Let’s talk about how this website works.

Hugo

This site used to be regular hand-typed HTML with some Bootstrap, but when it was time to add a bloggy thing it seemed like it would be a bad idea to try to maintain that manually.

The most common tool for this seems to be Jekyll, but it seemed like a good opportunity to try something new. So this website is built with Hugo. Here’s some of the stuff I learned:

Jekyll has a lot of dependencies, which makes it difficult to get a new development environment set up. This is especially bad when “edit a website” is supposed to be one of the easy entry-point contributions for newcomers. It requires an explanation of gems, possibly a detour into version managers, and might include some Python troubleshooting. Hugo only requires brew install hugo.

Hugo and Jekyll have different opinions on how to build templates, which was confusing at first. Jekyll uses the idea of one main template that is extended by increasingly-specific templates. Hugo uses the idea of creating building blocks and then putting them together into specific templates. This seems to provide more flexibility in design, but the learning curve for template name-spacing is not insubstantial.

Good example projects are invaluable. The Hugo documentation is comprehensive, which means that your answer is in there but also means that you have to figure out where to find it. Being able to flip back and forth between the docs and an example or two helps a lot. And it’s more useful if the examples are of specific elements, rather than a showcase of all of the options.

Vanilla.js

There are tons of open-source tools available, and—as much as we like to complain about this or that drawback of them—they’re pretty awesome. With a little bit of Bootstrap, a website can look decent really quickly, and with jQuery it can be interactive, too.

I’ve definitely taken advantage of that, but it comes at the cost of not fully understanding the basic layer that those tools are abstracting. For many things, this is fine: the time and consistency benefits are worth it. This site is a good opportunity to experiment, though, so I’m trying to minimize dependencies and force myself to work around them. That doesn’t mean no dependencies: adding a blog would be very tedious without something like hugo behind the scenes and I have some presentations that benefit from using big. But it is a goal.

Deployment

Jekyll also has the benefit of being the built-in language of GitHub Pages. You can use another static site generator, but you have to build locally and push up the generated site. That’s fine, but then you don’t get to see the source, or you weirdly have two repositories for the same site.

Luckily, I was already using Mail-in-a-Box to host the static version of the site. It’s a really great way to host your own email (much better than my previous setup of a computer in my basement), and it comes with nice bonus features like static hosting and DNS.

The basic way to update the site would then be to build it locally and use sftp to put it on the server. But that is reliant on me being on a computer with the right server keys, which limits the speed of making changes. It also creates parallel processes for posting to GitHub and posting to the site, which could allow things to get out of sync.

One approach is to set up a tiny server that listens for webhooks from GitHub and then rebuilds the site. I’ve gone a more basic route, and just set up a cron job on the box to rebuild the site every minute. Hugo is supposed to be fast, after all. The script is basically this:

git pull https://github.com/stvnrlly/stvnrlly.com.git
cd stvnrlly.com
hugo -d /home/user-data/www/stvnrlly.com