In the sense of a “form” just being lines on paper that you can arbitrarily add some text to — no, that’s easy.
Likewise, in the sense of a “form” being some defined input regions that accept your keystrokes and turn them into new text DOM nodes in the PDF itself — easy enough. Though, unlike HTML, there’s no concept of an <input> tag that just has the semantics of accepting keystrokes and turning them into (persisted) input; instead, this all has to be done through scripting [i.e. writing event-handlers, or having some PDF authoring software generate them]; and there are several incompatible scripting languages for PDF that get used, some of which are proprietary with no open specification.
But, doing form validation? Or, worse yet, making one of those fancy PDF forms that auto-calculates fields like an Excel spreadsheet? Now you’re getting into the hairy stuff, because IIRC none of the open-standard PDF scripting systems provide these sorts of mechanisms, so these are inherently proprietary things.
And when I say “proprietary”, I mean “like old versions of Word or Photoshop, where each version emitted its own in-memory data-structures to disk without formal serialization; and it was the job of authors of future versions to write importers to deserialize whatever format resulted.”
If we simply had print-to-HTML functionality which resulted in a document identical to what you view onscreen while editing, PDF could die the death it deserves.
But HTML+CSS somehow manages to suck just as much for common usage, so it persists.
And implementing a PDF viewer is already a major undertaking; adding the form functionality complicates things even more.
The vast majority of form is indeed not « ready » for input, requiring users to go through hoops to fill them. And that work is done again by the next person.
The problem I was referring to, is when one is expected to fill a PDF that was not built to be filled, such as the scan of a paper form.