When we set out to relaunch our company techblog, we had a few specific requirements:
- The blog should require no form of monitoring or on-call time.
- The blog must scale to meet (almost) any amount of traffic.
- Writing a post, submitting it for review, and getting it published should take as little time as possible away from the developer.
- All developers must be able to post, whenever they want to post.
- Developers must be familiar with the tools used to create the blog.
Of course, we could set up a dedicated Wordpress installation on one or several of our AWS-instances, and make it scalable with heavy caching and Varnish. However, this solution is prone to breakage, as we have seen with other Wordpress installations. Developers also have a tendency to not like WYSIWYG-editors, arguing that they create ugly and unreadable HTML. Not to mention that we would have to create users for every developer before they could start to blog, taking more time away from actual programming.
Then we discovered Jekyll. Jekyll is a simple, blog aware, static site generator. It is the main component behind GitHub Pages, a free hosting solution from GitHub. GitHub takes your Jekyll application (and static files too!), runs it through its compiler, and hosts the results. They even have a handy guide on how to create and publish your Jekyll application. Developers can fork, make pull requests, comment inline, use whatever editor they wanted, see diffs of content, and best of all, it needed little to no monitoring from our on-call engineer.
So with this in mind, I was tasked with creating the company blog using Jekyll, including a complete bootstrap stylesheet for Comoyo. I used Initializr to create the basic structure, using HTML5 Boilerplate, Bootstrap 2.0.4, Modernizr, jQuery and LESS. Initial development was slow at first, mainly because I didn’t know about the excellent bootstrap solution provided by Jekyll Bootstrap. After the initial folder structure is set up, the rest was pretty straight forward. Our folder structure now looked something like this:
. |-- _config.yml |-- _includes | |-- article_header.html | |-- author.html | |-- authors.html | `-- comments.html |-- _layouts | |-- default.html | `-- post.html |-- _posts | |-- 2012-06-12-how-comoyo-built-its-blog.md | `-- _1970-01-01-example.md |-- _site |-- assets | |-- css | |-- fonts | |-- img | |-- js | `-- less |-- .gitignore |-- 404.html |-- favicon.ico |-- Gemfile |-- humans.txt |-- README.md `-- index.html
With GitHub Pages, we no longer have a database, which would make it hard for ourselves to store comments, likes etc. However, as almost everyone else these days, we decided on using Disqus. The main advantage of outsourcing everything like this is that we don’t have to spend much money ourselves keeping the blog up, making it easier to justify spending developers time on writing posts, not spending time trying to keep everything backed up, monitored and live.
Of course, outsourcing has its drawbacks. Not controlling our own platform means that we risk unplanned downtime, data loss, sort of like setting sail in a ship who’s direction you no longer have control over. But since we use git, we will have multiple repos, thanks to the distributed nature of git. The risk of data loss is minimal. The only problem might be service outages at GitHub or Disqus, but then we would have bigger problems to deal with anyway.
Making Jekyll overcome its hurdles
Another issue was that Jekyll is by default a single-author blogging platform. We needed support for multiple authors, preferably with author profiles and metadata. After some quick googling, we found a solution. We also created two simple includes, authors.html which lists a short bio for each author, and author.html that shows the author profile when reading a post.
# _config.yml authors: dagingaa: name: Dag-Inge Aas position: Software Developer Intern gravatar: d4acca0bcb24dc67644055d4e44a6b29 web: http://www.daginge.com github: dagingaa twitter: daginge about: | Summer intern at Comoyo. Likes front-end development, new technology and cake. Currently active as Head of IT for [ISFiT](http://www.isfit.org).
We could then get the authors array through site.authors and iterate through this as one would expect.
Getting Jekyll ready for production
In short, we wanted our blog to scale to demand, use the latest web technology, be insanely fast, and become a digital playground were developers can create, write and modify everything the way they prefer. We believe that this is the best way to motivate developers to write for a company techblog.
Hi, we are Comoyo.comments powered by Disqus