This is a fri through mon inclusive. The sat was actually the day of startup school.
Orange = requests/sec; black = memory used; red = % time spent in gc; blue = median msec to serve an item page; green = median msec to serve the frontpage. The last two are unnaturally spiky because I use a very narrow window (because we use the same data to to drive our Leftronic screen at YC). So the blue spikes don't represent anything that would have been observable to users.
The slowly rising gc percentage (red) is a sign of memory leaks. They used to be a lot worse before Rtm finally tracked down the main culprit this summer.
We're now getting a bit over 900k page views per day.
Though fierarul's comment leads to a natural question. The original comment explaining the graph is really necessary for the link to make sense. Explaining the graph in a comment seems to break the (apparent) convention of submissions being self-contained.
I thought the reason for disallowing both a URL and text in a submission was to avoid this. Though, if a submission that requires a comment is acceptable, why not allow submitting both a URL and text?
There is a difference. A comment is voted up at the discretion of the users, while text associated with the url would be displayed at the top no matter what. Which would mean the first person who submitted a link could essentially make their comment upon it always be the top one, and define the discussion.
I think you're missing the point. Imagine if everyone on hn posted a link that directed people to half of the content, and put the other half in a comment. That would make for a pretty crappy trend. Note that your original comment has 25 points, so as you can see there's incentive for people to start doing this.
I think a non-vote-able comment at the top wouldn't do as much damage as you might think. Maybe, instead, make it a rule where half & halfs get filtered/blocked.
I really liked the flat-screen monitor up in the YC offices to track the site. I want one for my site. It seemed the site wasn't getting much traffic, but then I realized most of the users were in the room at the time.
It was in something I added quickly to fix a hole someone discovered. I made the mistake of memoizing the function that generated authentication keys for vote links.
HN is sort of a loss leader for ycombinator. Free advertising for all the YC comings and goings. HN is the leader in mindshare amongst those interested in startups, startup technology and startup culture and (obviously) startup news.
HN is already monetized, albeit indirectly, through discrete and relevant references to YC products. For example the reference to Leftronic in this post.
Orange = requests/sec; black = memory used; red = % time spent in gc; blue = median msec to serve an item page; green = median msec to serve the frontpage. The last two are unnaturally spiky because I use a very narrow window (because we use the same data to to drive our Leftronic screen at YC). So the blue spikes don't represent anything that would have been observable to users.
The slowly rising gc percentage (red) is a sign of memory leaks. They used to be a lot worse before Rtm finally tracked down the main culprit this summer.
We're now getting a bit over 900k page views per day.