I’ve dug into this somewhat. It’s something to do with caching images via the JS Service Worker in the UI. On first load, everything works. Refreshing then causes the images to break until you do a “hard” refresh to bypass the cache. Then they work until you do a regular refresh and the images “break” again.
As far as why it’s only startrek.website and no other instance (so far, anyway), I haven’t the slightest idea. As far as I’ve been able to examine the HTTP responses for images, there’s nothing weird going on. Perhaps the .website TLD causes confusion in the service worker? That’s just a guess, and is my only working theory. Unfortunately, I’ve only found two other Lemmy instances that use a .website TLD, and neither have any local communities that have images, so I’m unable to test that.
I use Photon as my main UI, and it doesn’t do any of that service worker image caching (just uses the browser cache), and it’s never experienced the issue. Funny enough, one release of Photon did implement that kind of caching via service worker, and startrek.website images broke immediately. Removing that part of the SW fixed it (the dev has also since removed that from the SW).
So I don’t think anything is misconfigured on startrek.website, but there is some kind of funny behavior when viewing its images from a federated instance that uses service worker caching.
Workaround for Users:
Chrome: In your Lemmy tab, open developer tools and go to the Application tab. Click “Service Workers” in the left tree and in the right tree, check the box “Bypass for network” and then refresh. Images hosted from startrek.website should work.
Firefox: Should be the same, but I seem to be missing the option in 113.0.2 :hmm:
Apparently that no longer works in the 0.18.4 UI. :sigh: Maybe ask your home instance admins to offer additional UI options like Photon or Alexandrite :shrug:
it doesn’t do any of that service worker image caching (just uses the browser cache), and it’s never experienced the issue. Funny enough, one release of Photon did implement that kind of caching via service worker, and startrek.website images broke immediately. Removing that part of the SW fixed it (the dev has also since removed that from the SW).
So I don’t think anything is misconfigured on startrek.website, but the
Thanks, that was really useful. Also explains, why Jerboa seems to work for me, but the web frontend looks broken.
I’ve dug into this somewhat. It’s something to do with caching images via the JS Service Worker in the UI. On first load, everything works. Refreshing then causes the images to break until you do a “hard” refresh to bypass the cache. Then they work until you do a regular refresh and the images “break” again.
As far as why it’s only startrek.website and no other instance (so far, anyway), I haven’t the slightest idea. As far as I’ve been able to examine the HTTP responses for images, there’s nothing weird going on. Perhaps the .website TLD causes confusion in the service worker? That’s just a guess, and is my only working theory. Unfortunately, I’ve only found two other Lemmy instances that use a .website TLD, and neither have any local communities that have images, so I’m unable to test that.
I use Photon as my main UI, and it doesn’t do any of that service worker image caching (just uses the browser cache), and it’s never experienced the issue. Funny enough, one release of Photon did implement that kind of caching via service worker, and startrek.website images broke immediately. Removing that part of the SW fixed it (the dev has also since removed that from the SW).
So I don’t think anything is misconfigured on startrek.website, but there is some kind of funny behavior when viewing its images from a federated instance that uses service worker caching.
Workaround for Users:
Chrome: In your Lemmy tab, open developer tools and go to the Application tab. Click “Service Workers” in the left tree and in the right tree, check the box “Bypass for network” and then refresh. Images hosted from startrek.website should work.Firefox: Should be the same, but I seem to be missing the option in 113.0.2 :hmm:Apparently that no longer works in the 0.18.4 UI. :sigh: Maybe ask your home instance admins to offer additional UI options like Photon or Alexandrite :shrug:
Thanks, that was really useful. Also explains, why Jerboa seems to work for me, but the web frontend looks broken.
Good to know that I broke image caching.
(this is serious)
I couldn’t figure out how to disable it and the cache sizes kept going into the gigabytes, but I somehow broke it recently and caches are sane now.
If you want a standard way to disable it, I made this patch for lemmy which makes it a config flag: https://github.com/LemmyNet/lemmy/pull/3897