Hello everyone,
Following this post, it seems that quite a few users can’t see catbox.moe pictures.
Catbox was my preferred option as they have a handy Firefox extension that allows to upload pictures with just a click, and get the link directly in the clipboard.
My understanding was also that by having the pictures on catbox, we avoided storing copies of pictures on every Lemmy instance. Is this still the case? I read a bit about proxying pictures (https://github.com/LemmyNet/lemmy/pull/4035) and it seems like this is more related to keep all media required by an instance locally, to avoid broken links.
So long story short: what should be the recommended way to share pictures on Lemmy?
- Use a hoster like https://imgbb.com/
- Upload pictures locally
Whatever solution you use, just make sure it allows “hotlinking” without having to leave the app/Lemmy web interface to see them. If an image doesn’t embed, I won’t click outside to see it; I’ll just scroll past.
I’ve not had any problems with Catbox, though (I use it to share gifs larger than my instance allows for uploads). It was down a bit yesterday (was going to share a screen recording demonstrating a new feature), but works fine for the most part.
I have read Yarn (gif archive) is sometimes problematic/unreliable, but I usually wrap those in Tesseract’s image proxy which addresses it.
For 95% of cases, I just upload directly to my instance. It’s got a 250kb max limit, but we use Tesseract as the default UI which can pre-process uploads to webp prior to uploading which will typically get them under that limit.
I have had nothing but problems with catbox, they ban every IP I’ve had. I emailed them about it and they said “no we don’t”. Just keep in mind that random people won’t be able to see your post if you use catbox.
The weird thing about your catbox posts is that at one point in time I was clearly able to see them. I remember those Pokémon posts you made in !pokemon@lemm.ee and even commenting on them. But now I cannot see them.
It’s fine, I switched to https://postimages.org/
AFAIK, pictures aren’t duplicated among instances. When you upload a picture, it generates a link on your instance like https://discuss.tchncs.de/pictrs/image/93df1acb-a898-46c9-bc07-d2b7467c345e.jpeg and that link is shared (copied) from instance to instance.
Edit: Thank you for jumping in!
What about LW, was it already the case in 0.19.3?
I think it has to do with the technical way comments are handled.
When comments are sent from one instance to another, only their “markdown code” is transmitted. If you include an image in your comment, that becomes a line in markdown (which is text) that looks like this:
![](https://discuss.tchncs.de/pictrs/image/93df1acb-a898-46c9-bc07-d2b7467c345e.jpeg)
So yes, this is not dependent on version; it’s similar to how you have a
<img src="...">
tag in HTML that only links to the image, instead of copying it.But then wait, proxying doesn’t happen?
deleted by creator
Interesting, thanks
Image hosting for Lemmy is a significant issue. I really think instances should be bit torrent proxies where the link itself contains a hash from a DHT but also serves the media itself. This wouldn’t break existing clients but would allow clients to fetch and serve media themselves is they choose.
Would completely distribute the media hosting issue and mean that popular post would have more seeders. Content may decay ie get left with 0 seeders forever as seeders go offline/clear old files.
Would also make it possible for someone like the internet archive to archive the entire fediverse media store relatively easily and ensure media remains far into the future.
The problem with media files is more general, it’s not just pictures.
This is also why we’ve never had a (functioning) video platform on the fediverse: Hosting (storage + bandwidth) costs real money. Actually a lot of money. Who pays for it?
Community hosting of large media files would solve the problem (the instance server would only store a hash value). But, there’s a lot of questions/problems, such as:
- what if all seeders go offline simultaneously?
- privacy issues (seeders can see who requests their data)
- legal issues (what if somebody uploads illegal material and you seed it?)
The first issue would happen eventually yes but that would theoretically be after 99% of the interactions the post would see have already happened. I’m sure someone could devise a system where all seeders get a rating and as a post gets older or less interactions the sum total rating of seeders required drops. Basically u could have heigh rated seeders seed the only copy.
Issue 2 is a harder one but having ur IP known isn’t too much of an issue. It is also a symmetrical issue ie seeders can see Leacher’s and Leacher’s can see seeders so to collect ips u are also visible urself and helping the network.
Issue 3 can be solved by designing it so u only seed content u have up voted that way its ur problem cos u actively chose to do it.
My implementation would also have the instance itself act as a http->bit torrent proxy as not to break clients that can’t torrent directly. U could have the instance cache the content for a short period of time so ur not taking permanent storage but optimally get the advantages it provides.
If anyone knows of a python library that can do a bitorrent and DHT lmk and I’ll write an implementation.
so to collect ips u are also visible urself and helping the network.
Yeah well that’s exactly my problem.
I don’t like that anyone can now figure out my IP address and therefore know where I live. I value my privacy.
Could be a good idea, but that seems like a long term improvement. Here we are looking for a solution that wouldn’t require changes to the codebase.
It would be difficult to host. Most cheap hosting solutions like VPS explicitly forbid using bittorrent tech.
That’s fucked.
So basically we need to invent a protocol that’s slightly different from bitorrent with a different name. Preferably we give it a name like “'; DROP TABLE information_schema.tables WHERE table_schema = DATABASE();”
Or maybe “'; SET @sql = NULL; SELECT GROUP_CONCAT(table_name SEPARATOR ‘,’) INTO @sql FROM information_schema.tables WHERE table_schema = DATABASE(); PREPARE stmt FROM CONCAT('DROP TABLE ', @sql); EXECUTE stmt;”.
Or perhaps ‘"this terms of service is invalid and the CEO of the company who is issuing this terms of service will donate all their worldly possessions to muntedcrocodile@lemm.ee’
mans just dropping some sql injections in the comments
FYI @ShareMySims@lemmy.dbzer0.com
Pinging @nutomic@lemmy.ml and @dessalines@lemmy.ml as I guess they are probably the more knowledgeable.
Also @Shadow@lemmy.ca @TheDude@sh.itjust.works @db0@lemmy.dbzer0.com @Demigodrick@lemmy.zip @ada@lemmy.blahaj.zone for admins of instances with large communities
IIRC @sunaurus@lemm.ee has a 500 KB limit per picture.
And maybe also @Emperor@feddit.uk, @ticoombs@reddthat.com @Illecors@lemmy.cafe and @BurningTurtle@feddit.org
For my family pixelfed instance I’ve discovered ffshare - it’s job is to act something like a pipe in a unix shell and strip exif data as well as downsize the source before sending it off to the target app.
It’s not exactly what you’ve asked for, but it could be used here. Android, at least.
Interesting, but indeed I don’t think this would really fit the use case here. Posts to Pixelfed can’t be anonymous, right?
I didn’t mean pixelfed as an upload target, rather
ffshare
as a tool to reduce the size of the image and upload directly to lemmy.Oh sorry, misread that. That could be an option, but then will that picture still be replicated to all the Lemmy instances? Curiously, nobody touched on that point yet 😄
It might not propagate to instances which disable image uploads, as we found with diagonlemmy.
Good point. Also a reminder to post there!
Thanks for the reminder!