Cap'n Refsmmat

Hmm, "Server Not Found" suggests a DNS problem. I haven't seen any sign of DNS issues yet (my personal website relies on the same DNS server). Was that intermittent, or did you have that issue for a while?

We're no longer on Invision's hosting after they kicked us off late last year. We're on a virtual server from Hetzner. I'm looking into tweaking settings so I'll have more information about the problem when we get these errors. I also increased the number of PHP processes available to serve requests, which might help with the 502s. I assume they're happening because no PHP process is available, but that's just a guess; I need to tweak the logging settings to keep more history so I know. I'm also going to update our robots.txt to ban some AI bots. Check out the top user-agents accessing SFN today: 11141 "meta-externalagent/1.1 (+https://developers.facebook.com/docs/sharing/webmasters/crawler)" 4692 "Mozilla/5.0 (compatible; Bytespider; spider-feedback@bytedance.com) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/70.0.0.0 Safari/537.36" 1254 "Mozilla/5.0 AppleWebKit/537.36 (KHTML, like Gecko; compatible; Amazonbot/0.1; +https://developer.amazon.com/support/amazonbot) Chrome/119.0.6045.214 Safari/537.36" 754 "Mozilla/5.0 (compatible; DataForSeoBot/1.0; +https://dataforseo.com/dataforseo-bot)" 552 "Mozilla/5.0 (compatible; SemrushBot/7~bl; +http://www.semrush.com/bot.html)" 510 "Mozilla/5.0 AppleWebKit/537.36 (KHTML, like Gecko; compatible; bingbot/2.0; +http://www.bing.com/bingbot.htm) Chrome/116.0.1938.76 Safari/537.36" 422 "Mozilla/5.0 (compatible; AhrefsBot/7.0; +http://ahrefs.com/robot/)" 420 "Mozilla/5.0 (Linux; Android 7.0;) AppleWebKit/537.36 (HTML, like Gecko) Mobile Safari/537.36 (compatible; PetalBot;+https://webmaster.petalsearch.com/site/petalbot)" 401 "Mozilla/5.0 (Macintosh; Intel Mac OS X 10_15_7) AppleWebKit/605.1.15 (KHTML, like Gecko) Version/17.4 Safari/605.1.15 (Applebot/0.1; +http://www.apple.com/go/applebot)" 279 "Mozilla/5.0 (compatible; DotBot/1.2; +https://opensiteexplorer.org/dotbot; help@moz.com)" 258 "Mozilla/5.0 (Linux; Android 5.0) AppleWebKit/537.36 (KHTML, like Gecko) Mobile Safari/537.36 (compatible; Bytespider; spider-feedback@bytedance.com)"That's how far you have to go before you get to humans. And there are only 25,000 requests in that log file, so more than half are bots.

Upon spotting the bat-signal, I checked the server load graphs and found this: That's slightly suspicious, but everything looks fine now. Since our move at the start of the year, we've been on a virtual server hosted by Hetzner. I'm wondering if something went wrong with the physical server it's running on, or the networking in the datacenter. If it happens again I'll try to catch it while it's happening to diagnose.

Where I live, it's common for people to rent an apartment and be astonished by their electric bill in the winter, because they didn't notice the apartment has electric resistance heat. (Not a heat pump, just a giant toaster.) They tend to make confused posts online, somewhat like the BBC article, quoting the total cost but no other information about usage and pricing. As electricity prices rise, electric resistance heat will get disproportionately more expensive.

The traffic seems to have actually ramped up in early December, which resulted in us getting unplugged for a couple of days. However, the hosting service didn't tell us the reason and just turned the site back on without explanation, giving us no chance to understand what was going on before they pulled the plug again.

Hard to say. When I searched around, I found a thread on reddit describing an essentially identical attack (same countries, same User-Agent). Nobody's emailed us demanding ransom or anything like that, and it wasn't nearly big enough to actually overwhelm a decent server. Feels like the low-level background radiation of the Internet.

On Monday morning we were notified that Invision, who make the software and also provided the hosting, were kicking us off their hosting. Apparently SFN was attracting a denial-of-service attack that was costing them money to deal with, and instead of trying to filter it out, they just cancelled our service. Dave and I thus spent some time switching to a new (cheaper!) hosting service and getting everything hooked up again. We're back, though a few things will probably still be broken from the move. I suspect notification emails won't be working right now, and I disabled registration for the time being so we can fix the email system. If you notice anything else weird, let us know. (oh, and we already filtered out the denial-of-service attack, which involved several megabits per second of requests to the All Activity page)

I have approved your request. (Which is also helpful so we understand what happens when we use this feature.)

I just found a "Spam Analysis" feature that oddly didn't turn up the last time I searched for "spam" in the admin panel. I've turned it on, and allegedly "any content submissions made by a user with less than 5 posts, or dormant users who have not posted in 6 months, will be analyzed for potential spam". Highly likely spam will be blocked and possible spam will be held in the mod queue. There's another page with spam statistics, and it claims to have blocked 54 known spammers from registering so far this month. That includes 29 marked "known spammer" just yesterday. But it seems very inconsistent from day to day.

I just logged in to look at the spam settings and saw there was an update available. Coincidentally, the update has a new CAPTCHA option (Cloudflare Turnstile). I switched to that, so let's see if it makes any difference.

Unfortunately I don't see a setting to control it. Ideally it would only auto-recognize links starting with https://, but there's no setting for that.

I'll take this opportunity to point out that yes, bulleted lists with emojis are a feature of LLM output, and yes, LLMs have a distinctive writing style. Oddly enough this is one of my research areas now, collaborating with an English professor who knows how to quantify various features of writing style. We wrote a paper (free preprint) where we generated a corpus of LLM and human writing. ChatGPT, for instance, really loves participial phrases and an information-dense style: lots of nouns, noun phrases, and nominalizations (things like "development", where the verb "develop" is turned into a noun). They're features that display authority. More recently I looked at formatting, and while I don't have full results, the more recent models seem to particularly love using bold, bullet points, and sub-headings. I should add some code to count emojis.

Looks like bad design: the counter showing the number of upvotes gets in the way. But if you click on the heart symbol directly, you'll upvote the post, even if you can't reach the arrow. So this mainly stops you from downvoting, and we can call that a feature to encourage positivity.

It looks like what you copied is a heading. Under the ¶ menu, try changing it from "Heading 1" down to "Paragraph".

Hmm. The AdminCP won't let me change the sitemap URL settings, I think due to a bug. I'll have to ask Invision to look into it.