Worklog: June 2020

How I built The worklog from June 2020.

17 minute read Published: Daily Worklog

In June, I worked on the initial implementation of It’s supposed to be a website builder that lets visitors choose a template from a gallery of websites, copy the template to GitHub, and then deploy to a static hosting site such as Netlify or Vercel. There’s some programming required to copy a template to GitHub. In May, I explored using a Ruby function running on Vercel to access the GitHub API and duplicate template files. In early June I began building out the Ruby function that copies template files. should offer a choice of website templates from a “template gallery” (sort of like Envato’s ThemeForest). Initially I intended to build the template gallery using a static site generator (SSG). In mid-June, I looked at popular SSGs: Hugo, Zola, and Eleventy. I quickly became frustrated by the complexity of the static site generators. I began exploring web components, a newer technology that is not yet well known. Web component technology is exciting because it simplifies a great deal of web development. It seems like a great technology to showcase on It may mean that can offer not just simple websites, but sophisticated websites using web components without frameworks or build tools.

I spent the second half of June learning about web components. I couldn’t find a good tutorial for beginners, so I began writing my own tutorial “Build a Web Component” (published eventually as Create Your First Custom HTML Tag). From the start, I was planning to create tutorials to assist users. Writing about web components seems important, as the technology is not yet well known and it could make web development easier for beginners, with a path from simple sites to more sophisticated web applications.

By the end of June I was convinced that web components are the future of web development. It’s possible that we may begin building sophisticated websites without frameworks or build tools. As July arrived, I began building web components to use for a suite of sites. As I learn more about web components and write basic tutorials, I’ve postponed plans to build website templates for

June 1, 2020

I found an example of a gallery using Bulma CSS with modal cards. I also found a Bulma form builder. I created a gallery of templates (actually just one template for now). Clicking a template opens a modal that contains a form with fields for project name, website name, and description. Clicking a “Deploy Now” button submits an HTTP POST request to the Vercel function, which takes the form parameters and retrieves template files, modifies files, and saves the selected template to a GitHub repository. I’m excited to see this. It’s the full functionality (in skeleton form) that I need for Yax. With this feature, creating a website can be as simple as browsing a template gallery, selecting a template, and seeing a website deploy.

June 2, 2020

GitHub Authentication

I can programmatically create a website project in my own GitHub account. I started implementing GitHub OAuth Authentication so Yax users can create websites in their own Github accounts. OAuth authentication is difficult to implement. In my experience, it takes a few days of work, even for an experienced developer. Today I used Dark to prototype the GitHub authentication web flow so I’m familiar with each step.

June 3, 2020

GitHub Authentication

I want to create a one-step process for selecting a Yax website template, authenticating the user with GitHub, and saving the website project to the user’s GitHub account. The standard GitHub authentication workflow expects the user to authenticate on the GitHub site and then return to a web application using a supplied redirect URL. At first, it looked like any data from a form submission would be lost during the authentication flow. I noticed that the GitHub authentication API can accommodate a state parameter that is passed onward in the callback request. So I hacked together some JavaScript that captures the form data and transforms it into an encoded string that is appended to the authentication HTTP GET request as the state parameter. Seems to work as I prototype it in Dark.

June 4, 2020

GitHub Authentication

I can initiate the GitHub authentication by clicking a simple link or button on a web page. But the link doesn’t contain the state parameter I need to pass to the Yax API endpoint. I tried initiating the HTTP GET request from JavaScript using fetch with the state parameter in the query string. That doesn’t work because I get a CORS error. Thought about it some more and realized I don’t need to use fetch because I’m not trying to get back any data. Instead I can just use JavaScript, "_self") with a URL that contains the state parameter in the query string. Works nicely.

June 5, 2020

Yax API Endpoint

Until now I was just prototyping the interaction between the Yax template gallery and the Yax API endpoint using Dark. Today I switched the URL to use the Yax API endpoint running as a Ruby function on Vercel. With what I learned during prototyping I quickly updated the Vercel deploy.rb function to receive a form submission from the Yax template gallery, complete GitHub user authentication, create a repo and write a file to the repo.

I improved the user experience by adding a pageloader to the template gallery page (the alternative would be a progress bar or spinner icon). I used a Bulma extension which places a page loader overlay on the whole page after the user submits the template form.

