Reaching Utopia, and Why It's Not

I've been programming in ExtJS (the Javascript framework) a lot recently. We were sick of writing the same stupid widgets and plumbing code every time we needed a list of names from the database. Since PHP doesn't seem to have a framework that fit our needs, we decided to just cut out as much backend code as possible and go with ExtJS and their Direct functionality, which let's you call your PHP directly from the Javascript. You can instantiate stores, wire them directly to your backend, and configure your widgets to display it, all without ever typing XmlHttpRequest. Cool stuff -- I recommend you check it out.

Except, I don't like it.

But, it sounds so good! Look how easy creating a one of these stores is!

new Ext.data.DirectStore({
    reader : new Ext.data.JsonReader({
        idProperty : 'id',
        fields : [
            { name : 'number' },
            { name : 'id'},
            { name : 'stage_time', mapping : 'stage', type : 'date', dateFormat : 'H:i:s Y-m-d'},
            { name : 'return_time', mapping : 'return', type : 'date', dateFormat : 'H:i:s Y-m-d'},
        ]
        }),
    autoLoad : true,
    api : {
        read : ISPATIAL.mission.get_all_missions
    }
});

That's production-ready code! It sets up a store, reads the data from our PHP function (mission class, method get_all_missions), formats the data for the client-side widgets, and automatically loads on creation or update of a widget that uses the data store!

Oh, wait. It doesn't work. I forgot a line: root : 'data', under JsonReader. Without that line, ExtJS will helpfully throw errors like "ds is undefined", or maybe it won't maybe my data just won't load. There goes my afternoon. [Also, for extra points, it doesn't work in IE. Figure out why, and I'll give you a gold star.]

But here's the thing. This isn't ExtJS' fault. Let's take a step back and talk about why I like programming. I've said it before, but programming is one of only a few disciplines where your ability to produce value is only limited by how smart you are. I've always loved that if I thoroughly understand a problem and know the correct solution, I can write that solution within 30 minutes. Those 30 minutes gives me plenty of time to manually test all my code for syntax bugs (most of which are caught by my editor anyway). The downside is that trying to understand everything about a problem is daunting, so I use that most powerful of CS tools: abstraction. I can break larger problems down into small, testable, reusable modules. Building these back up into a useful solution allows me to think about where I'm going, instead of getting bogged down in the details of "How do I make sure my database connection isn't opened until I actually need it?".

But even that isn't good enough. When I have enough of these tiny, generic, reusable modules sitting around, I need a way to organize them. So I group them together in logical ways. But the default action isn't always right, so I add some configuration options to them -- parameters, config files, whataver. That way, I can think "Connect to the DB with the dev credentials, not the live site credentials" instead of actually working out where all those things should be stored. Great.

In the future, so many of those hard problems will have been solved, the coders can address problems for real human being without having to discover Paxos consensus or red-black trees, and they'll just configure these generic solutions. They might have even specialized it to domains, like embedded devices or web coding. They'll be able to set up all the options they want, without hardly thinking about the way it actually works, and their application will look, and feel incredibly mature on day 4 of development. There will be some much information, they'll probably have to have documentation, and maybe they'll call it Ext-

So here I stand, in the future. A programmers utopia, where I am free to tackle whatever set of problems I want, equipped with the tools upon tools from 50 years of genius programmers, and I am unhappy with it.

I don't want to deal with those tools. I don't value 5 years of experience in NetBeans, and I don't see the fun in writing Javascript literal config objects, no matter how useful or eye-catching the end result is. Pushing somebody else's config values are can unquestionably produce something of value, but that's not the kind of value I want to create. I want to build something that you can build something better on, and in the meantime, it saves 500,000 from having to spend next Saturday afternoon from doing something they don't want to do. I want to understand why what we have is so hideously complicated that it takes a degree and 2 years of experience to send a Word document over something that isn't email to a friend, and I want to fix it.

I want to build and have tools that are so good I forget they're there. I want to stand on the shoulders of giants and forget how tall I am, because I know my tower can be miles taller.