http://ycombinator.com/newsguidelines.html
Up- and downvoting depends mostly on karma thresholds, though there are also some protections against karma-bombing and voting rings. You can't post comments on items that are dead or over 45 days old.
Most of the details can be found in the source:
why one is sometimes unable to reply to a comment
There is a cooling off period when there is vigorous activity on a subthread to avoid flame wars. I think this is a recent change.
"the title of a submission is often changed arbitrarily"
The guidelines are pretty clear on this one: "You can make up a new title if you want, but if you put gratuitous editorial spin on it, the editors may rewrite it."
"down- or up-voting is sometimes disabled"
I've never seen that. Downvoting for comments is only available to logged in accounts with a karma of 500 I think, and not for stories.
And why does a thread sometimes turn up locked, with no explanation Is HN now Wikipedia?
I don't know what the dead-ing mechanism is, but it certainly not Wikipedia - HN is a private site and I expect the editors to do what they like.
[Edit: Thanks to rms and tokenadult for the corrections below]
It is - I think it's one of the best (the algorithm apparently scales too - so longer waits the deeper the thread). it's stopped me posting a rash response once or twice.
Some submissions are autokilled because they are links to sites on the list of banned sites.
http://news.ycombinator.com/item?id=499044
Some submissions are autokilled after they have been flagged by a sufficient number of users.
Some submissions are manually killed by curators for the usual reasons every public website needs curators.
Most of these questions are answered in the FAQ, or turn up sufficiently often that I would expect a casual reader to know them. Clearly I'm wrong.
So this goes for other sites as well: How can you arrange information such as this, so that both new and experienced users can find it easily, instead of being driven to ask questions?
Can we create a really useful question answering system of which questions like this can be asked, and which mines a database for the answers?
There's a start-up waiting to happen ...
Can't upvote or down vote a story? white-out the arrows and put an alt="can't (up|down)vote story, you don't have enough foo".
Can't reply to a post? remove the link but leave a white "reply" text in a box with gray sunken borders. Hovering over it should show the reason why.
Each deactivated post can also log its status and reason for disabling right in the title, or somewhere in the header area; maybe the top-color can change?
People shouldn't be sent to an Information Desk; instead, the system should be self-documenting, if not self-evident.
However, and this is where I wish I were more eloquent, and more of a writer than doer, because I'm sure I'm not going to say this as well as I want ...
I can't help but feel that there are cases where the ideals of instant obviousness and total transparency actively prevent the potential for power use. Insisting that everything is self-documenting is, I feel, inherently self-limiting.
I feel sure that in some cases it's right to have a learning curve, and some effort required to be expended. Insisting that everything be instantly obvious to everyone who stumbles across it is, I think, creating the expectation that everything should be like that.
It's a gut reaction. To paraphrase an old, hackneyed and much criticised observation: Everything should be as obvious as possible, but no more.
Perhaps there's a place for the unobvious.
What we need is to study and consolidate "Adaptive User-Interfaces" (a well stablished field of study, btw) with software engineering, and specially the web (and even more specially, with respect to dynamic programming languages and in my case Lisp and CLOS.)
Something I have been doing lately was reverse-engineering a few google apps from their rendered HTML and minified JS. I don't wanna throw speculations around, but I think Google has some sort of a "learning" engine which adjusts the shown content to the estimated level of user expertise. They do this with a "newbie cookie". See the sources for any google apps help pages, for example, and you will see this:
var global_newbie_cookie = "newbie_11";
http://www.google.com/support/a/bin/answer.py?answer=77003
Your global newbie cookie should be different than mine.
Regards.