June 6, 2020

Deploy a Website with Netlify or Vercel

For building a website with Yax, the user story should be: Choose a website template from a gallery, submit a form with a website name and description, save the modified template to the user’s GitHub account, and deploy the website to a hosting service such as Netlify or Vercel. Today I took a look at the deploy step. Ideally the new hosted website would appear in the user’s browser at the completion of the process. I don’t think that’s possible. Instead, the user will see the README page from the template, which will contain deploy buttons for Netlify and Vercel. Both Netlify and Vercel provide deploy buttons. I tried setting up a Netlify Deploy button in the template README. It’s not quite ideal as Netlify insists on creating a duplicate repo in the user’s GitHub account. If the project is named “my-yax-website,” the user has to give the duplicate template a new name such as “my-yax-website-netlify.” It adds friction to the deployment process but Netlify doesn’t have an optimal solution for my use case (where I want to deploy the repo that contains the deploy button). I also set up a deploy button in the README for Vercel. It works the same as the Netlify deploy button. There’s a little friction during deployment (and potential for user confusion) but it’s still pretty cool to choose a template from a gallery and see it deployed as a website in less than a minute.

June 7, 2020

Yax Manifest File

So far, the Ruby function on Vercel only retrieves two files, a README and an index.html file. The function needs to retrieve all the website files that are needed for the template, including multiple HTML files, CSS files, images, and more. I’m not sure how to get a list of all the files in a GitHub repository so for now I’ll just read a manifest file and retrieve a list of files from the manifest. I’ll use the YAML file format and name the file yax.yaml. I added a YAML parser to the Vercel Ruby function and wrote a recursive method to iterate over a hash that contains strings and nested hashes to build a file list. Now I can retrieve all the files listed in the manifest, modify each as needed, and save the files to the user’s specified GitHub repository.

June 8, 2020

Testing Website Deployment

Did some testing and resolved an issue I had with Vercel giving me a “bad credentials” error when I deployed the websites. Now everything works to duplicate and deploy Yax templates that contain multiple files.

June 9, 2020

I’ve prototyped selecting a single website template and deploying with Yax. Now I need to build one or more gallery pages with a curated collection of templates. Static site generators are typically used to build blogs but they can be used to build a repetitive pattern like a gallery as well. I’ve been using Hugo but I really don’t like it. It’s fast to generate websites but I don’t like the templating language (templating provided by the Go language). And the layers of overriding template files is confusing. It’s just too complex for my tastes (and my current need). I looked at Zola. It’s built in Rust and has a cleaner templating system and no tooling required (just download a binary) but it’s not very popular. I decided to try Eleventy (a JavaScript-based static site generator) because it’s very popular. It uses the popular Nunjucks (from Mozilla) for a template language and it can construct repetitive page elements from data files so it is suitable for a template gallery.

June 10, 2020

Trying Out Eleventy

I got Eleventy set up and working using the HTML files I already created for the site. It requires an ugly JavaScript configuration file full of curly braces. Plus it requires NPM as a build tool that creates a node-modules directory with 163MB of dependencies (otherwise known as cursed hellspawn). The next step is to use the Nunjucks template language to set up a base layout page pulling in a header and footer as “partials” or “includes.” This is such a basic function I really wish it was part of the HTML specification from the dawn of the web. We’ve been solving this problem (DRY or “don’t repeat yourself” HTML code) with Dreamweaver and PHP and fuck all since 1991. Looked at Nunjucks and it’s a hell of curly braces. Really, haven’t we solved this problem yet. I don’t want to offer Yax templates to do-it-yourself webmasters that require a build process (and brittle tooling) and contain HTML files polluted with weird template languages. I don’t want to inflict this on neophyte Yax users and I don’t want to inflict it on myself. I’ll look at using web components to DRY up my HTML.

June 11, 2020

Learning About Web Components

Started learning about web components. I like the idea that web components don’t break the declarative nature of HTML. Web components just add custom HTML tags you create yourself (or get from a library), loaded on a page using some JavaScript. Mostly I’ve heard they are useful for “design systems” which is something (in my experience) that only very senior frontend developers worry about. I thought about the reasons design systems are recommended (consistency, maintainability, reusability). Really, when you want to DRY up HTML for standardized layouts and consistency of elements along web pages, that’s a design system. So as soon as you want to reuse a footer across multiple webpages, that’s a design system. We’ve been struggling with a need for that since 1991.

