Ask HN: How to improve dup management?

Suppose there were a 'dup' button on any post (next to 'flag', maybe) that not only marks a post as duplicate content, but lets you link it to the particular other post that you think it's a dup of.

This, you see, is the hard part.

Every other thing you might imagine doing about duplicates can be algorithmic. Some suggestions are below, but any good ideas we have require something like a graph of duplicate posts. Some of them only require a tree-based (parent-only) annotation, but if you represent it internally as an undirected graph you can topo-sort it on various criteria (popularity, chronology, etc) depending on the algorithm you want to play with.

Some things you could do with the graph:

- Distribute karma among the various posters.

- Merge discussions among the various posts.

- Include a "previously posted as" link on the discussion page

- Display all the previous and subsequent dups (e.g. in an inset float-right box, or in a list above the discussion)

- "Lend" post karma from one dup to another in some fashion, so the popularity of one post (and its rank on the front page) affects another.

I'm not suggesting that marking a dup would automatically kill the dup---in fact, I'd suggest that that not happen---but it seems to me that the only way we can do anything intelligent with duplicates is if we have reliable information on which posts are duplicate content (and dups of what).