This is a journal of my thoughts—for my own reference and to have something to play with—while figuring out how to create this static website on GitHub Pages. Step one is to create a repository with the right name. For me, that’s

Rambling about Frameworks

I think I should use Jekyll. It’s suggested at I’m confused whether this is the only option with GitHub Pages or if there are other frameworks that can be used. A Google search for “blog github pages” mostly results in matches about Jekyll, but also something about Pelican and Hugo. I suppose options other than Jekyll will require committing the generated HTML pages in addition to the files used to generate them… I’ll just use Jekyll.

Installing Jekyll

The ArchWiki suggests gem install jekyll. There’s also a gem called github-pages that, I think, gives us whatever version of Jekyll is used on GitHub’s side. Sounds good. I’ll use that version.

The README of the github-pages gem says we need Bundler. The “quick-start guide” from the official Jekyll docs says so too. The docs’ installation page and the ArchWiki don’t. Huh. I guess I’ll install it too.

$ gem install bundler
$ gem install github-pages

Creating a Jekyll Site

Back to the quick-start guide. The next step is to create the basic scaffolding of a Jekyll site using jekyll new.

$ cd
$ jekyll new .
Conflict: /home/meribold/ exists and is not empty.
$ jekyll new . --force
Your user account isn't allowed to install to the system RubyGems.
You can cancel this installation and run:

    bundle install --path vendor/bundle

to install the gems into ./vendor/bundle/, or you can enter your password
and install the bundled gems to RubyGems using sudo.

Password: ^C

Bail. Installing stuff outside of $HOME with something other than the system package manager is a mistake I made with pip.1

$ bundle install --path vendor/bundle


$ jekyll new . --force

Errors… let’s start over. I removed everything created by jekyll new and bundle install.

$ jekyll new . --force --skip-bundle
New jekyll site installed in /home/meribold/
Bundle install skipped.
$ bundle exec jekyll serve

Looks promising but http://localhost:4000 only shows my index.html from my initial commit (28919f5). Delete it (rm index.html).

That’s better: some kind of demo page (1327f24). Let’s add this Markdown file as described in _posts/2017-09-02-welcome-to-jekyll.markdown (4f8937c):

$ mv _posts/$(date -I)/

When I serve the site locally, the new post shows up but isn’t themed. On GitHub it somehow is themed. Anyway, it probably needs some front matter: 34c1623. That does it.

At this point I looked at and edited all those other files jekyll new created.

Fork the Theme

I want to include my theme in my repository so I can make major changes easily. This is covered here. It’s also possible to use a gem-based theme and selectively override defaults while still getting updates to the theme used.

The goal of gem-based themes is to allow you to get all the benefits of a robust, continually updated theme without having all the theme’s files getting in your way and over-complicating what might be your primary focus: creating content.
Tom Johnson

I don’t want my theme to randomly change, though. (And probably break my overrides in the process.)

$ bundle show minima
$ cp -r /home/meribold/.gem/ruby/2.4.0/gems/minima-2.1.1/* .
$ git checkout # oops, restory my readme file
$ git add -A
$ git commit -m 'Fork the theme'

See commit f978760.


I removed almost everything from the Minima theme (de12477) to figure things out from the ground up. By the time you read this I hopefully created custom styling that doesn’t suck.


Or rather, “being noticed by search engines at all”: Google only has two results for “” (the site currently has 10 pages). Following the advice of whatever ranks highly when searching for “jekyll seo” seems logical.

The jekyll-seo-tag plugin will apparently automate adding a bunch of metadata tags that will help (a6d1930). I’m also using the jekyll-sitemap gem now (435a157) to generate this sitemap.xml2 and submitted it to Google. Sitemaps “inform search engines about URLs on a website that are available for crawling”.

A day later (09-13), this hasn’t worked out. There is currently one Google result and the snippet’s title isn’t updated. I added some semi-made-up modification date metadata to most pages (d1404b2) that jekyll-sitemap will use for <lastmod> tags (most pages were lacking this tag completely, though it was present when building locally). Then I submitted the Sitemap to Google again (this is probably unnecessary?) and also requested recrawling of my site.

Another day later (09-14), there is still one result with an outdated title. Yet another day later (09-15), there now are no results… here’s a table:

Date Results Outdated Titles
09-13 1 1
09-14 1 1
09-15 0 0
09-16 0 0
09-17 0 0
09-18 2 1
09-19 3 1

At some point I noticed that and didn’t give the same results, so I started checking both of them.

Date Results on .com TLD Outdated Titles Results on .de TLD Outdated Titles
09-20 4 1 4 2
09-21 5 2 5 2
09-22 4 1 5 2
09-23 3 1 3 1
09-24 3 1 3 1
09-25 3 1 3 1

It looks like the results are stabilizing but that’s just chance. To the contrary, they change way more frequently than I noticed before. Here’s a table with the number of results at various times (given in UTC) of the same day (2017-09-25):

Time Results on .com TLD Outdated Titles Results on .de TLD Outdated Titles
10:04 5 2 5 2
13:31 6 2 5 2
14:07 6 2 6 2
15:50 5 2 6 2
18:44 5 2 5 2
19:35 3 1 3 1

Some observations:

  • On 22., at least one more page could be found by adding keywords. Generally, including words from page titles in searches gives new results sometimes.
  • Maybe my description meta tags shouldn’t repeat the descriptions that are already on the projects page.
  • What pages have outdated titles didn’t change, so I can probably try updating the timestamps (<lastmod> tags) of those: /projects/ and the current one.
  • I think all pages were indexed on the 20. That’s what Google Search Console says.3


This section is a gallery of tricky and not-so-tricky content that tests how the CSS copes.

This was an h2 heading.


This was an h3 heading.


This was an h4 heading. It looks like h3 and I’m trying not to use it on this site.

The Way

This was an h5 heading. It looks like h4 and I’m trying not to use it on this site.


This was an h6 heading. It looks like h5 and I’m trying not to use it on this site.

Code Snippets

The following code block has quite long lines. It may extend to the right of where the main text is wrapped if it needs the space and the viewport is wide enough. If you’re using a phone it probably just has a horizontal scrollbar.

// This is a code block with quite long lines.  It tests how the CSS copes with something like that.

int sometimes_people_use_variable_names_which_are_much_too_long = of_course_this_can_be_somewhat_annoying();

int main() {
    printf("Hello, world!\n");

Here’s another code block. This one has a pretty ridiculous amount of characters per line; it may extend to the very right of the viewport and still have a horizontal scrollbar.

// In actual practice some hard line breaks should probably be added to a code block with as many characters per line as this one.

int sometimes_people_use_variable_names_which_are_much_too_long = of_course_this_can_be_somewhat_annoying() + especially_when_all_names_follow_the_same_pattern();
  1. In my defense, pip does make it pretty easy to shoot yourself in the foot. 

  2. It also generates robots.txt

  3. It says /sitemap.xml was submitted on 2017-09-12, processed on 2017-09-20, and 10 out of 11 links were indexed. I assume this is the one that wasn’t.