I started with Perl in ‘96 and have lived through what feels like everything — PHP, jQuery, Drupal, Backbone, Node, Angular, ClojureScript, React, GraphQL, and NextJS. Htmx feels like a divergence from the trend, and is worth thinking about.

Htmx asks us a good question: “does the complexity of your work reside essentially on the server or essentially on the client?”

The complexity for the vast majority of websites resides essentially on the server. Most of us are not building Figmas and Google Sheets. Most websites, even if heavily interactive, are just CRUD apps with pleasant interfaces. Frameworks like NextJS attempt to rectify the problem of overly complex clients by moving React to the server, but this often magnifies complexity rather than minimizing it.

Wouldn’t it make sense to remove React from the stack? For complex clients, skip the DOM and JS with canvas and compiled web assembly. For complex servers, use some form of server-driven granular updates to the DOM.

The problem I see with this approach is that although most complexity reside on the server for most websites, there are almost always a few high-complexity task that needs to reside on the client — image editing, real-time sorting/filtering/calculation, drag-touch gestures, etc.

A hybrid approach is necessary. It isn’t good enough to allow compatibility. Although htmx and React can be used on the same web page, they need to be kept in isolation. But I’m not looking for isolation; I’m looking for fundamental integration.

My ideal framework would allow for reactive granular updates to the DOM while also being tightly integrated with compiled web assembly powering complex client operations. I’d write all my code in a powerful language rather than JavaScript. My debugger would operate on both server and client because the difference between the two has disappeared. It would be true full-stack development — a single-stack application (SSA)

Clojure + ClojureScript comes close to being an SSA, but only superficially.

If ever there was a killer framework for Common Lisp, I think an SSA fits the bill.