The two aren't the same, one is about credit, and the recommendation is about hosting. The first can be solved by the author contacting W3C, and if they don't respond, then that is poor form by them.

> just point the original document and forget about hosting a copy. The Harvard website is not going anywhere.

Assuming the Harvard site might not, that document could go missing. It happens all the time and will continue to happen no matter how confident you may be in your systems and how many pinky promises you make. For the same reason that StackOverflow discourages links to answers instead of summaries on their site, keeping the documentation together means that there is continuity.

Imagine you click a link to view the RSS spec from W3C's site and you are hit by a maintenance page. That might be acceptable for blog posts linking to other sites, but for a spec site it feels important for everything to be as self contained as possible.