Me: 2015 retrospective

My 2015 retrospective. Kind of like 2014, but less terrible and more snow.

I spent a good portion of 2015 ending projects:

  • elasticutils: I ended this in 2015.

  • Douglas: Douglas was a successor to PyBlosxom which I stopped using a few years ago. It fixed a bunch of issues and added a few nice features, but it was clear that I needed it to do some additional things like inter-blog-post linking and better handling of images and I just didn't feel like implementing those and maintaining them.

    Thus I moved my blog over to Nikola. I chose Nikola because I could trivially get it to keep all the urls I had on my blog already. Also, it was pretty easy to set up and had the features I needed.

    In switching to Nikola, I ended Douglas.

  • ernest: Ernest was a dashboard for Mozilla's Bugzilla that showed SUMO and Input sprints. I threw the initial bits together in a couple of days at a Mozilla Summit, then we kind of kept piling on things. It worked pretty well and met my needs more or less.

    Towards the end it had a lot of technical debt and then something changed either on Heroku or Bugzilla (I forget) that required a bunch of work, so instead I ended it.

  • Harold: I worked on Harold as an experiment combining the AAQ of SUMO with the Feedback section of Input in a way that didn't require accounts.

    That's about as far as I got. Since I never continued this experiment, I ended it.

  • django-godfrey: One of the problems I have with Django is that if you're using an API and expecting JSON back, but request an invalid url, then you get an HTTP 404 with HTML.

    I threw this together as an experiment, but after a while, I decided maybe this is better as a middleware. It's entirely possible this is solved already somewhere by now.

    Anyhow, this was a terrible idea so I nixed this project. I never did any releases, so it shouldn't affect anyone.

I took over as maintainer:

  • bleach: This was originally written and maintained for a long time by James Socol. After a chat in October or thereabouts, he decided it was a good idea to pass along maintenance to Jannis Leidel and myself and have it be owned by Mozilla.

    Since then, I've done some minor work in regards to defining the new security bug reporting process, documentation tweaks, and (slowly) working over existing issues and pull requests.

    Jannis and I are doing slow maintenance on this project, so expect releases for security issues, big bugs and then maybe twice a year.

I also started a couple of new projects:

  • pytest-wholenodeid: I wrote this pytest plugin to show the entire node id for failed tests so that it's easier to rerun the test after I've fixed it.

    It's not the most exciting library ever, but it helped me a lot.

  • Puente: I wanted to transition projects off of Tower, but the Django/Jinja2/django-jinja/Babel situation didn't cover all the requirements the projects I was working on had. I really didn't want to start a new library because I'm tired of maintaining things long-term, but I didn't see any other good options.

    Thus I created Puente to act as a stop gap for now to make it easier to transition off of Tower and identify the upstream changes that would need to happen. I spent a lot of time writing documentation to make it really clear how to transition off of Tower to Puente and also writing tests to make sure Puente correctly covered all the cases it needed to cover. Input and MDN transitioned without issues. SUMO had some issues. Marketplace had some issues, too. I think it was successful, but I'm sure there were things I could have done better.

    I got a lot of help from Rob Hudson and Mike Cooper. They pointed out a lot of issues and their comments and veiled threats fixed a lot of issues.

At work, I spent most of my time on fjord which powers Mozilla Input.

2015 was a lot less interesting than 2014---which is a good thing.

  • I passed off writing up Django security release bugs to the Aaron and the security team. This is great because they get the two-week heads-up and I don't. Plus they generalize this work and do the same with security releases from other frameworks, systems, etc. (Security)

  • I continued to hone Dennis and make it better. I regularly use Dennis to identify issues with msgids before I push them to Verbatim/Pontoon for translation. (L10N)

  • I spent a lot of time getting Input off of Tower. I put all the new code in Input, but then after talking to Rob about how he should do the same for MDN, I decided to turn it into a library with a migration guide. Thus was born Puente.

    My work on Dennis made it a lot easier to wrap my head around design decisions for Puente balancing translators' needs and work flow requirements with reducing the divergence of our l10n tools from normal Django projects.

