While this is a bit of an after-the-fact view, it seems like there was a moment where they basically said "we've got enough tags." Richer and special-case capabilities tended to move into (first) applet platforms like Java and Flash, and later JavaScript and CSS enhancements atop a fairly thin platform.

I feel like this was the same stall-out that created "it took until HTML5 before we had a more or less universal slider form control" and "there's still nothing resembling a OS-native dropdown menu or rich-text widget in pure HTML."

That time period sort of coincided with the window of a couple years-- when we still had to think about 40-bit and 128-bit crypto offerings, you had to explicitly navigate to a SSL domain to check out, and things like Flooz were still considered a viable idea. Nobody really knew how to take a payment online, so we ended up basically copying the payment from from a mail-order catalog in HTML. A strong "this is the secure and easy way with big clicky buttons" proposal, supported by the leading browsers, would have steered the nascent market effectively.