Hey folks,
You may have noticed we took about 2.5 hours of downtime this morning for a
performance improvement.
Previously we memoized the number of upvotes and downvotes on both the stories and comments tables.
Many queries sorted by score, which they had to calculate per-row with upvotes - downvotes.
No queries directly used upvotes, so I replaced it by memoizing score instead.
The migration for this locked the tables, so we took downtime - that’s now completed. @355E3B and I are squishing have squished some small followup bugs; please file github issues or let us know in chat if you see anything odd.
And then as long as finishing that migration meant running it five times on my machine to iron out bugs, I picked up some other features we’ve been figuring out for a long while.
I’ve redesigned the tags page with categories to help make sense of it. This was prompted by @355E3B’s analysis of our tags and the page’s longstanding design issues. I’m leaving the issue open because there’s still room for improvement. Perhaps use our newly added svg graphing library to add sparklines showing usage?
I know categories are a really tempting bikeshed, so let me emphasize that categories and tags are not intended to be an epistemologically complete ontology of the entire field of computing. They make the page look nice. Maybe, like domains, they’ll slowly grow functionality, but let’s not overthink them too quickly.
I’ve revamped the comment flagging UI to further emphasize that they’re for alerting mods when a user doesn’t want to engage and curb their abuse for punishing disagreement. Visually, flagging comments now looks like flagging stories with a text link rather than an down-pointing arrow, and flagging a comment collapses the tree. Nothing on the site refers to “downvoting” any more.
I’ve also expanded the about page with material on flags and topicality compiled from many meta stories and subthreads about topicality and the purpose of flags. These changes also address a recent thread on whether the warning about being heavily flagged by trying to improve the signal of flags so that granular warnings aren’t necessary. (These links are not exhaustive.) I’ve also deactivated the ‘Incorrect’ flag reason that was mostly used to disagree with people rather than indicate serious problems.
I hope this is a roughly accurate collection of community norms and expectations. As usual for changes, these are not the final version we’ll ever have. If you see ways to improve these things, I appreciate help identifying what’s wrong, proposing solutions, and implementing them.
Finally, last month I asked kyle if he was still interested in being a moderator, as it had been a long time since he was active. He replied that he’s not not active online much anymore and not interested in continuing as a mod. A few days later he deactivated a bunch of online accounts, including this one. He leaves with our thanks for volunteering and best wishes.
- Peter
Always happy to help. It was a whole lot of fun working with you.
I think this is a great change. A lot of times people downvote out of disagreement. It would be interesting to see how this change is going to affect this behaviour in practice. I’d imagine that psychologically people will be less inclined to flag on disagreement than to downvote on disagreement. Downvoting is cheap and (seemingly[1]) anonymous; whereas with flagging you are explicitly asking a moderator to engage their attention.
[1] Of course, the site mods have full access to voting data, so it is not actually anonymous. Though this is not the case on other sites like reddit.
I wrote a detailed comment on how this works a while back. The relevant bit is that flags and flaggers have always been visible to mods.
Really great work folks, thank you!
V.v.V
Great work as always. There is likely no version of collectivized content curation that totally removes all noise from the signal, but I’m continually impressed with this site’s high signal-to-noise ratio and deep cultural commitment to subjects of technical interest.
Keep it up. Your work is appreciated, even when that goes unstated between feature upgrades.
Oh wow, I’ve been misusing and abusing flags without realizing. I figured they were analogous to downvotes on other platforms. Changing the UI to make this more obvious was a good move. Thank you so much for your hard work!
As always, thank you for your service! :)
Thanks!
Interesting. I wonder whether that helps to curb the customary “unkind” flag in response to suggestions to read the linked article when it’s painfully obvious that the person hasn’t read it.
On kyle’s profile, it says he submitted -3 stories. What?
https://github.com/lobsters/lobsters/issues/771
Apparently the fix is not correct, or at least not complete.
I’ve seen that sometimes on accounts that have been deactivated.
So no downvote possible at all?
“Downvotes” were never possible really, it was always intended as a way to flag that there is a (significant) problem with the post, rather than “I don’t agree” or “I don’t like it”.
I think the gist is that you either agree with / appreciate a comment and upvote it, or you think it’s inflammatory / misleading / trolling enough that you flag it. Disagreement was never the intended reason for downvoting comments.
Lobste.rs keeps getting better and better. Thank you!
A tacit admission that no one actually programs in Perl.