That's not to say HN doesn't have flaws. But these aren't the droids you're looking for.
The droids you're looking for have to do with solving the problem of growing a community and keeping it intelligent. Inevitably, the tyranny of the majority will prevail, people will leave HN and go to some place smaller and newer, as did those who came here from reddit who came here from digg who came here from slashdot who came here from John Carmack's .plan file.
Maybe this is the essence of what all these points amount to.
This reminds me of research I'm doing with Vivek Wadhwa on how the effectiveness of entrepreneurial networks (say TiE) decreases as they grow (or any innovation networks for that matter; maybe even Google) :)
1. It takes only 1 hour for a post to disappear from the “New” page.
Inevitable for a news site.
2. The Point system sets the wrong incentives
Most people seem to be in it for the points.
This could happen, but I haven't seen many signs of it. Personally, I try to avoid making comments that might get a negative karma score, but I don't much care how high a positive score I get. IMO The points system works well; I especially like the way that negative karma posts are 'named and shamed'.
3. There is no way to search
This is a reasonable point.
4. Using HN as a discussion forum doesn’t work.
Ask and Tell HN threads appear frequently, so this doesn't seem to be a problem. And I find the discussions on HN to be excellent.
5. There is no way to revive old discussions
IMO, this is a good thing for HN. It prevents it turning into a traditional web forum - not that web forums are a bad thing, they just tend to be more insular, and there's more of a split between regulars and newcomers.
5 is also a very nice feature that allows never-ending fights known from forums to die. There's no way to bring back some old rant by replying "You're wrong!". Unfortunately that happens on mailing lists and forums quite a lot.
Otherwise I can imagine the front page would consist of "git vs hg", "dvorak vs qwerty", "python vs ruby", "nosql vs rdbms", "lisp vs everything", ...
For searching HN, http://searchyc.com is a fantastic service, much better than the HN Search linked at the very bottom of HN.
If you're interested in more discussion with startup folks / if you find that Ask HN posts are usually your favorite, you may also enjoy: http://answers.onstartups.com
About searching: Interestingly Google loves HN. At least I when I search for stuff that is mentioned in the comments, I often get the very comment as a search result.
I think the fact that conversations go stale quickly is a feature. It keeps things from becoming too unproductive, and makes "the last word" in a debate arbitrary.
If you want to continue a discussion that has gone stale, you write a new blog post about it and post it. Discussion continues.
Point taken. But still, as a data-mining person (yes, shame on me), it feels like a lot of information and opinions that have ben created by some of the smartest guys around the internet are discarded of.
Maybe a best-of page would help collecting really thoughtful, long and well-crafted comments on a subject?
That's true, but if you keep meeting smart people in the hall, maybe you would like to start a more substantial discussion with them. This is currently not directly achievable (except for writing blog posts back and forth)
HN is written in the 100-year programming language Arc using the first Unlimited Scalability tactic of keeping everything stored as serialized hash tables. Any performance problems you may experience are due to your lack of ability to perceive perfection.
That's not to say HN doesn't have flaws. But these aren't the droids you're looking for.
The droids you're looking for have to do with solving the problem of growing a community and keeping it intelligent. Inevitably, the tyranny of the majority will prevail, people will leave HN and go to some place smaller and newer, as did those who came here from reddit who came here from digg who came here from slashdot who came here from John Carmack's .plan file.