CURRENT | Echoes

Infrastructure Inflections

When HTML Forked: The Compromise That Never Ended

By Nora Kaplan— December 3, 2025

Feature image for article: When HTML Forked: The Compromise That Never Ended

In February 1996, the HTML Editorial Review Board struck a deal: Netscape would remove blink tag from the specification if Microsoft removed marquee tag. Both companies agreed. Browsers kept supporting them anyway.

The negotiation revealed something structural. Between 1991 and 1996, HTML had forked into two incompatible paradigms—markup that describes what content is versus markup that controls how it looks. The compromise became permanent. Today, when you extract data across thousands of sites at scale, you're not just parsing HTML. You're interpreting two design philosophies that were never reconciled.

Infrastructure Inflections

When HTML Forked: The Compromise That Never Ended

In February 1996, the HTML Editorial Review Board struck a deal: Netscape would remove blink tag from the specification if Microsoft removed marquee tag. Both companies agreed. Browsers kept supporting them anyway.

The negotiation revealed something structural. Between 1991 and 1996, HTML had forked into two incompatible paradigms—markup that describes what content is versus markup that controls how it looks. The compromise became permanent. Today, when you extract data across thousands of sites at scale, you're not just parsing HTML. You're interpreting two design philosophies that were never reconciled.

Nora Kaplan

Nora Kaplan, former collaboration platform product leader turned technology writer. Studied human-computer interaction and spent years designing tools for knowledge work. Now writes about AI agents, work transformation, and how enterprise software reshapes human capability at TinyFish.

One Echo This Week

WebSocket Broke Every Assumption About Connections

December 2011: RFC 6455 standardized WebSocket. Developers celebrated real-time communication without polling hacks. Infrastructure teams got a problem that still hasn't resolved.

HTTP assumed connections die fast. Single request, single response, close the socket. Every layer of the network stack optimized for this: firewalls tracked brief sessions, load balancers expected quick turnover, proxies aggressively timed out idle connections.

WebSocket flipped it. Now servers hold thousands of connections open indefinitely. Firewalls don't recognize the traffic pattern. Proxies kill sessions they think are stalled. Memory pools sized for transient requests exhaust under persistent load.

Fourteen years later, HTTP polling remains more common than WebSocket across the top million sites. The infrastructure never caught up. When you choose between polling and persistent connections today, you're navigating consequences of that 2011 inversion.

One Echo This Week

WebSocket Broke Every Assumption About Connections

December 2011: RFC 6455 standardized WebSocket. Developers celebrated real-time communication without polling hacks. Infrastructure teams got a problem that still hasn't resolved.

HTTP assumed connections die fast. Single request, single response, close the socket. Every layer of the network stack optimized for this: firewalls tracked brief sessions, load balancers expected quick turnover, proxies aggressively timed out idle connections.

WebSocket flipped it. Now servers hold thousands of connections open indefinitely. Firewalls don't recognize the traffic pattern. Proxies kill sessions they think are stalled. Memory pools sized for transient requests exhaust under persistent load.

Fourteen years later, HTTP polling remains more common than WebSocket across the top million sites. The infrastructure never caught up. When you choose between polling and persistent connections today, you're navigating consequences of that 2011 inversion.

Connection assumptions:

HTTP 1.0 explicitly designed for minimal TCP lifespan, with entire network stack optimized around brief request-response cycles and rapid connection recycling.

Infrastructure mismatch:

Firewalls and proxies built for transient sessions struggle with persistent WebSocket connections, causing unexpected timeouts and socket exhaustion at scale.

Adoption stagnation:

Analysis of top million websites shows HTTP polling remains dominant, with WebSocket growth plateauing despite protocol maturity and developer preference.

Deployment lag:

Many production firewalls and proxies still lack proper WebSocket support, creating reliability gaps that force teams toward older polling patterns.

Pattern persistence:

Long-lived session protocols (WebSocket, SSE, HTTP/2 push) share operational challenges because short-connection assumptions remain embedded throughout network infrastructure.

