I'm relatively new here so I readily admit to having a very limited perspective: I used reddit years before I came here, and now I use reddit for just a few specific subreddits; HN has more than enough technical and bigger-picture pieces to occupy me. Certainly, the discussion threads are consistently of the highest quality, so much that I have to fight to read the OP first rather than jumping to the comments. I also admire the mod philosophy of discouraging off-topic, even if clever, posts...it's a fine line between cultivating fruitful discussions and being overbearing, but HN seems to do it better than just about anyone else.
I think it's a universal trait among communities to think that things were better back then. Maybe they were, but part of that mindset is out of a natural regret of years passed and nostalgia. But it'd be interesting to see more data about the quality of HN...obviously it's not easily contained in a metric, but some analysis of discussion length, upvotes per user, upvotes in a time period, distribution of top posts among domains, etc. would be a place to start.
A couple of months ago, someone started an epic discussion about how HN was going downhill. But the crux of his complaint was that people weren't giving him enough positive feedback, and he saw the non-glowing reviews of his two projects as negative sniping. I checked out his two sites (one was defunct, the other didn't seem to have much presence) and the various links-to-them that he posted. There was very little, if at all, sniping, and there were at least a few very constructive, helpful comments. It was disappointing to see so many users, including pg, jump into agreement with the disenchanted user. His overall argument may have had merit, but the reasoning and experience he used to get there were questionable.
But back to the OP: I think a -1 karma per link isn't a bad idea at all. OTOH, I think the (as I've perceived it) penalty for posting too frequently also fulfills that role.