Web development best practices in 2021

What the data tells us

HolyJS, 2021-11-02

Who is this guy?

  • Currently: Data Evangelist at Netlify
  • Previously:
    • Co-founder at npm, Inc.
    • Co-founder at awe.sm
    • Web developer at Yahoo
  • Web developer since 1996

What you'll learn from this talk

1. The current state of web development

What You'll Learn

2. "Popular choice"

is not the same as "best choice"

What You'll Learn

3. A framework for choosing web technologies

What You'll Learn

How do I

know this stuff?

  • The Jamstack Community Survey
    • 10,000+ survey responses
      • 3000+ in 2020
      • 7500+ in 2021
  • npm downloads data
  • Redmonk survey data
  • Other surveys

What do we know?

  1. Who we are
    • Demographics
  2. What we're doing
    • App types
    • Device targets
    • Dev priorities
    • Audience sizes
  3. How we're doing it
    • Design tools
    • CMS choices
    • Editors
    • Programming languages
    • Web frameworks

Sources of error

  1. The Jamstack Community Survey
  2. Probably me being dumb sometimes

Back-end beware

Where else can I get this kind of data?

Data tells a story

But sometimes you have to explain what it is

Data always raises more questions

Section 1:

Who we are

82% engineers

Lots of students

Average experience: 6+ years

50%

50%

Increasingly geographically diverse

85% EU/NA

52% EU/NA

Developer

archaeology

85% EU/NA

52% EU/NA

Tech vs. non-tech

Industry

39% say "pure tech, no particular industry"

Huge shift

to remote work

40% remote

in devs with > 2 years of experience

Most likely remote: DevOps

What does this all mean?

Section 2:

What we're doing

Everybody works on their personal site

B2B and Consumer

are almost equally big categories

E-commerce is big

but lead capture by itself is enormous

Internal tools are big

Do we give them resources to match?

Application types

  • Fully static
  • Partially dynamic
  • Fully dynamic
  • Single page app

 

All sites, not just Jamstack sites

Fully static and fully dynamic are almost equally common

Target devices

Are we "mobile first"?

Mobile second

Are we paying enough attention to mobile devices?

or are thought leaders paying too much?

Data leads to questions:

Development priorities

Performance then uptime

So being down is okay if it's fast when it's up

Move fast and break things

is apparently the industry standard

Avoid vendor lock in

unless it's in any way inconvenient

Audience sizes

What affects performance most on a site with only 1000s of users?

What does performance look like for fully dynamic sites?

Section 3:

How we're doing it

Metrics

1: Awareness

bubble size

well known

not well known

2: Usage

horizontal axis

usage

2: Satisfaction

vertical axis

want to

use more

(happy)

want to

use less

(unhappy)

Year on year changes

happier

sadder

lost users

gained users

Technologies that gain users usually lose satisfaction

Late adopters

are grumpier

Figma leads design

Everybody else

Content Management Systems

Why pay so much attention?

WordPress leads traditional

Contentful leads headless

Webflow leads design-led

WordPress leads overall

CMS conclusions

  • Headless CMS is the most popular type
  • Contentful is the best bet in headless
  • Strapi and Sanity are worth considering
  • Webflow is the leader in design-led CMS

Everybody uses IDEs

Top 10 programming languages

for (mostly front-end) web developers

TypeScript is a big deal now

TypeScript rivals JavaScript

Hey look, SQL!

Unhappy languages

The new class

The wider world

Source: RedMonk Programming Language Rankings: June 2021

The JavaScript Trap

Developers writing for the web are forced into JavaScript

WebAssembly

aka WASM

WASM can

free the web

from JavaScript

Conclusions: languages

  • JavaScript is inescapable (for now)
  • Learning TypeScript is a good idea
    • Especially at a big company

Server-side techniques

Serverless functions: 46%

Serverless functions and fully dynamic sites

Web frameworks

are good, actually

The time before

web frameworks

It was very inefficient.

Ruby on Rails

Changed the development economy

Frameworks always solve a human problem

and only sometimes a technical one

Frameworks make you go faster

because you can get other people to help

Why not use a framework?