    I wrote up several upstream bugs which I'll fix when I get a chance (or pass off to others--Rob Hudson already fixed one of them). That'll allow us to ditch Puente, too. (L10N)

  • I stopped doing the Input status reports every other week. As far as I could tell, they didn't have any impact on anyone. Given that, it seemed like wasted work. I did continue to do the quarter-in-review reports. (Project Management)

  • Through 2015, I wrote up project plans for all the major projects we tackled. I'm still on the fence about whether anyone noticed or cared much. Still, if anyone did, they're there. (Project Management)

  • We upgraded to Django 1.7 and then to Django 1.8. The plan is to stick with 1.8 because it's an LTS. In doing this, we dealt with a lot of technical debt and ditched a bunch of libraries. Input is a lot closer to a "normal Django project" now. (Software engineering)

  • We added a suggestions framework to the Thank You page. Users would leave feedback and Input would analyze it and offer suggested links.

    For example, in some cases, people are leaving feedback about a problem they're having that has a KB article on SUMO that addresses it. In these cases Input should guide the user to that KB article. In this way, we get the feedback that this is still a problem for some people and we also help the user solve their problem.

    There are some problems that only a handful of users experience that we're unable to reproduce. For these, we can create a triggered suggestion that looks for certain keywords and then if those show up in a user's feedback ask the user for some additional information. This makes it a lot easier for us to find users to work with to get the information we need to fix the problem. This comes up a lot with video cards and drivers on Windows.

    The suggestions framework allows for multiple suggesters which are essentially structured as plugins. Creating new ones is pretty easy to do. (Software engineering)

  • We switched to py.test. At first, I figured it wasn't a big deal. As time went on, I'm very happy I switched. I really like a lot of things about py.test and plan to use that going forward on everything.

    Tests on Input take about 30 seconds to run. I spend a decent amount of time making reducing bitrot in the tests and making sure each test tests something important. (Software engineering)

  • Wrote up a blog post for upgrading from Django 1.7 to Django 1.8 for Mozilla sites that were based on Playdoh or had Playdoh-like things. I've heard that blog post helped a bunch of other Mozilla sites. I hope so because it was a pain in the ass to write. (Communicating)

  • I tried out Atom for a couple of weeks, but I just couldn't get over the sluggish horizontal split and how much space it wasted. This is something I do in Emacs all the time, so it was really hard. I didn't want to go back to my configuration of Emacs, so I decided to follow Josh Mize's suggestion and try out Spacemacs and I've been using it since.

    It's a little goofy because one of the key features is vim keybindings (known as Evil mode) and I'm fairly proficient with vim, but I decided to use the Emacs keybindings (also known as Holy mode) instead. It's been nice.

  • Just before the year ended, I finally built a new desk. It's a little wobbly side to side which is easy to fix. I set up the monitor arms and so I've got a lot more horizontal space and hopefully don't have to play Towers of Hanoi with equipment and paperwork anymore.

Challenges for the year:

  • In February, it snowed every couple of days for 6 weeks. We got more snow than anywhere else in the USA and 3x more than we usually get in an entire winter season. It was insane. I had snowbanks on either side of my driveway taller than I am. Our 6 foot chainlink fence was completely buried. I gave up on one car and the section of driveway it was parked on--we didn't see it until Spring.

  • I still have too many things I'm working on, but it's a lot easier now than it was. In the next few weeks, I'll be putting down pyvideo, richard and steve. I spent some time working on them in 2015, but more importantly, I spent a ton of time feeling guilty and otherwise terrible about them. More on in another blog post.

  • Wearing a lot of hats on Fjord/Input is hard. Plus it feels like a constant battle to get anyone outside of User Advocacy and my manager to care.

    That struggle is at an end for me. I'm switching to MDN. On the one hand, I've followed MDN for a long time as a project since SUMO and MDN are sort of like siblings in a lot of ways so I'm a little wary. Having said that the team is awesome. I'm psyched to get a chance to work with them. I hope I'm helpful.

In 2016, I want to:

  • Find more time to tinker. I had like zero tinker time in 2014 or 2015--it was just too crazy. No tinker means I'm falling behind (or at least feeling like I'm falling behind).
  • Clean up my office. It's a mess. It really bugs me.
  • Find some friends. I have family and I have work and that's pretty much it. It'd be nice to do other things.

That's 2015 in a nutshell!

Want to comment? Send an email to willkg at bluesock dot org. Include the url for the blog entry in your comment so I have some context as to what you're talking about.