Latest changes and plans for this site

Jul 17, 2019

The website functions, but its evolution will never stop.

I ran three tests (of two types) regarding the site:

  1. Official Lighthouse, results: 100 / 97 / 85 / 100
  2. webpagetest.org Lighthouse, results: 67 / 97 / 93 / 100
  3. webhint.io scanner, results: 66 hints

Obviously, I’ll aim to achieve scores of 100 across the board for Lighthouse and 0 hints for the webhint.io scanner in the future. The following roadmap materialized in my mind:

  1. I’ll change the underlying framework. Although I upgraded to Bootstrap 4.3.1 + jQuery 3.3.1, the latter contains a medium security vulnerability. My website is static, so it’s not really an issue, but a newer / better framework could help with better scores, and those achievements should translate to real-world enhancements. In the post a week ago, I mentioned two Material links and ZURB Foundation. Turns out that some Material themes are for React, which I won’t use (just like Vue or similar frameworks), since I try to be minimalist and this is a static website after all. I’m pretty sure now that I’ll go with the Material Design Lite framework (MDL). The only feature for which I don’t see a clear equivalent in MDL is the carousel. I’ll either get rid of the carousel, develop my own based on the old 2018 summer version of android.com (see GitHub issue), or somehow pull in the feature from mdl-ext.
  2. When changing frameworks, I’ll try to work with the SCSS version so I can get rid of any unused fluff at an early stage. I’ll possibly take the same approach with the JavaScript assets, keeping them as minimal as possible.
  3. I’ll definitely need to do some image optimizations. That will possibly involve generating WebP images or applying heavy optimization, and possibly generating mobile versions of the large carousel images.
  4. I’ll remove unused CSS at the end of the build with a Gulp step, or possibly with a Jekyll Ruby gem.
  5. As far as performance goes, Bootstrap and MDL both use some fonts from fonts.googleapis.com. For a quicker first paint event, it’s advisable to use the font-display CSS descriptor of the @font-face rule. However, these Google fonts are pulled in by a CSS @import directive, and Google generates the @font-face rules. Unfortunately, they don’t place any font-display descriptor there; see more about this problem on CSS-Tricks. Right now, I have just scraped the CSS rules from the file served by Google, but that has some hashes, so it may not work forever. For a better solution, I’ll try to use one of the service worker solutions; they are kind of hacky, but we’ll see.
  6. Finally, I’m considering the use of a third-party blog post CMS / API service, like Siteleaf, but only if I find a satisfactory free service. Note to self: see the list at headlesscms.org.
Comments loading...