Performance is not a general argument

Staleness

The case against frameworks

Frameworks grow for 3-5 years

and then very, very slowly decline

Types of frameworks

  • Component libraries
  • "Kitchen sink" frameworks
  • Static site generators (SSGs)
  • Others

Component libraries

  • React
  • Vue
  • Svelte

"Kitchen sink" frameworks

Middle

  • Next
  • Nuxt
  • Nest
  • Vite
  • VuePress
  • Sapper

Newest

  • SvelteKit
  • Blitz.js
  • Remix
  • RedwoodJS

Older

  • Angular 1.x
  • Angular 2+
  • Meteor
  • Hapi
  • Ember

Static Site Generators

SSGs

  • Gatsby
  • 11ty
  • Hugo
  • Jekyll
  • Hexo
  • Gridsome

Special purpose frameworks

  • Docusaurus (documentation)
  • Stencil (web components)

On web components

Usage < 10% but growing, not well tracked in this data

Other frameworks

Legacy frameworks

  • jQuery
  • Backbone
  • Dojo
  • Express

Not tracked

  • Astro

React at 67% usage

Gatsby leads SSGs

but for how long?

Server-side vs. serverless

is a very blurry distinction these days

Conclusions: frameworks

No clear winner,

but some solid ideas.

A component library is a good idea

Conclusions: frameworks

A kitchen sink framework is not for everyone

But if you want one:

  • Next
  • Nuxt
  • Vite
  • SvelteKit

Conclusions: frameworks

SSGs may be a better option

if performance is your goal

  • Market leader: Gatsby
  • Satisfaction leader: 11ty

Conclusions: frameworks

Some kind of framework is a good idea

Conclusions: frameworks

What about

"No Code"?

Maybe some day.

React may be the new jQuery

but I mean that in a good way!

jQuery never died

It became part of the browser

Can React transcend?

If it keeps growing, maybe

Web Components

vs

React

How about Web Components + React?

Phew.

We are nearly done.

Big sites:

Audiences of Millions

Breakouts

Big sites have more specialized roles

Fewer "full stack", more "front", "back" end

Breakouts: Big Sites

Big sites have different priorities

Compliance rises in importance at scale

Breakouts: Big Sites

Big sites are

mobile first

Appealing to everyone where they are

Breakouts: Big Sites

Big sites prefer certain frameworks

  • React
  • Next.js

Breakouts: Big Sites

Students

This wasn't useful, just funny

Breakouts

Students have weird preferences

They're just all over the map

Breakouts: Students

Enterprise developers

They have a lot of money

Breakouts

Enterprises use heavier back-ends

Containers, orchestration, microservices, etc.

Breakouts: Enterprise developers

Enterprises don't care about phones

Mobile whenever

Breakouts: Enterprise developers

Enterprises use a lot of TypeScript

Help your team scale with types

Breakouts: Enterprise developers

E-commerce developers

E-commerce? Ecommerce? eCommerce? Whatever.

Breakouts

E-commerce uses lighter back-ends

Ideally, written and maintained by someone else

Breakouts: E-commerce developers

E-commerce cares most about phones

Mobile first, sometimes mobile only

Breakouts: E-commerce developers

E-commerce doesn't use TypeScript

Maybe just to mess with Enterprise devs?

Breakouts: E-commerce developers

What does

it all mean?

Demographics

  • Full stack is most common
  • Developers are getting more diverse geographically
  • 40% of developers have gone remote
    • even more in DevOps
    • but not if they were new

What we learned

What we do

  • Consumer and B2B are equally big
  • Lead capture and landing pages are huge
  • Internal tools are a big deal
    • And maybe we don't give them enough time

What we learned

Site types

  • Static and dynamic sites are equally common
  • We don't care about mobile as much as we might
    • Unless we work in e-commerce
  • We care about performance and uptime
    • Even if it's insecure and illegal

What we learned

Technology choices

  • Figma leads design tools
  • WordPress leads CMSes
    • Contentful is strong
    • Sanity and Strapi are rising

What we learned

Languages

  • JavaScript still runs the world
  • TypeScript is a big deal
  • WASM may be worth a look

What we learned

Component libraries

  • React is the tool of choice
  • Can web components play a role?

What we learned

SSGs

  • Gatsby is the favorite
  • 11ty is coming up

What we learned

Kitchen sink frameworks

So many! But maybe pick:

  • next.js
  • Nuxt
  • Vite
  • SvelteKit

What we learned

What should I do?

Data can be overwhelming.

Enjoy the web

The best way to reach me is Twitter:

@seldo

Web Development Best Practices: What The Data Tells Us

By seldo

Web Development Best Practices: What The Data Tells Us

  • 4,751