I’ve noticed very recently (last two days or so) that when I get a direct link to a specific comment, it won’t load anything under certain circumstances.
Following the link from the button on the comment, either from the profile or from viewing other comments on the post, will work, but using the link as the URL for a post in a different community, or opening a new tab and pasting the link in, don’t.
In the former situation, the UI of the post I gave it as the URL for just freezes until I press the back button, still showing the post. In the latter, Lemmy doesn’t load at all.
I’m using Librewolf 116.0, based on Firefox 116.0.
I can confirm the problem, it’s been gong on all week. It really impacts anyone on another instance with a link, they will fail.
As I understand the situation, Lemmy.world has been suffering from performance problems and certain comment links were being attacked by distributed clients. So they basically have firewalled /comment links for everyone (I assume using nginx based on behavior, or maybe the front-end cloud distributor).
Personally I’m interested to know which specific comment links cause the PostgreSQL performance problems as I’m trying to track down and fix those issues. But I haven’t seen anyone detail which specific post/comment threads cause the problems… I’ve just seen the developers reduce loading to 50 and 300 without creating testing scripts to reproduce the issue for other developers to study.
I’m hoping lemmy.world can implement a less-drastic solution than 100% block of comment links from non-local referral origin… such as a rate limit on those links of 3 per 5 seconds or something low like that. Anyway, I hope you are having a good weekend.
Ah, that’s good to know. For me the problem is that most of my non-lurking activity on Lemmy is transcribing posts to improve accessibility; I post the URLs to an archive to make all the transcriptions easy to find for those who need them, and I know there are at least a few people from rblind subscribed to it, but with this going on the archive is essentially useless.
Is that what they said? Often with single page app setups you have unrecognized routes return the main index.html with JS files etc. by default, and then detect the URL path and feed that into the SPA router. From checking in inspector it looked like it was failing to authenticate when loading the route via XHR, like some kind of on-page context was missing that would be required for that request, i.e., it’s not redirecting to the index page and is maybe rewriting straight to the API endpoint. Not sure myself though.