other half of the code, code without test is not complete. you should always write code with test code side by side. Some people even advocate write test first.

If you just keep in mind of 2 principles above, you'll be doing very ok.

Development lifecycle - Iterative

Break big apps into iterations, each iteration with requirements from most critical down to nice to have. People call this agile development. In my example, my final goal for groups.wuyaSea.com is to create a full featured social network site. First release only have most essential features, user can discuss in forum, and write articles, and make friends. that's it. No search (outsourced to google search), no photo sharing, no private messaging, not even tagging (now does). With capistrano, it's easy to keeps on pushing out releases, I often push out new builds like 5 times a day.

Major reasons for doing iteration is to minimize risk, cuz bean counters like that.

• each release requires shorter timeline, less coding effort, and more predictable in time and money.
• after you give each release to user acceptance testing, you'll find out what features users use most of time, and what features users rarely touch. It's known that in any app, 20% of code does 80% of work, so why write 80% of code nobody uses? In my case, the web is my user acceptance test.
• so release early, release often. In venture capital world, it's called startup fast, fail fast. Think your startup will be the next facebook or youtube? think again, for every successful startup, there're dozens failed ones you never heard of, so why waste 1 year to learn the truth if you can do it in 3 months?
  
• one side effect by releasing early is you get to earn that fabulous adsense dollars right away (although i haven't earned a dime yet). Cash flow is always important for anyone who bootstraps.

Maximize your effort. User requirement changes, site traffic changes, life sucks, blah blah. there's no one design that satisfy all the requirements. Changes are unavoidable, eventually rewrite may be the best solution. The best design is one that satisfy the most requirements with least cost, don't over engineer something.

Failing is learning

I can tell you what i've done. When I was creating groups.wuyaSea.com, i went through 2 prototypes, after a few months, each one becomes too rigid and unmaintainable, then throw it away, and start over. Current site is my 3rd implementation. Making webapp is like doing everything else in life, you learn by just doing it. keep cranking, if it fails, keep trying again.

hope that helps.

Dorren