hackergotchi

treitter

Let's Push Things Forward

Maximizing social utility for fun and (modest) profit


Previous Entry Share Next Entry
hackergotchi
treitter

Answering the question: "How do I develop an app for GNOME?"

During the GNOME Developer Experience Hackfest this week, one of the major goals we identified was the need to pick a single language to give a simple answer to "how do I write a GNOME app?".

Right now, if you ask that question, you'll get about 8 different personal-preference answers, which drives people away from our platform. Having to potentially evaluate several different languages and their stacks gives potential developers a lot of unneeded extra work.

There was broad consensus in the hackfest for this goal because it allows us to:
  • It allows us to focus when we write developer documentation, fixing bugs in the development environment and the development of tools. This reduces our maintanence costs and enables us to be vastly more efficient.

  • It enables code and knowledge sharing to occur, so that people can easily copy and paste code from existing applications, or find information about common problems and challenges.

  • It provide a coherent and easy-to-follow path for new developers.

  • It allows us to include the full GNOME framework within the language itself.

We spent a long time discussing the different options that are available to us, and there were a variety of opinions. However, at the end of the day, we had to recognize that no language is perfect and there will always be disagreement. The important thing was that we had to make a decision.

It's critical that everyone understands this decision as a plan to elevate the language, bindings, tools, and documentation to a level of quality we have not yet achieved. It is not a decision to abandon any other language bindings. We will continue to distribute other bindings and documentation as we do now and compatibility for the other languages will continue to be developed as they are today by the developers involved with those modules.

Our decision is to support JavaScript as the first class language for GNOME application development. This means:
  • We will continue to write documentation for other languages, but we will also prioritize JavaScript when deciding what to work on.

  • We will encourage new applications be written in JavaScript.

  • We will be working to optimize the developer workflow around JavaScript.

C will remain the recommended language for system libraries.

Why JavaScript?
  • Our language of choice needs to be dynamic and high level.

  • There is already momentum in the GNOME Project for JavaScript -- it's used in GNOME Shell and GNOME Documents.

  • There's a lot of work going into the language to make it especially fast, embeddable, and framework-agnostic.

  • JavaScript is increasingly being seen as a first class desktop programming language -- it us being used in Windows 8, mobile platforms, and for local web applications.

  • JavaScript is self-contained -- it doesn't come with its own set of core libraries, which makes it more convenient for us when integrating it into the platform.

This is the start of a process and there's obviously a lot of work ahead of us. However, prioritizing a single language will enable us to turn GNOME into a compelling platform for application developers in a much more effective and efficient manner.

I'm interested in static analysis of programs to improve quality. Given that JavaScript is dynamic in just about every direction, and not at all amenable to static analysis, how can we expect to improve tooling so that programming errors are caught before release? (That was the reason I wanted to attend the DX hackfest.)

As an example, this is the kind of thing which I'm interested in: http://fedoraproject.org/wiki/Features/StaticAnalysisOfPythonRefcounts. Possible with Python (as an arbitrary example), probably not possible with JavaScript.

(Note: linting is not the same as static analysis. Linting is useful, but has limitations.)

Re: Discussion question

treitter

2013-02-03 03:52 pm (UTC)

Yeah, in the very least, I think we should make linting happen for JavaScript code in the same places we compile our compiled code to minimize risks of, eg, runtime compilation errors (the type that are caught earlier in static languages).

I'm not personally an experienced JavaScript developer (more C, Vala, Python, etc.), so maybe others can point out other areas we can bridge the gap between static and dynamic languages. I'd certainly welcome that.

Re: Discussion question

davemauldin

2013-02-04 08:14 am (UTC)

With jslint (or gjslint) and plenty of jsdoc comments, you can get what amounts to compile-time type checking.

I'm not sure if there's a simple way to do it without Closure, but since I use Closure most of the time, I find it very convenient to use Closure's compiler to find many easily overlooked issues for me, across many JavaScript files.

You are viewing treitter