That might work... but only if people actually read that page. Given how few people look at /newest (as estimated by the fraction of votes which are cast before submissions hit the frontpage) I'm not optimistic.
How about only placing comments into the "pending" purgatory if the submission they're attached to has received more than X comments in the past Y minutes? I assume it's the chatty discussions which you're concerned about cooling down, so this would handle the problem case while avoiding the side effect on quiet/abandoned threads.
I think I like this idea quite a bit. I don't know how many users there are with >1000 karma, but will they be motivated to keep endorsing everyone else's pending posts? Sometimes good discussions do happen on quieter threads, or way down the list that the 1000+ users might not see.
Well, I guess it really comes down to just how much (proper) endorsing ends up happening. One thing I like about HN is that it's open and fast to use. Having a pending mode on everyone's comments affects not only the troll-users, but many of the normal ones too who aren't abusing the system ><.
I was thinking the same thing. There's currently no known harm for upvoting, so we do it to promote good discussion, despite the fact that there's typically no personal benefit. But if endorsing posts not only provides no personal benefit but also bears the risk of possible harm, people might be too cautious with endorsements for this system to work.
One concern is that there's no direct feedback to the endorser, so he or she would have no sense of the relative importance of their endorsement to keeping the discussions rolling along. With visions of the Stack Overflow police, how high a bar should a comment have to pass? Although maybe that's a positive -- after all, we no longer directly see comment scores, only the derivative effect of thread reordering.
So, have a new meta-karma that is equal to the mods given to the posts you endorsed? (An alternative would be giving karma directly, or some fraction of karma).
Perhaps that could work, although it goes a bit down the road toward heavier and more explicit mechanics, as at Slashdot and Stack Overflow. In another comment somewhere in this thread, pg mentions that he'd like to keep it as simple as possible. Unspoken is the "...but no simpler" part.
It's fun to consider that a dynamic system could provide someone on the back end (or a smart algorithm) with a variable nozzle controlling comment flow.
But endorsing not at all guarantees loss of that privilege, in that you're never actually exercising it. If the penalty doesn't extend further, I'm not sure the cost will discourage terribly much.
>I don't know how many users there are with >1000 karma
There are about 5500 such users. There have been around 245,000 users to ever post on Hacker News and around 85,000 users have posted over the last year.
Source: I'm working on a fork of the Hacker News Karma tracker (not ready to be live yet) that uses Algolia's new HN Search API; I also downloaded every comment ever made similar to how minimaxir downloaded all of the submissions.
I wonder how many upvotes we would collectively have to give each day to bring comments out of pending?
Looks like in the last 10 minutes we've had 30 comments on HN - that's about 180 and hour - that's 4,320 comments in a 24-hour period.
Let's say 50% of those are worthy of being seen. If we assume it takes two upvotes per comment to bring it out of pending status then that group of 5500 people need to cast 4,320 upvotes a day collectively to to stay caught up.
Given the fact that A: it's unlikely that all 5500 of these users are still active, and B: it's very unlikely that they would be upvoting the same comments then it seems almost certain that there will be a SIGNIFICANT backlog of pending comments created each day.
Well, hopefully new users can hit the 1000+ mark quick enough to not have it affect them too much.
But, I wonder how the lack of visible scores on comments has affected the overall comment scoring rate. For the average new user today, how long do they have to wait to get to 1000 karma? How long was it a few years ago? These would be interesting questions to answer.
I joined more than 2000 days ago. I don't have 1000 karma. This way I'll maybe never will reach it. Seems a bit of a high threshold to me. But maybe my commentary just isn't good enough.
Karma per post is not well correlated with the actual quality of your contributions. You can get hundreds of karma for posting superficial observations early on a post that become popular. Someone that has observed HN for long enough could probably get 1000 karma in a day from just trying to optimize for karma instead of quality.
If you really do want to reach some karma threshold I would suggest trying to get it from submissions instead. You can easily get hundreds of karma per submission if you submit the newest release of some popular software product or the latest Zed Shaw rant with very little invested. This also has the bonus that you are not actually lowering the quality of your commenting for more karma.
That, and doing some basic research to post additional useful information (facts, numbers, stats, references, background info, etc) on a relatively new thread.
I'm mentioning this because I believe such posts are genuinely useful and surely deserve the points they get.
BTW I don't think I ever got near a hundred points for a comment. But I suppose it can happen in the above circumstance in a popular thread.
Maybe the easier solution would be to have two thresholds - one for un-endorsed posting and one for endorsing. Or having a heuristic that includes ave. comment score as a factor. Or... Or...
Honestly, the more I think about this, the more complicated it seems, which is usually not a good sign.
I'm not sure it is a worthwhile endeavor to hope that new users hit 1000 karma. I think the discourse would be better served by having a wide variety of people saying something worthwhile occasionally rather than trying to say something popular more often to play a karma game.