June 12, 2020

Learning About Web Components

To learn about web components, I started with these Awesome lists:

Also, after I learned that lit-html is popular, I found great resources on the Awesome lit-html repository.

I liked this quote from Leon van Wijk:

The concept of Web Components promises a future where web applications are built without frameworks. Instead of frameworks, developers could stick to ‘vanilla’ JavaScript and use the Web API’s made available by the browser.

June 13, 2020

Learning About Web Components

I read about the history of web components. Significantly, on January 15, 2020 Microsoft released a version of its Edge browser that is based on Google’s Chromium, assuring web components are now supported by all major browsers including Google Chrome, Safari, and Firefox. I also read about All the Ways to Make a Web Component and learned (as of June 2020) that there are 33 different ways to make a web component.

June 14, 2020

Learning About Web Components

I started examining web component libraries and looking at sample code that implements web components. Some large corporations have released open source versions of the web components they use across their websites (AXA, ING, SAP, Pearson, Adobe, Salesforce). It seems most of the web component libraries are documented with Storybook which provides a UI for a collection of web components (but loads really slowly on my machine).

June 15, 2020

Learning About Web Components

I looked more closely at Google’s Lit-Html and Lit-Element (released February 5, 2019). It seems many of the newer web component libraries are based on Lit-Element. The Open-WC project has documentation and guidance for building web components with Lit-Element.

June 16, 2020

Learning About Web Components

I looked at how to use web components to implement a gallery. Lit-Html supports templating with iteration, so elements like a photo or card can repeat, generated from data. This is what I need for a template gallery for It’s great if I can do this without a static site generator. Just wondering if there’s better performance to use a static site generator or use web components that are populated from data.

June 17, 2020

Learning About Web Components

I looked for starter packages or generators to build web components. The Open-WC project has a scaffold generator that will kickstart a new web component. The documentation makes some assumptions and has some gaps but the project is informative. I also looked at an interesting project from Abraham Williams named Nutmeg to “Build, test, and publish vanilla Web Components” but the installation instructions led to errors.

June 18, 2020

Learning About Web Components

I looked for a beginner’s tutorial for building web components with lit-html but all the tutorials I found were aimed at experienced front-end developers, so I started writing my own tutorial “Build a Web Component.”

June 19, 2020

Writing a Tutorial About Building a Web Component

I wrote sections “Create the Project”, “Add Html”, and “Add JavaScript” for my “Build a Web Component” tutorial.

June 20, 2020

Writing a Tutorial About Building a Web Component

I wrote a section “Try a One Page Demo” for my “Build a Web Component” tutorial. It’s great to be able to show how to build a web component with just a text editor and a browser. No need to install build tools or a framework!

June 21, 2020

Updated and published the worklog for

June 22, 2020

Writing a Tutorial About Building a Web Component

I wrote a “View in a Browser” section for the “Build a Web Component” tutorial.

I saw Brian Haferkamp is working on a project to build a basic personal website with in-page editing via Mavo. That’s the kind of template I want to make available with

June 23, 2020

Setting up

Now that I’ve got a draft of a first tutorial, I want to set up a site at I want to find a nice template for tutorials that uses the Bulma CSS framework as a starting point for my own design. I looked for templates used for documentation. Turns out it is difficult to search for “Bulma documentation theme or template” because all the results are about “Bulma documentation.” Finally, I found documentation for the Fastify web framework that uses Bulma styling with Nunjucks templates. That’s a good beginning point.

June 24, 2020

Setting up

I started down the path of setting up a tutorials website using Eleventy as a static site generator with Nunjucks for templating and Bulma CSS styling using files borrowed from the Fastify website (Fastify uses the MetalSmith static site generator with Nunjucks and Bulma). Very quickly started to dislike the entire Nunjucks templating system with layouts, partials, front matter and so on. The Hugo static site generator is complicated but Eleventy and Nunjucks is almost as complicated. I just want to use HTML custom tags for a header and footer and have Markdown files pulled into a page.

