PHP 8.2 strict in 2026: why it is still our floor for SMBs

Every time someone reviews Suite HUB internally, the first reaction is the same: “PHP? In 2026?”. It is understandable — the discourse on X and Hacker News does not do it many favors. But the decision was not by inertia or nostalgia. It was a calculated choice based on who our client is, where the system runs, and what lasts longer than the technical fashion of the year.

The context: $5/month hosting and cPanel#

Our typical client is a Panamanian SMB paying for shared Hostinger or a modest VPS. No Kubernetes, no Vercel, no Lambda edges. cPanel is still the operational reality of most accessible hosting in Latin America.

PHP runs out-of-the-box in any of those environments without touching anything. Node needs PM2, systemd, gateway, monitoring. Go or Rust need cross-compilation and binary deployment. For a client without a DevOps team, that is friction that kills the project before it starts.

PHP 8.2 strict is not the PHP of 2010#

People who see “PHP” mentally load version 5.x with mysql_query() and loose files. But PHP 8.2 with declare(strict_types=1) gives you:

<?php declare(strict_types=1);

namespace HubCore\Modules\Pos;

final readonly class CartItem
{
    public function __construct(
        public string $sku,
        public int $quantity,
        public Money $unitPrice,
    ) {}

    public function subtotal(): Money
    {
        return $this->unitPrice->multiply($this->quantity);
    }
}

final readonly class, strict return types, named args, enums, match expressions, first-class callable syntax, never return type, intersection types. Ergonomics are at the level of modern C# or Kotlin.

Composer is real, PHPUnit is real#

Hub Core gen 2 has 37 PHPUnit tests verifying 32 specific DGI rules. The PAC adapter breaks in CI before reaching production. Composer manages the dependency tree with composer.lock just as robust as package-lock.json or Cargo.lock.

The narrative that “PHP has no tooling” has not applied for a decade.

Frontend is not where decisions hurt#

For UI we use Tailwind 3 + Flowbite + Alpine.js. Alpine is the “Swiss army knife” of declarative frontend when you do not need React. In HUB Pro we add HTMX 2 for richer server-driven interactions. This covers 95% of cases without the complexity of an SPA.

Sounds minimalist? It is the right decision when your app lives embedded in a cheap ARM tablet browser with unstable 4G connection. Every KB counts.

The day we do pivot#

There are cases where PHP is not the right tool:

• Real-time intensive computation: we prefer Go or Rust as sidecar.
• Persistent WebSockets at scale: Soketi in Node runs parallel.
• AI/embedding services: Python with FastAPI lives in its own container.

PHP is the transactional heart. Specialized workers are satellites called via HTTP. That architecture distributes complexity without forcing us into a Node monolith that later no one knows how to operate.

The real advantage: finding talent#

In Panama there is a huge base of PHP developers with real experience, many with WordPress / Laravel background. Recruiting to extend Suite HUB is trivial. Recruiting to extend a Rust + GraphQL + microservices stack is practically impossible locally.

That talent availability is what makes the stack economically sustainable for our scale.

Conclusion#

PHP 8.2 strict is not the aspirational option or the one that earns clout on social media. It is the right option for our context: Panamanian SMBs, accessible hosting, small disciplined teams, 10-year maintainability. Technical modernity matters, but it matters more to solve the client’s problem with the tool that survives the hype.

And honestly: seeing a well-written PHP system serving 2,300+ real electronic invoices per month from a $40/month VPS, without dropping once, is an argument that rarely any nostalgia can match.