Haha!
Ok, this one was from over 24 hours ago.
So it includes the text from some posts but not others. And you can’t just copy and paste the text and poster’s info.
Thanks for implementing wide mode. I’ve found it only makes posts wide if the ‘Post metadata below post’ option is also enabled.
No bother, and I did it way too quickly, so I’m not surprised there are things wrong with it. I’ll take a look. So hang on, are you the guy who lives in the old Luma factory? I’m probably gonna be wrong again here, but I asked someone else yesterday!
Yeah, I’m not sure what that’s about. That one should. A bug, I guess.
I will be doing the bookmark thing; in fact, I would’ve done it already as I looked at it this morning, but it turned out to be a bit more complicated than I initially thought, so it’ll take a bit longer.
Is there a way to block ai scrapers from accessing the site? That’s kind of gross, worse than ad scrapers even.
Hey N, well no way to block them in a sense, strap yourself in... They make themselves look like normal web requests, because if there were some clear marker it were an AI bot, people would just block them. What a lot of work here has been to do is to at least cope with that situation as best I can.
Traffic is placed in different buckets:
Anonymous traffic
Logged in users
The front page
People trying to log in
Also anonymous API requests (computer to computer messages)—these are just blocked
And within that there are two controls
How many concurrent requests you are allowed (i.e. requests at pretty well exactly the same time), and
How many requests in a thirty second period you are allowed
The numbers are unimportant and I don’t know what they’ll actually be yet; they’ll require tweaking, but:
Anonymous users (AI etc.) get a certain amount of computer, but it’s capped at a certain point (based on the two aforementioned criteria)
People trying to log in get certain amount of computer—quite small, as it doesn’t require much, but it’s ring-fenced from the anonymous user stuff so it’s not overwhelmed—it’s so real users are able to log in during an ‘attack’
The front page also gets it’s own allowance to make sure people can find the site, even if it’s ‘under attack’
Logged in users get unlimited access (i.e. the rest of the available compute)
So you can’t prevent this stuff happening, but if this works the way I intended, then it should mean that even under attack, logged in (and trying to log in / front page) users can still operate as if nothing is happening.
I have tested it at 5–10,000 transactions per second on my laptop, depending on the type of transaction. By way of comparison, the existing site managed 400 transactions per second when I last tested it, although that was years ago.
Thanks,
Lemmy.
Oh and if an AI thing tried to register as a user and do it that way, I haven’t done this yet, but it’s quite easy to deal with, because real users don’t throw thousands of requests per second at the machine, so I would just cap logged in users at something way above real users generate, but that would stop AI bots easily—basically at that point if they’re logged in, then you can identify them.
The live notifications of new posts makes me fantasize about a Keith is typing... sort of notification so you know who’s imminently about to post also.
Sorry, I draw the line about fantasizing about Keith.
Wide view seems to work! Thank you Keith