The state of CSS, reconsidered
In Eric's year-end wrap-up of the 2020 series, the most salient takeaway is that the web is full of old technology — and that this is not just acceptable but good. Many voices independently raised the same point: old tech is dependable, predictable, and the majority of what we are actually doing.
- Lea's numbers on web technology usage across a wide slice of the internet support the "old tech" thesis.
- Kilian blamed the feeling of being behind on inflated estimates of how widely new tools are used.
- Shawn illustrated that "old" and "new" depend on when we started; we can't move our "Year Zero" but can try to see as a beginner.
- Manuel said keeping up isn't a game worth playing, but learning more about languages we already "know" holds surprises.
- Cassidy noticed old ideas resurfacing in new contexts — notably static file hosting as a fresh, smart approach.
- Jeremy had success building with raw HTML, CSS, and JavaScript, and his projects outlived the tooling cycle.
- Ire observed the way websites are built barely changed this year — and didn't have to.
Other threads
- Design and typography: Adam framed logical properties as layouts that adapt across written languages; Erik connected great design to great fonts; Miriam defended CSS art as a genre that can even have practical benefits.
- Process and mindset: Andy said every problem he faces is solved by simplification; Kitty urged readers to stop chasing hype; Brian learned to accept not knowing everything; Natalya argued this year is all wrong for creativity and productivity.
- Specific technique: Eric explained that indexing colors in a PNG with a scoped palette can save tremendous file size before any compression is applied.
Looking at the edge
The editor's own contribution for the year centers on the edge: global CDNs with more capability than caching static assets. Serving images from a nearby server is faster than from one 2,000 miles away, simply because of physics. That reasoning is now applied more broadly — a Jamstack site can serve literally everything from a global CDN — but servers are still needed for some things, such as logging in a user and pulling the items that user owns from a database.
A classic single-origin server handles that. Buying a server means paying for a physical machine in a physical location, like AWS's us-west-1. Serverless cloud functions such as AWS Lambda remove the server purchase, which the author calls cheap, fast, secure, and easy — yet, these functions also still run from a single physical location, which helps keep them inexpensive. Cloud functions are starting to run at the edge as the picture shifts.
As I write, Lambda@Edge is about 3× the cost of Lambdas in one particular region.
The author wants cloud functions at the edge because it is better for literally everyone where performance is concerned; the question is only whether the cost is affordable today. Given that costs trend downward while capability rises, the expectation is a world where all cloud functions run at the edge at all times.
Extend that reasoning further and the default becomes all-edge-all-the-time: static assets, compute, and data storage all at the edge. Physical infrastructure will always be required, but as it covers more of the world, the hope is that edge-first becomes the default way to build for the web.



