Another option would be to allow the withdrawal of pending comments by the submitting user. Reply -> you already have a pending comment (with a link to the explanation of the system). Would you like to withdraw your comment X in thread Y and post this comment? -> Y/N.
A good idea but race conditions could still cause both comments to be posted. Maybe if we decide the user can't do that reliably, we cod just let that case slide.
Surely allowing two comments to be pending (and posted) rather than one, isn't a big deal? I do wonder if this can scale, though. I usually open up the top 10-15 interesting (to me) stories, along with a list of my old threads, whenever I take the time to check hn. I then reply to old threads (if that seems to make sense), and post the occasional comment across the stories that I find interesting.
Do we really believe positive moderation will be in the range of a few seconds to a minute?
I know there's been really great experiences[1] with negative moderation/flagging -- but then all users could flag, not just a subset -- and flagging something that's clearly wrong (as in goatsex wrong) is much less effort and much higher incentive than approving a somewhat contributing comment to a story.
Which brings us to what the goal of a comment policy should be. Should we really work towards discouraging people to post things like github-links to stories missing them, because sometimes they'll be beat to the punch by someone else, and now have to wait before contributing to the discussion on a different post?
Perhaps allowing "one pending post per story" might work better?
[1] I'm not sure which talk this was from, but I think it was "building web reputation systems" with an example from Yahoo that touched on flagging (users flagged in sub-second time, much better than automated spam detection). Not sure if this is the same thing(s), but they seem relevant to this discussion:
