I've been spending the day at the future of TV conference I mentioned below, and am hearing recurring themes from people here, which mesh with a lot I've been thinking.
I spoke about a key technological driver, which is storage. Classic broadcast TV had everyone watching at the same time, because there is no storage at all. The next iteration is a central archive, but these are fragile - witness the BBC's loss of large numbers of programs from the 80s due to videotape erosion. Tivo, iPod and BitTorrent are manifestations of storage moving to the edge, streams becoming files, and giving us more choice. Streaming is a throwback to the 2nd model - Rick Prelinger says 'streaming is for sissies'. Multiple copies spread to the edges are more resilient, and more remixable, and so create a larger cultural footprint.
Old school search looks at the content, and tries to derive metadata from it. With text we can find keywords, but with audio and video this is difficult. The classic attempt is to scrape closed captions, try shot detection, try phonetic transcription, but all of these don't help much.
Traditional media production goes through a broadcast funnel that strips out all the structure that was there while it was being made - shots, scenes, production notes, alternative takes. DVD production is finally preserving some of this, but it is often explicit recreation.
Then there is the afterlife of media, which is its cultural impact and the discussion, recommendations, remixing and inspiration that goes on afterwards, and which provides the richer context and description.
What is happening is that the edge culture, the long tail, is spreading a bigger footprint, while the locked-up media from the centre is shrinking it context. My cousin Robert does video restoration for the BBC, and often relies on discovered amateur recordings to reconstruct destroyed recordings.
So how do we help this? Tagging, citing and annotating are already working for text and pictures, lets do this for audio and video too.
A good step would be to converge on a way to add media metadata that is easy to create and share. We've started a microformat process by collecting media metadata examples.
Another thing I heard from multiple people is a desire for a way to pay for the media created by remix culture. My mediAgora idea comes to mind.