big hate to all users who post photos as PNG without thinking about each one of them weighing a gazillion megabytes

wanna help out your fedi admin / help your fedi instance / fedi as a whole? COMPRESS YOUR SHIT

that “just 1MB” turns into “just 1GB” if just 1k instances fetch it. can be more w/o mediaproxy, or for large profiles

(and full-res pictures are rarely “just 1MB”, it’s usually a couple megs in size. size baloons accordingly)

@domi Is it worth sending a feature request to some of the Fedi software to make it warn on large file upload? I have to actively work to make sure I don't do it - even on slow Australian networks it doesn't take long to upload large images.

@voltagex it’s worth educating people who misuse the software

@domi you got it

@trucy you uploaded it as a png, fuck off

@domi How would someone who isn't tech savvy realise they're misusing it?

@domi alright for my defence I just did a copy/paste. I just jpegged it

@trucy that’s not a defense!!!! that’s a part of the problem!

@domi i feel frontends should show attachment size and mime types, and not immediately upload when you drag-and-drop/paste/add attachment
would be generally good to have —helene

@domi i think instances should run optimizers like oxipng on uploaded images, and also deliver them as lossless JPEGXL when possible. both approaches easily save 20-60% space because most PNG libraries don't optimize very well and PNG itself is not the best-compressing format.

@domi
Based on personal experience, I guess that some fraction of photos posted as PNGs are the result of people pasting the clipboard into the mastodon web UI.

@domi imo it's the instance softwares job to pick a decent format/compression tradeoff. Also 1 MB is pretty okay for photos? I remember a time when Mastodon jpeg'd all my photos to death which was *so* frustrating, because it made photos I was proud of look nothing like what I created.

I don't believe that bandwidth for media assets is the scaling issue when it comes to running instances today. If it would take a minute or two to federate my hi-res post via my Germany-Grade home connection, then so be it. But most multi-user servers live in cozy datacenter racks anyways. The usual scaling problem is the thundering herd and exhaustion of sockets/workers/database connections/cpu, many of which are caused by ancient OS defaults.

That being said, I, as an instance admin, have different priorities (and resources?) as you. That's good, it's the beauty of federation.

@fionafokus FWIW I too, am an instance admin. 1MB may not be an edge case extreme enough; “couple” megs for a photo of a mediocre resolution (1200px?) is perhaps more fitting

and, to wit, despite me being an instance admin, it’s not my server that’s suffering. In the last couple of weeks I’ve seen a bunch of remote instances serve me media at a whopping pace of tens of kilobytes per second, which feels like my german internet is at fault, until I realize that my downlink is hundreds of times faster than that. My suspicions have been towards a bad non-CDNed file storage on the other side, leading to a DDoS-like scenario from all the fedi server trying to fetch the picture (and saturating the bandwidth in the process) but heck, maybe it is poor OS defaults

@filmroellchen @domi JPEG XL is unfortunately not supported by most browsers (Chromium dropped support and Firefox requires an about:config option to enable it that does not seem to work on ESR)

@noisytoot @filmroellchen only because google has an effective monopoly on the browser market!