Joerg Kantel, aka der Schockwellenreiter. is the second keynoter talking about Learning from Weblogs: Community, Peer-to-Peer and Authenticity as a model for future knowledge collections.
The paper is in German.
Basically he starts out explaining, just like the original internet aimed at, that with large collections of knowledge you are very vulnerable if this collection is concentrated at one point, thus you have to distribute the collection. Tools bloggers use are suited to bring this distributed knowledge collection into existence.
One example would be the use of CSS and templates by bloggers. If the content of a collection (maybe packed in XML) is stored independent of the template, different organisations can serve the content in their own style.
Trackbacking, pinging and RSS-syndication would keep the distributed collection together.
Also (probably to the delight of Lilia and Dina, who have been arguing for this already) you would need some alternatives to HTML as ‘humanities scholars can’t be expected to know anything about html or worse xml’ to be able contribute to such collections of knowledge.
The other important thing is meta-data.