Why The World is Ready for Dexy

At one time or another, I suspect that most programmers and technical writers have attempted to "fix" technical writing in one way or another. It's a big problem space:

  • Everything, or at least many things, need to be documented, because undocumented features and behaviors cause problems that *one really ought not need to review the source code and understand the engineering to fix (potentially) trivial problems, every time the occur.
  • The people who write code are both not suited to the task of writing documentation because writing code and writing documentation are in fact different skills. Also, I think the division of labor makes some sense here.
  • Documentation, like code, requires maintenance, review, and ongoing quality control, as the technology and practice change. That's a lot of work and particularly for large projects, that can be a rather intensive task.
  • Lots of different kinds of documentation are needed, and depending on the specific needs of the user, a basic "unit of documentation," may need to be presented in a number of different ways. There are a number of ways to implement these various versions and iterations, but they all come with various levels of complexity and maintenance requirements.

The obvious thing to do, if you're a programmer, is to write some system that "solves technical writing." This can take the form of a tool for programmers that encourages them to write the documentation as the write the code, or it can take the form of a tool that enforces a great deal of structure for technical writing, to make it "easier" for writers and programmers to get good documentation. Basically "code your way out" of needing technical writers.

You can probably guess how I feel about this kind of approach.

There is definitely a space for tooling that can make the work of technical writing easier, as well as space for tools that make the presentation of documentation clearer and more valuable for users. Tools won't be able to make developers to write, at least not without a serious productivity hit, nor will tools decrease the need for useful documentation.

It's a difficult problem domain. While there is a lot of room for building programs that make it easier to write better documentation, the problem is that the temptation to write too much software is great. Often the problems in the technical writing process, including high barriers to entry, complicated build/publication systems, and difficult to master organizational methods, which are easy to address in programs. Meanwhile, most of these issues can be traced to overly complex build tools and human-centered problems, which are harder to address in code.

And since documentation takes the form of simple text, which seems easy to deal with, developers frustrated by documentation requirements, or technical writing teams, are prone to trying to write something to fix the apparent problem.

Which brings us to the present, where, if you want to write and publish documentation, your choices are:

  • Use a wiki, which isn't documentation but the software generally does a good job of publishing content, and wiki engines mostly don't have arcane structures of their own that might get in the technical writer's way. Downside: it's the wrong tool for the job and it forces writers and editors to maintain style themselves across an entire corpus, which is difficult and eventually counterproductive.
  • Use some other existing content management system. Typically these aren't meant for documentation, they have difficult to use interfaces, because they're meant to power websites and blogs, and they almost impose some sort of structure (like a blog,) which isn't ideal for conveying documentation.
  • Use an XML-based documentation tool-set. This is probably the best option around, at the moment, as these tools were built for the purpose of creating documentation. The main problems are: they're not particularly well suited for generating content for the web (which I think is essential these days) and as near as I can tell they make humans edit XML by hand which I think is always a bad idea.
  • Build your own system from the ground up. Remember text is easy to munge and most of the other options are undesirable. Downside: homegrown projects take a lot of time, they're always a bit more complex than anyone (except the technical writers?) expect, and it's easy to almost finish and that's bad because half-baked documentation systems are most of what get us into this problem in the first place.

So it's a thorny problem and one that lots of people have (and are!) trying to solve. I've been watching a tool called dexy for the last few weeks (months?) and I've been very interested in it's development and the impact that it, and similar tools, might have on my day-to-day work. This post seems to be the first in a series of thoughts about the tools that support technical writing and documentation.

comments powered by Disqus