In order to migrate legacy content, one might resort to importing or attaching old documents into a newly fangled KM repository. This is a slippery slope. Importing/attaching becomes a crutch for users who are just too lazy to get used to authoring content in-line (using rich text editors). They continue to author all documents off-line. So what? Is there is a problem?
As Google has so successfully demonstrated, it is often the hyperlinks between documents that are more valuable than the documents themselves. Imported/attached content almost always lacks relevant hyperlinks to other documents in the repository. They hang about like dead limbs of a tree. What is worse, the offline versions often aren't discarded after import. They continue to evolve offline, unbeknownst to other users. They get attached to emails and soon no one cares about the orphaned version in the repository. The Google docs video captures this sitution well.
But of course, we love the power of our offline office tools and will continue to have powerful spreadsheets and presentations created using them. Sure, just don't tack them on as appendages to the one big repository. There is another repository better suited for offline content. It is called a file server. Better yet, use a web-enabled version control system backed by a file server. This gives us permalinks to the latest version of every document. Get them indexed by an enterprise search product. Oh, but our users aren't version control savvy. Come on. It takes fifteen minutes of training to understand update, add/edit, commit. After all, we all claim to be organisations of savvy knowledge workers, don't we?