Circa 2007 or so it seemed a to be a huge mistake because it didn’t play well with the so-called “model-view-controller” paradigm where, most importantly, the request handler could decide which view to render. (e.g. you fill out a form and if you mad a mistake it redraws the form with an error message, otherwise it might render one of several different forms depending on what you filled out on the first form.)
He thing about HTMX is that it really wants support on the server side. For instance, you might have a <select> that gets drawn as part of an HTML page or that gets updated with new <option>(s) via HTMX when you push a button. Similarly you might have a form that can be rendered as the only form on a page or that can be rendered as a modal dialog that is pulled from the server when you open the modal (as opposed to unhiding it.). Either way the code that draws the HTML fragment (both the template and any database fetching/thinking) has to be runnable in more than one context. (Modal dialogs are so fun to implement with HTMX.)
It’s something you can do in a situational way or that you could have a framework to help with, but it is part of the HTMX puzzle.
