[Paper] The Case for HTML First Web Development

Source: arXiv - 2602.17193v1

Overview

Juho Vepsäläinen’s paper makes the case for “HTML First” web development – a disciplined approach that treats HTML as the primary building block of web apps, only reaching for JavaScript frameworks when the markup can’t do the job alone. By leaning on modern HTML5, server‑side rendering, and hypermedia principles, the author shows that developers can shrink codebases, boost performance, and simplify maintenance without sacrificing interactivity.

Key Contributions

• Definition of “HTML First” – a concrete set of design guidelines that prioritize native HTML semantics, server‑driven logic, and minimal client‑side scripting.
• Empirical evidence – quantitative comparisons (code‑size, load time, and memory usage) between traditional SPA stacks and HTML‑first implementations, including a real‑world case study on the Yle news portal.
• Integration with htmx – demonstration of how the lightweight htmx library can bridge the gap between pure HTML and dynamic UI needs, effectively replacing large JavaScript frameworks in many scenarios.
• Performance and maintainability analysis – holotype‑based benchmarks that reveal up to 45 % reduction in JavaScript payload and 30 % faster Time‑to‑First‑Byte for content‑heavy sites.
• Discussion of trade‑offs – exploration of where HTML‑first shines (content‑driven sites, form‑heavy applications) and where richer client‑side state may still be required.

Methodology

1. Literature & standards review – the paper surveys the evolution of HTML (from 1991 to HTML5) and the rise of SPA frameworks, establishing the technical backdrop.
2. Design of “HTML First” principles – a checklist (semantic markup first, server‑side routing, progressive enhancement, minimal JS, hypermedia‑driven navigation).
3. Prototype implementations – two parallel versions of the same functionality were built:
   • A conventional React‑based SPA.
   • An HTML‑first version using server‑rendered templates + htmx for incremental updates.
4. Holotype benchmarking – identical content‑driven pages were measured for bundle size, network payload, rendering time, and CPU/memory usage across browsers.
5. Case study conversion – the public Yle news site was refactored to follow HTML‑first guidelines; metrics were collected before and after the migration.
6. Qualitative developer feedback – a small group of front‑end engineers reported on code readability, onboarding speed, and bug‑fix turnaround.

Results & Findings

• Codebase shrinkage: The HTML‑first prototype required roughly one‑third the amount of front‑end source code.
• Performance boost: Reduced JavaScript payload translated into faster network transfer and lower main‑thread work, especially on low‑end devices.
• Maintainability gains: Developers reported quicker onboarding (average 2 days vs. 5 days) and fewer bugs related to state synchronization.
• Yle case study: After migration, page load times fell by an average of 0.4 s, and server‑side caching became more effective due to the increased reliance on static markup.

Practical Implications

• Smaller, faster front‑ends – Teams can ship leaner assets, which is a direct win for SEO, mobile users, and cost‑effective CDNs.
• Simplified stack – By moving much of the interaction logic to the server, you can reduce the number of build tools, polyfills, and runtime dependencies.
• Easier hiring & onboarding – New developers only need solid HTML/CSS fundamentals plus a light library like htmx, rather than deep expertise in a specific SPA framework.
• Better accessibility & SEO – Server‑rendered, semantic HTML is inherently more crawlable and screen‑reader friendly.
• Incremental adoption – Existing projects can adopt HTML‑first gradually—replace isolated widgets with htmx‑driven fragments without a full rewrite.
• Alignment with emerging trends – The approach dovetails with AI‑assisted coding tools that excel at generating template‑based markup, potentially accelerating development further.

Limitations & Future Work

• Scope of applicability – Highly interactive, real‑time dashboards (e.g., collaborative editors) may still benefit from full‑blown SPA frameworks.
• Tooling ecosystem – While htmx is lightweight, the broader ecosystem (debuggers, testing frameworks) is less mature than that of React/Vue, which could hinder large‑scale adoption.
• Quantitative generalization – Benchmarks were performed on a limited set of content‑heavy sites; more diverse workloads (e‑commerce, gaming) need evaluation.
• AI‑driven development – The author flags an open question: how HTML‑first practices intersect with AI code generators that currently favor component‑based outputs.
• Future research directions – Larger‑scale field studies, automated migration tools, and deeper integration with server‑side frameworks (e.g., Django, Rails, ASP.NET) are proposed to validate and extend the findings.

Authors

• Juho Vepsäläinen

Paper Information

• arXiv ID: 2602.17193v1
• Categories: cs.SE
• Published: February 19, 2026
• PDF: Download PDF