Over the current decade, there has been a big hope among the mass software companies that they all can do a Flash and have their own format dominate the web. Silverlight, Google Gears, and so on seeking to get a mindhold on content developers without instant success, and the various Rich Client Platforms and RCP-augmented browsers seeking to get mindhold on application developers, again without instant success but real contenders.
I am not sure whether this came from or caused the stagnation of HTML over the last decade, but thankfully the HTML 5 effort is showing how much scope there still is for improvements to HTML. The alternative to HTML 5 is for websites based on cross-platform APIs: not just browser sniffing but platform sniffing: generate SVG Tiny rendered content for browsers that support that, Silverlight-rendered content for those browsers, and similar for Google Gears, Flash and so on.
What happens then, of course, is that after cross-platform support becomes the prevailing mentality for slick sites, then a period of consolidation becomes important: the distribution system, technical qualities and box office appeal of the platforms (hardly the same thing!) will tend to promote a winner. It then becomes also natural to make a cross-platform markup language that can cope with a wrinkle-smoothed common set of platform features, the pendulum swings and the API gets turfed out of the driver seat in favour of markup.
And this is, of course, where HTML 5 comes in. I think as well as seeing HTML 5 as a way to ward off the evils of proprietary formats, we need to figure out how to use it to neutralize the negative impacts of these formats: if HTML 5 and CSS can be augmented in ways that take advantage of slicker rendering and interaction by the specific-vendor platforms, then their presence becomes a net gain not a challenge to interoperability.
The lesson of the Web is that people prefer reach to slickness, and speed to beauty. But the greatest of these is reach—ubiquity—which is why ultimately any strategy apart from fitting on top of HTML 5 seems dubious, even if we have to get there through a cross-platform API-centric period. It is in the interests of all the platform vendors to make HTML 5+ as good a platform for feeding their platforms (whether by transformation or directly) as possible: richer, more extensible, better degradation.