We have paused all crawling as of Feb 6th, 2025 until we implement robots.txt support. Stats will not update during this period.
Robots.txt is a lot like email in that it was built for a far simpler time.
It would be better if the server could detect bots and send them down a rabbit hole rather than trusting randos to abide by the rules.
It was built for the living, free internet.
For all ita dark corners, it was better than what we have now.
It is not possible to detect bots. Attempting to do so will invariably lead to false positives denying access to your content to what is usually the most at-risk & marginalized folks
Just implement a cache and forget about it. If read only content is causing you too much load, you’re doing something terribly wrong.
False positives? Meh who cares … That’s what appeals are for. Real people should realize after not too long
Every time I tried to appeal, I either got no response or met someone who had no idea what I was talking about.
Occasionally a bank fixes the problem for a week. Then its back again.
While I agree with you, the quantity of robots has greatly increased of late. While still not as numerous as users, they are hitting every link and wrecking your caches by not focusing on hotspots like humans do.
You need a bigger cache. If you dont have enough RAM, host it on a CDN
Sure thing! Help me pay for it?
It would be better if the server could detect bots and send them down a rabbit hole
Already possible: Nepenthes.
ANY SITE THIS SOFTWARE IS APPLIED TO WILL LIKELY DISAPPEAR FROM ALL SEARCH RESULTS.
I’m sold
stoped
Well, they needed to stope. Stope, I said. Lest thy carriage spede into the crosseth-rhodes.
Did someone complain? Or why stop?
No idea honestly. If anyone knows, let us know! I dont think its necessarily a bad thing, If their crawler was being too aggressive, then it can accidentally DDOS smaller servers. Im hoping that is what they are doing and respecting the robot.txt that some sites have.
Gotosocial has a setting in development that is designed to baffle bots that don’t respect robots.txt. FediDB didn’t know about that feature and thought gotosocial was trying to inflate their stats.
In the arguments that went back and forth between the devs of the apps involved, it turns out that FediDB was ignoring robots.txt. ie, it was badly behaved
Interesting! Is this over a Git issue somewhere? That could explain quite a bit.
might be in relates to issue link here
It was a good read, personally speaking I think it probably would have just been better off to block gotosocial(if that’s possible since if seems stuff gets blocked when you check it) until proper robot support was provided I found it weird that they paused the entire system.
Being said, if I understand that issue correctly, I fall under the stand that it is gotosocial that is misbehaving. They are poisoning data sets that are required for any type of federation to occur(node info, v1 and v2 statistics), under the guise that they said program is not respecting the robots file. Instead arguing that it’s preventing crawlers, where it’s clear that more than just crawlers are being hit.
imo this looks bad, it defo puts a bad taste in my mouth regarding the project. I’m not saying an operator shouldn’t have to listen to a robots.txt, but when you implement a system that negatively hits third party, the response shouldn’t be the equivalent of sucks to suck that’s a you problem, your implementation should either respond zero or null, any other value and you are just being abusive and hostile as a program
lol FediDB isn’t a crawler, though. It makes API calls.
They do have a dedicated “Crawler” page.
And they do mention there that they use a website crawler for their Developer Tools and Network features.
Maybe the definition of the term “crawler” has changed but crawling used to mean downloading a web page, parsing the links and then downloading all those links, parsing those pages, etc etc until the whole site has been downloaded. If there were links going to other sites found in that corpus then the same process repeats for those. Obviously this could cause heavy load, hence robots.txt.
Fedidb isn’t doing anything like that so I’m a bit bemused by this whole thing.