The best idea I've heard so far is a timeout, such that even if no one endorses your comment is to let you comment again anyway after a day.
I'm still not sure what the benefit of adding endorsements is compared to using a generic upvote as the endorsement.
> How about only placing comments into the "pending" purgatory if the submission they're attached to has received more than X comments in the past Y minutes?
Fluff will pass through.
Another option would be to enable the pending machine on big stories for the time they are on the home page, then auto-validate everything (but mark the unendorsed comments as such).
It would be nice if users with a lower karma could validate the replies to their own comments, once they have been validated.
I like this idea; perhaps it could be a continuous function. There's already some sort of velocity-like calculation, with a time coefficient, so stories have to kick increasingly hard to keep their heads above water. It seems logical to attach the posting threshold -- a new cooling saucer dimension -- to that same calculation.
Let's see how much of a problem there is first. I wanted to start with the simplest possible thing. If it breaks in some cases I'll add stuff to fix those.
The are discussions where user name matters, for instance when refering to the nth parent in the same thread, or when trying to bring more context to a point made previously or even retracting a comment (knowing it's the same user posting is important).
We could get away with temporary user names, changing on a per thread basis for instance, but that might be heavy to implement.
>for instance when refering to the nth parent in the same thread
waterlesscloud's suggestion is to hide them only while the comment is pending. Once the comment is approved, then the username can be displayed as normal, allowing references.
My immediate thoughts were very similar to cperciva's. In the status quo, there's already a disincentive (for those who care about average karma) to comment on any posts that require scrolling down, especially if posts above them are heavily nested or if they themselves are. I can think of a particular discussion I had with a well-known HN user who has professional/financial incentive to care about his karma statistics, and as our two person back-and-forth got slightly too nested but very much unresolved, he merely liked my final comment and never replied.
If pending comments are applied to anything except the top-level, I could see this having disastrous effects on the quality of response in discussions since responses in low-traffic branches will likely not even show up.
With the availability of browser plugins and user scripts, I anticipate an "off-HN" application popping. Interested people can shadow the "canonical" HN discussion and continue a discussion that has legs, possibly grafting it onto HN itself.
Of course, it may be simpler to bot up a subreddit and do the same thing via convention. That may be the best result, redirecting the reddit-like dross back to reddit, where it belongs. Throwaway accounts will be mechanically discouraged along with the me-too, ya rite, and other useless posts.
Hopefully this will prove overly pessimistic. The way I see it, either we'll be proven wrong or pg will revert to the system we know and love until he comes up with a better method for improving the quality of comments.
I actually found it optimistic, in the "destroy a village to save it"[1] sense. HN, as it currently stands, will cease to exist. I suspect the volume of submissions pointing directly to old wikipedia articles will dry up. I consider this a good thing.
I suspect the volume of submissions that are reposts piling onto something already on the front page (Erlang, Erlang, Erlang, Erlang, Haskell, Haskell, Haskell, Go, Go, Go, Snowden, Snowden, NSA, Erlang, Lisp, Lisp, Lisp, Lisp-flavored Erlang, NSA, Erlang, Erlang, Bacon and Spam, Javascript, Framework, Framework, NSA, Erlang, Haskell, Haskell, Erlang, Lisp, 2048, will dry up. I consider this a good thing.
I suspect I will spend less time on the site, either because conversations will become static expressions of views or because I won't have to filter through as much content, even though much of which marginalia I find quite engrossing. I consider this a good thing.
What comes next is open to conjecture. It could be a more mature salon full of reasoned discussions or it could become a ghost town with lots of great, old, discussions.
That seems rash. The 'simplest thing' is a first-order approximation of a good thing. It could quash conversations, lock out users for days perhaps. And how can someone even talk about problems, if they're locked out? Catch-22.
At least measure the results, including people who give up and go away. It doesn't take much frustration to discourage even an active user with cogent remarks. I can see the quality taking a dive when regulars are driven away.
Looking at it statistically - there's always going to be a chance that a given comment will remain pending indefinitely. Over a long enough timeline, a greater and greater percentage of contributors will be unable to post.
Toxic replies in threads are even worse than toxic top-level comments. A toxic top-level comment probably will drop to the bottom of the page, and, more importantly, the toxic reply is personalized.
I totally agree here. The replies are the real issue.
Just from personal experience, if I have something to share on the topic, but the discussion already has a couple hundred comments, I look to contribute as a reply to an already highly rated comment. I'm much more likely to get actually engagement that way.
The issue isn't comment quality. It's UI. New comments, even on busy articles, should be discoverable. It should be possible to have discussions past the front-page-life of an article. It should be easy for the reader to decide whether to explore a given thread of conversation in depth or skip it altogether.
How about only placing comments into the "pending" purgatory if the submission they're attached to has received more than X comments in the past Y minutes? I assume it's the chatty discussions which you're concerned about cooling down, so this would handle the problem case while avoiding the side effect on quiet/abandoned threads.