layout: post title: Refactoring from Jekyll to Dat date: 2019-05-22 10:00:00 -0800
This is a live walkthrough of refactoring a Jekyll blog into a Dat site. Comparing our site on
dat:// is quite interesting. I'm currently encoding entries in markdown, which is natively supported by Beaker -- not so much by the major browsers. However, with the site hosted on Github, Hashbase, and my own computer, we will have a strong hand of backups - and mutations - to examine.
We initialize a new Dat in the already-existing git repository. My repositiory was setup as a Jekyll blog, so there's already a few files existing in the folder. We're going to hold onto some of the configuration files and the old
/_postsfolder, and build a new architecture around them.
Fig. 001: Directory after Dat init.
Creating a new Dat from Beaker Browser's main menu initializes the .datignore files. Compared to a commercial solution, we obviously have a lot of control here. For now, I intend to stay agile while exploring the ecosystem of Dat & Hashbase, and their border with
https://. So, I'm keeping the architecture will minimal while I focus on writing new content for the site.
Jekyll left behind a lot of
yamlfiles, so I delete most of them. Some of the data in
_config.ymlI transfer over to
dat.json. Json formatting is more common and standardized than Yaml, but the beautiful thing is I have the freedom to pick either. Unfortunately, editing a
dat.jsonin Beaker's UI deletes all data other than what their UI opens to you (e.g., Name & Description), as I took issue with on Beaker's tracker. For that reason, I highly recommend editing
dat.jsonlocally, and dealing with the lack of titles in Beaker. As for the information I'm including in
dat.json, I'm not developing for a specific API, but I have the information that all of them use. It would be pretty easy to interact with sockets on the
datecosystem that use common fields like title, url, email, twitter_username, and github_username, the handful I saved from jekyll's
_config.yaml, and I can easily add more.
Finished with the initial rollback of jekyll, I had to decide on a static site generator, or try rolling my own. After surveying a few options detailed in my "Survey of Four Static Site Generators" entry, I've decided to roll with
mdBook. The short size of the code base was a big factor, plus I'm pretty stoked on rustlang. Check out that entry to see how the Big Bads like Ghost and Jekyll competed with underdogs like Metalsmith and our winner, mdBook!
Before we continue our easy-reading experience with mdBook, I'd like to take a moment to remember this site's alien birth, when it couldn't decide which dimension to live in and cried out for stability and comfort.
- Fig. 002: The frankensite on
Fig. 003: Frankensite humanized on
Dat appears much more user-friendly, as Beaker Browser natively renders Markdown. I really wish that was true in major browsers like Firefox, Opera, Chrome, & Explorer - why do we keep all this html metadata around? - as well as on mobile browsers. Not supporting markdown is why I need to pick a static site generator, which is really more hassle than needed, as shown by Dat.
- However, it is very interesting to see their mutated attempts. The Hashbase
https://mirror only shows the file structure; users can't view files in-browser. Github Pages renders my
nav.mdfiles, which lack any yaml frontmatter - but 404's on my actual posts. Navigating to the raw posts on github.com, however, does show the yaml frontmatter - rendered as a table.
- I used a few different protocols here that were not designed with each other in mind. This feels as much like an art experiment as a devlog. Or a breeding lab, feeding radioactive data to different browsers and seeing which can adapt, survive, & thrive.
- Fig. 002: The frankensite on