Adding forms in 1995 did almost
nothing to change HTML from hyperlinked text documents to anything resembling modern web apps. It took about a decade before (somewhat cross-browser compatible) asynchronous Javascript appeared on stage, what changed the whole game. 10 years of forms gave us rainbow sparkle generators and rudimentary webmail. Then 7-8y of "AJAX" gave us a generation of kids that truly does no longer intuitively grasp the difference between native apps and web apps.
While the form element was probably a great improvement (I wasn't there, not until '98 until my first steps in HTML), think about literal meaning of the word "form" for a bit. It's still a document. HTML was normal documents with hyperlinks in it, add form elements and now you also got fill-in document with hyperlinks.
Yes you could potentially make web apps that way, but everything (everything) required a server-roundtrip. It may be a hard to imagine what a far cry that was from what we now consider a "web app". Every single UI change had to go through a form submission request and a response from the server that contained the entire page regardless of what was updated. "... but .. frames!", I can hear you wonder, well IMO frames weren't all that. They didn't really make selective updating possible in pretty much the same way as forms didn't really make web-apps possible either. Plus, I can't remember if there's been any time people weren't hating on frames (it broke navigation and bookmarking--sound familiar?).
I say this, because I remember discussing the idea of web-applications in 2002/03 or thereabouts with a friend in college. I had just discovered the SCRIPT SRC trick so you could actually load data from the server without reloading which was paradigm-shattering magic, at the time (the technique is still in use as part of JSONP, btw). I had already written a chat-client and it was beautiful, I wanted to pick his brain for more ideas of what to do with this new ability.
(please remember, this discussion occurred during the height of the Browser Wars, 1-2 years before the tide started turning. if you weren't there, or doing webdev at the time, imagine it 10 times worse. digital trenches, man)
My friend told me pretty much the exact sentiment I see pop up in this thread here, that HTML was intended as a document markup language, not for application building, and that while we could hack all these cool interactive features on top of a markup language, and surely these were some quite impressive feats of creative thinking, it would always remain on shaky grounds because you should build an application framework from the ground up with application framework foundations, not by bolting things onto a markup language. Back then, I knew this to be true (one picks up a few things about software design, studying CS) even though I didn't want it to be, because the tech was so cool and the possibilities so enormous (hackers should be able to relate to that feeling :) ).
Now, many years later, I got what I wished for, and my friend got to be right.
(at some point CSS and JS incompatibilities started to slowly disappear and the Browser Wars came to a sort of uneasy peace--this was a slow realization I can't really say when it happened, but I don't have to resort to black magic anymore to write a basic webpage that works in FF,IE,Chrome and Opera--particularly if I write it in Opera first)
Looking at the state of web-apps today, we are using that technology ("AJAX") to build things beyond my (I won't say "wildest") imaginations back then. But it's still clunky here and there. Javascript is still a pretty crummy language (though I love coding for that weird prototypal beast), and as soon as you write more than 5 lines of DOM-manipulation, you're going to want a framework like jQuery. And for some mysterious reason, people still manage to write JS code that works in one or two, but breaks in the latest versions of other major browsers. Minefields are hard to clean up I guess. There is more, but you get the idea--and IMO a lot of the problems can be traced back to bolting application-behaviour capabilities on top of a document markup language.
Still, I do believe we're heading in the right direction. Things can be made much cleaner than say 5 years ago, and there's new standards looming on the horizon that may make things even better. Sure, a from-the-ground-up design would have been so great and so clean. But the industry would have probably found a way to crud it up regardless. And given the global scale and heterogeneity of the Internet as a communications medium, the immense market forces that come with such an environment, personal computing capabilities over time, maybe this was the only way it could've been pulled off, who knows?