Jon Udell posits some relatively plausible scenarios that could drive the growth of linkable continuous media. The ability to point at chunks of audio/movies from given offsets is severely lacking with current formats and players. Even worse is integration of the presentation of these chunks with other media, e.g. accompanying text. Podcasting and VOIP may be the drivers for improvements in this area. Of note is the connection by Real Networks' Rob Lanphier to the open source, Helix toolkit, which provides a hackable version of Real's player and server software and could form the foundation for making this easier. SMIL is the technical underpinnings of it all.
It'll be an interesting time when the world easily and routinely links into more than web pages.