Angular 2, Vue 2, and a Durability Rubric for the Framework Decade

September delivered a natural experiment in framework economics, and we have been waiting for it.

On September 14, Google shipped Angular 2. A ground-up rewrite of the framework whose 2013 rise our ‘JavaScript Ate the Page’ piece chronicled, sharing a name and a steward with its predecessor and almost nothing else. Organizations running large Angular 1 sites are describing the migration path, charitably, as a second implementation.

On September 30, Vue.js shipped its own version 2. An independent project that began as a single developer’s synthesis of the era’s good ideas, now matured into a serious ecosystem, delivering its major version with a migration guide measured in an afternoon.

Same month, same species of event, opposite economics. Together they crystallize the question that has migrated from our engineering standups into every technology-selection engagement this year: how does an organization choose technology it will have to live with? The house has a rubric. Clients pay for its application. The rubric itself should be public, because the vocabulary matters more than the verdicts.

The rubric, five questions asked in order

• 1. What is the migration history? The single most predictive question, and September is why. A steward’s past behavior at major-version boundaries is the best available evidence of its future behavior. Projects with a record of incremental, guided migrations, the WordPress compatibility ethic our platform coverage has long credited, will probably do it again. Projects that have already rewritten once have demonstrated the appetite. Angular 2 is not disqualified by being a rewrite. The rewrite fixed real architectural debts. But the rewrite is now data, and organizations adopting it should price the possibility of being rewritten at again.

• 2. Who governs, and what do they need? The Reader lesson, applied to code. A framework is free the way Reader was free, and the steward’s incentives are the terms of service nobody reads. Corporate stewardship, Angular’s Google, React’s Facebook, means the framework serves the steward’s internal needs first. That cuts both ways: enormous investment, and a roadmap that answers to someone else’s product calendar. Community and single-maintainer governance, Vue’s model, funded by its own users, reverses the risks. Alignment with users, fragility of the bus-factor kind. Neither is disqualifying. Both are terms to read.

• 3. What does leaving cost? The adopt-with-an-exit rule, now a permanent fixture of any sound evaluation. The framework’s real price is the cost of the day you stop using it. Concretely: how much of your code is framework-idiomatic versus portable logic? Does the framework wrap your data and business rules, or tangle them? The architecture patterns that keep the answer favorable, thin framework layers over framework-agnostic cores, cost control now and buy freedom later, whichever box you choose.

• 4. Can you hire for it, and for how long? The ecosystem-employment test. A framework is also a labor market, and the mid-market institutions we serve cannot pay scarcity premiums for exotic skills. React’s market is now the deepest. Angular’s enterprise base sustains a durable second. Vue’s is growing from the agencies and shops where pragmatism lives. The question is not today’s depth but the five-year slope, which the governance and migration answers largely predict.

• 5. Will it be boring by 2019? The house’s summary test, stated the way we say it internally. The technologies worth institutional adoption are the ones on a visible path toward boringness. Stable, documented, unsurprising, absorbed. Per the absorption pattern that runs from responsive design to HTTPS. Excitement is a cost. The correct amount of it, in an institutional stack, is nearly zero.

The current field, sorted honestly

Applying the rubric aloud, with the reminder that the sort is dated the day it publishes and the rubric is not.

React scores strongest for institutional adoption today. Incremental version history and the deepest labor market. An architecture (view library, not lifestyle) that keeps question three’s answer favorable. Its governance is corporate with the attendant terms, and this month’s licensing grumbles in its community are exactly the kind of terms-reading the rubric prescribes. Watch that file.

Angular 2 is a genuine enterprise framework whose rewrite reset its history clock. Right for organizations already deep in Google’s enterprise gravity or wanting the everything-included frame. Priced with question one’s honest answer.

Vue 2 is the rubric’s interesting case. Superb migration manners. User-aligned governance with bus-factor risk. A gentler learning curve our teams confirm. A labor market one should check again each year. We now field it comfortably for appropriately scoped work.

And the standing reminder our JavaScript-ate-the-page piece installed: the option of no framework remains on the menu. For the content sites that are most of the institutional web, it remains the house default. The rubric’s first output, always, is whether the box is worth opening at all.

Running the rubric in the room

A process note from the engagements where this rubric earns its keep, because the instrument matters less than the administration.

Run it as a scored worksheet completed independently by the people who will live with the answer. The engineers who will build. The managers who will hire. The stakeholders who will fund the eventual exit. Before any group discussion.

The rubric’s diagnostic power is in the disagreements. When engineering scores exit cost favorably and management scores the labor market poorly, that gap is the real conversation. It never surfaces in the demo-driven format where the loudest enthusiasm sets the agenda.

Require written answers to question one, migration history, with citations to actual release notes. It is the question most often answered from impression rather than history, and the history frequently surprises.

And put the no-framework option through the same worksheet as a genuine candidate rather than a strawman, scored on the same five questions. It wins more often than the room expects. When it loses, it loses for documented reasons that become the project’s justification file.

The output is not the score. The output is an organization that can explain its choice in five sentences to the auditor, the new hire, and the steering committee of 2020. That is the artifact the excitement-driven selections never produce.

The decade question

The one-line version, for the steering committees who will read this month’s release coverage as a prompt to relitigate their stacks. You are not choosing the best framework. You are choosing the best decade, and the two questions differ everywhere it counts.

The best framework is a benchmark argument. The best decade is a bet on migration manners, governance incentives, and exit costs. On labor markets, and the path toward boringness. September’s paired releases just handed the industry a clean controlled study in why those, and not the feature matrix, are the decisive variables.

File the rubric next to the CMS-selection criteria it descends from. Reapply it annually. Distrust any recommendation, ours included, that arrives without its reasoning shown. The frameworks will keep shipping majors. The rubric keeps its shape. That asymmetry is the point.