June 25, 2020

Setting up

I decided to explore if I can use HTML custom tags to generate pages from Markdown files for a simple tutorial site without a build process or static site generator. I found a web component prism-markdown-element from German Martinez Solis that converts a Markdown file to HTML, applies syntax highlighting to any code blocks, and displays the content on a web page. Nice to see it can be done! That means I can load Markdown files into web pages for a tutorial site without a template system or a static site generator.

June 26, 2020

Setting up

The prism-markdown-element web component depends on NPM packages for Markdown conversion (commonmark) and syntax highlighting (prismjs). The author of the web component assumes developers use some JavaScript tooling to install the NPM packages. I don’t want to install and bundle Node modules, I want to use JavaScript directly from the browser without a build process. Really! I read about ES6 modules in a good book from Axel Rauschmayer. Then I found my Cape Town friend Guy Bedford (who is a netgod of JavaScript system internals) recently (June 19, 2020) set up the CDN at to deliver any NPM package in a form that can be used directly from a web browser. No JavaScript tooling required for bundling packages with Rollup or WebPack. I replaced filesystem imports with CDN imports in the prism-markdown-element. Very pleased to see it work (thank you, Guy Bedford!).

June 27, 2020

Repository for assets-yax-com

If I’m going to have a few web components (starting with yax-markdown, yax-header, yax-footer) I probably should recognize I’ve got the beginning of a web component library and create a separate GitHub repository to manage all the web assets that will be shared between the sites,, and the main site. So I created the assets-yax-com GitHub project. Haha now I have the beginning of my own design system library and I’m not even a very senior front end developer.

June 28, 2020


It’s been two months since I started building I really want to get to the point where I can get user feedback and find out if there’s a market for what I’m building. So far I’ve just been diving deeper into technology. And that’s the problem I want to address with It shouldn’t require two months of exploring the newest technologies for web development to build a website to test an idea. I need a website for tutorials. There should be a good template on that I can deploy and then upload tutorials written in Markdown. Anyway, I’ll get there. I know this is always a problem for tech entrepreneurs. We get caught up with solving technical problems and defer engaging with potential customers.

Sources of happiness

  • I’m thrilled about web components (specifically, HTML custom tags). I’ve really wanted this since the early days of the web. It’s been a long time coming. I’m also thrilled about JavaScript ES6 modules. That’s also been a long time coming; JavaScript has needed a build-in package manager since day one. Web components and ES6 modules enable modularity for HTML and JavaScript so we get separation of concerns, reusability, and maintainability right in the browser without frameworks or build systems. It’s a bright new dawn for web development in 2020.
  • I was thrilled to find out Guy Bedford released the CDN so I can import any NPM package directly in the browser without a build process.
  • I’ve been happy with the Bulma CSS framework. I’m implementing design and layout quickly without feeling like I’m fighting with CSS.
  • I like building API endpoints in Ruby and deploying them on Vercel. It’s sweet and easy.
  • I’m happy to have the GitHub API to create repositories and copy files. It’s capable of providing the key infrastructure I need for my project.
  • I like Mavo for its instant in-page editing and its extension of the declarative HTML language syntax.
  • I’m happy I’m living in Bali. I’m staying healthy and living well, with a great opportunity to focus on the project. I’m blessedly lucky that I had the opportunity to come here on March 19th.

Mixed feelings

  • I was intrigued with the Dark concept of an integrated editor, API query tool, and web server with its own advanced programming language. It was very useful for prototyping but a language bug kept me from going further.
  • OAuth authentication hasn’t gotten any easier but I managed to implement what I needed for the GitHub API.

Sources of unhappiness

  • I’m disappointed with static site generators (specifically Eleventy and Hugo) and templating languages (Nunjucks and Go templates). The SSGs are basic to the JAMstack approach and I was hoping they’d make web development easier. I’ll continue to use Hugo for this blog because I’ve got it figured out for this simple use case. But there’s just too much cognitive overhead to sort out the configuration maze to build a site for my template library. And with web components, I think it’s just not needed.
  • Looking at the world beyond Bali and my own little tech project, I’m saddened and dismayed to see the pandemic raging with no indication that it is subsiding.

Previous: Worklog for May 2020 - Next: Worklog for July 2020