Patterns Repeating Right Now

Infrastructure doesn't evolve randomly. It follows patterns that repeat across decades, visible in every technology transition if you know where to look.

The web you're building on today carries weight. Countless architectural decisions, each creating ripples that reach your production systems right now. Understanding these patterns changes what you build next.

These six patterns show up everywhere: in your Kubernetes clusters, your cloud bills, your hiring struggles, your architecture debates. They're not historical curiosities. They're the infrastructure reality you're navigating today, echoing decisions made long before your systems existed.

Patterns Repeating Right Now

Infrastructure doesn't evolve randomly. It follows patterns that repeat across decades, visible in every technology transition if you know where to look.

The web you're building on today carries weight. Countless architectural decisions, each creating ripples that reach your production systems right now. Understanding these patterns changes what you build next.

These six patterns show up everywhere: in your Kubernetes clusters, your cloud bills, your hiring struggles, your architecture debates. They're not historical curiosities. They're the infrastructure reality you're navigating today, echoing decisions made long before your systems existed.

Abstraction Layers

Complexity Always Gets Buried

Kubernetes hid container orchestration. Platform engineering teams now hide Kubernetes. By 2026, 80% of large organizations will run internal developer platforms hiding the platforms that hide the infrastructure. Watch any technology long enough: someone builds an abstraction layer. Then someone abstracts that abstraction. The cycle never stops.

Adoption Curves

The S-Curve Reveals Everything

Slow start, explosive middle, grinding plateau. Every technology follows this arc. Early gains feel effortless. Later improvements demand exponential effort for marginal returns. AI adoption in 2025 sits somewhere on that curve. The shape never lies about where you are in the cycle or what comes next.

Distributed Systems

Complexity Compounds Faster Than Solutions

Add three nodes, get nine connections. Deploy microservices, coordination overhead explodes geometrically. The distributed monolith exists because coupling persists despite architectural intentions. Infrastructure drift becomes inevitable. Management overhead grows faster than the systems being managed. Always has, always will.

Skills Gap

Infrastructure Outpaces Human Learning

Three point four million cybersecurity jobs sit unfilled globally. Hybrid clouds need expertise in both legacy and cloud systems simultaneously. Platform engineering emerged partly because skilled operators are scarce. Each technology wave amplifies the gap. Infrastructure evolves on Moore's Law. People don't.

Architecture Pendulum

Centralize, Decentralize, Repeat

Mainframes to PCs to cloud to edge. Gartner predicts 75% of enterprise data at the edge by 2025, up from 10% in 2018. Each swing solves yesterday's problems while creating tomorrow's. Edge computing is just the latest decentralization wave. The pendulum never stops moving.

Cost Optimization

Performance First, Bills Later

Build fast, discover costs afterward. Cloud spending hit $290 billion in 2023, growing 18% while organizations struggle to justify basic network upgrades. Infrastructure drifts toward poorly optimized states. The cycle repeats: optimize for performance, face cost pressure, re-architect, create new performance problems. Different decade, same pattern.

Papers That Built Infrastructure

CAP Theorem Proved the Impossible Triangle

Brewer's 2000 conjecture, proven by Gilbert and Lynch in 2002, revealed what distributed systems builders already suspected: you cannot simultaneously guarantee consistency, availability, and partition tolerance. The theorem didn't create this constraint. It formalized an impossibility that was always there, forcing every database architect since to choose which guarantee to sacrifice.

What did this change?

Every eventual consistency decision in modern architecture accepts this proven impossibility as fundamental constraint.

Where does this surface?

NoSQL databases exist because CAP forced architects to choose availability over immediate consistency at scale.

Fielding's Dissertation Documented the Web's Hidden Architecture

Roy Fielding's 2000 dissertation extracted the architectural principles already embedded in HTTP and the web. While co-authoring HTTP 1.1, he recognized that constraints like statelessness and uniform interfaces weren't limitations but the source of the web's scalability. REST described what was working, not what should work. Today's API debates still argue over patterns he documented.

What's the insight?

The web scaled because of architectural constraints, and Fielding showed us which constraints actually mattered.

Why does this echo?

