Let’s run some fun scenarios to start this one off, shall we?
- “We’ll decide what language to use at a meeting next month. Until then, get to writing code!”
- “We might deploy internally, though Azure is a strong option, and we’ve discussed GCP. Now get operations up and running, we’ll tell you when we’ve decided.”
- “The code is changing constantly, get moving and write the documentation.”
All three of these scenarios are equivalent. The difference is how we, as an industry, view them. The first two, we (collectively) would never tolerate. The third? We do that to the documentation team every day. In DevOps and agile, we do it every day in a never-ending iteration. And it has to stop.
Two Different Documentations
For those not aware, there is documentation for internal use and documentation for external use. While we’re certain you have internal covered with self-documenting code, this blog will focus on external use, because if you expose and API, documentation is mandatory. If you’re not an end-user product, any UI requires documentation, too. So to be clear, for this blog we are talking generally about external documentation for B2B systems, and specifically about APIs, though UIs can have the same issues, depending upon how complex your market is.
We Generate Docs
Ever tried to use auto-generated documentation? I’ve yet to try one generated set of API documentation that I liked (they say Google Places has it, but my understanding is devs actually write that documentation and it’s reviewed by tech writers, so maybe we count that as one I liked, but its process is not yours).
Generated docs are only as good as the developers working on the code–very much like self-documenting code. If someone puts “This parameter is a placeholder for the spline development iteration” into a generatable comment, users are going to be totally clueless what that means. Heck, I just dreamed that statement up, and while we can all see it turning up in generated API docs, even I don’t know what it might mean.
Documentation Slows Up Dev
Duh? If you are loathe to commit to completing a project until you know the parameters, why would you expect a documentation team to be any different? Of course they need the final product for review, just as the business wants to see the final app before release. The product your docs team produces is detailed documentation about what is. If you change what is underneath them, they look bad for having inaccurate docs. When really it should be IT that looks bad for not considering the entire value chain when delivering. If you deliver an API that is poorly documented and people don’t use it, you have delivered nothing of value to the organization.
What to Do?
Put documentation on your DevOps teams. I mean an actual person whose responsibility is documenting. If you’ve got technical writers, use one of them. If you don’t, find someone who likes helping people use your toys, and give them time to do it. For some organizations, technical evangelists can fulfill this role if you don’t have them too busy helping active users. Have them sit in on meetings, treat them as just another piece of both agile and DevOps. They need to know when functionality, URI or whatever else changes, so they can update the docs. The best place to get that info is at stand-ups or design meetings.
Just like always, they’ll want to be there for design, but they’ll also want to be there for implementation decisions/changes. Because current documentation is the only worthwhile documentation. Give them sign off on release–not release of code (that’s a done deal) but publication of APIs. Before you let users at the access points, make sure they have solid instructions in how to use them.
Everything is a trade-off, but this one has some pretty easy choices to make–fewer people use your API because your documentation isn’t great (outright sucks in many cases), or you run some risks of slower iterations and you dedicate resources to documentation. Given that if no one uses your API, it is all wasted time, a bit of slow-down and man-hour investment is an easy choice to make.
Tech writers are generally cautious people who want perfection–because this is their product, and they want it right. This can cause unnecessary slow-downs in an agile/DevOps world. The easy answer here is to look for the same traits in documentation DevOps team members that you look for in other DevOps team members–ability to adapt, understanding that everyone is just trying to make it better and willingness to work through changes quickly to deliver.
Just like you don’t want the DBA that says “no” to every request and you have to argue with just to move the project forward–I worked with this guy, genius, but really frustrating to have on an agile team–you don’t want the documentation person that adds weeks to timelines so they can get it right. This risk is easy to manage. You know how to choose people for a fast-moving environment, just follow your gut.
Well, okay, maybe not profit, but deliver a solid product quickly with sufficient documentation that users not steeped in your corporate shared knowledge will understand and be able to use. That’s the goal of public-facing APIs, to be used. And it’s the goal of docs people to make certain those calling the APIs understand what they’re getting into and how to make the most of it.
Keep on cranking out cool solutions and crank out better docs. Your customers are counting on it. I’ve had my team go find alternative solutions based on shoddy API documentation. If I can’t use it, why would I keep trying if there are alternatives?