I have lots of raw notes, but I'll wait to see whether the presentations show up at the Open Repositories 2008 conference repository (for some reason, I keep wanting to spell this "respository").
One of the main themes that I've heard in terms of doing science with repositories over the past couple days is that presentation formats, particularly PDF, are destroying the data (e.g. chemical structures and reactions) that we have so carefully assembled. Then we have to make machines work really hard to try to reconstruct this data, which is madness to me (although I accept it may be the only practical solution in the near term).
I would argue that HTML plays a similar role in emphasizing "what looks good" rather than adding to that "and is also usable by machines under the hood".
And in a different way, PowerPoint, with its constraints of display and its style of bullet points, discards our complex ideas and presents them in a lossy, radically oversimplified way (with a dependency of course on the skills of the presenters).