Every RESTful API inherits decisions made for browsers, whether those decisions fit the problem or not.

MapReduce Made Distributed Processing Accessible

Dean and Ghemawat's 2004 paper revealed how Google processed petabytes by hiding distributed systems complexity behind two functions: map and reduce. The abstraction made parallel processing accessible to programmers who didn't need to understand distributed systems internals. Hadoop emerged two years later, carrying these patterns into every data pipeline since.

What's the breakthrough?

Google was running 100,000 MapReduce jobs daily when they published production lessons, not theory.

Where's the impact?

Hadoop adopted these patterns wholesale, shaping how an entire generation processes large-scale data today.

Bigtable Showed How to Store Petabytes

Chang's 2006 paper documented how Google stored petabytes across thousands of commodity servers using a sparse, distributed, persistent sorted map. Already running in production since 2005 with over 60 Google products depending on it, Bigtable proved you could build flexible storage that handled both batch processing and real-time serving. Every NoSQL database since borrows from this architecture.

What did this prove?

Flexible schema design could scale to petabytes while serving both analytics and production traffic simultaneously.

How does this echo?

HBase, Cassandra, and Cloud Bigtable all trace their lineage directly to patterns documented here.

Papers That Built Infrastructure

CAP Theorem Proved the Impossible Triangle

Brewer's 2000 conjecture, proven by Gilbert and Lynch in 2002, revealed what distributed systems builders already suspected: you cannot simultaneously guarantee consistency, availability, and partition tolerance. The theorem didn't create this constraint. It formalized an impossibility that was always there, forcing every database architect since to choose which guarantee to sacrifice.

What did this change?

Every eventual consistency decision in modern architecture accepts this proven impossibility as fundamental constraint.

Where does this surface?

NoSQL databases exist because CAP forced architects to choose availability over immediate consistency at scale.

Papers That Built Infrastructure

Fielding's Dissertation Documented the Web's Hidden Architecture

Roy Fielding's 2000 dissertation extracted the architectural principles already embedded in HTTP and the web. While co-authoring HTTP 1.1, he recognized that constraints like statelessness and uniform interfaces weren't limitations but the source of the web's scalability. REST described what was working, not what should work. Today's API debates still argue over patterns he documented.

What's the insight?

The web scaled because of architectural constraints, and Fielding showed us which constraints actually mattered.

Why does this echo?

Every RESTful API inherits decisions made for browsers, whether those decisions fit the problem or not.

Papers That Built Infrastructure

MapReduce Made Distributed Processing Accessible

Dean and Ghemawat's 2004 paper revealed how Google processed petabytes by hiding distributed systems complexity behind two functions: map and reduce. The abstraction made parallel processing accessible to programmers who didn't need to understand distributed systems internals. Hadoop emerged two years later, carrying these patterns into every data pipeline since.

What's the breakthrough?

Google was running 100,000 MapReduce jobs daily when they published production lessons, not theory.

Where's the impact?

Hadoop adopted these patterns wholesale, shaping how an entire generation processes large-scale data today.

Papers That Built Infrastructure

Bigtable Showed How to Store Petabytes

Chang's 2006 paper documented how Google stored petabytes across thousands of commodity servers using a sparse, distributed, persistent sorted map. Already running in production since 2005 with over 60 Google products depending on it, Bigtable proved you could build flexible storage that handled both batch processing and real-time serving. Every NoSQL database since borrows from this architecture.

What did this prove?

Flexible schema design could scale to petabytes while serving both analytics and production traffic simultaneously.

How does this echo?

HBase, Cassandra, and Cloud Bigtable all trace their lineage directly to patterns documented here.

Today's Debates Yesterday's Decisions

HTTP Archive's 2024 Web Almanac on Protocol Persistence

One in five pages still runs HTTP/1.1, twenty-seven years after its release. 1997's protocol decisions live in production today.

Deloitte Measures the Developer Time Tax from Technical Debt

Developers spend a third of their time managing architectural choices made before they arrived. Past decisions compound into present-day overhead.

Academic Research on Shared Databases Blocking Microservices Migrations