To migrate your local storage and settings to another machine, simply copy the 'storage' and 'config' folders over to the project directory before running.

To install custom css themes, add them to the /themes folder. Currently you have to either restart or change the theme once to get the list to update.

You can configure how long you want post hashes and corresponding post quotes to appear in config/config.json. The default is 16 characters, truncating the 64 character post hash to reduce visual clutter.

tenor.gif, 22.20 kB

Can anybody familiar with Peerbit explain the following: so if hosting of text and media is distributed across users (is it? when poster goes offline his content is gone too or is there some data replication?) - how does boards moderation functions? When I delete some post, it gets permanently hidden only on my instance of Gladden but still available for other instances, right? So if everyone would run own Gladden to access the network, it would be practically ungoverned? I'm sure I misunderstood something but not sure what

>>99c4c78c

>when poster goes offline his content is gone too or is there some data replication?

Yep there's data replication, once the rest of the swarm gets the data it will generally persist. However in our situation each board is its own topic and will only be seeded by others also seeding those boards. Files are shared in a single board-agnostic pool.

>When I delete some post, it gets permanently hidden only on my instance of Gladden but still available for other instances, right?

Yeah, when you delete something you will no longer seed it to other people, but others will continue to seed it. When you delete something you also broadcast a "deletion operation" to the swarm. However, this will only be accepted by others who consider your node to be a moderator. So with the default configuration this means it will only be deleted from your perspective. If everyone else in the swarm also deletes something however, in practice you won't receive the deleted data.

>So if everyone would run own Gladden to access the network, it would be practically ungoverned?

By default yes, unless you subscribe to a moderator to delete things for you, or you delete things yourself.

Thank you for the explanation.

>If everyone else in the swarm also deletes something however, in practice you won't receive the deleted data.

How do client choose peer to load data from? Is it original seeder, or closest peer, or just random peer?

>>112d76f7 When we attempt the query the database (eg. to find the most recent threads to build the index page), there are two types of queries, "local" and "remote".

>local

Query our own node to find the data

>remote

Query other notes in the swarm who we expect to have the data. In Gladden, we do both types of queries, first we check local but if we for some reason are not up to date on the current state of things, then we also ask other peers. There is also two keys concepts "sharding" and "replication fraction. Peerbit has dynamic allocation of the database into "shards" by default. Each piece of data (a post, a chunk of a file) is considered to have a different location in a "content space", each node replicates a certain fraction of this content space. So you will first check if your node should have copy of the data given it's location in the content space, and then if not, it asks other peers (who are also broadcasting to the swarm which parts of the content space they are replicating). In practice this doesn't work so well for our purposes of an anonymous imageboard with heterogeneous access-control-logic (eg. perspective-dependent moderation), and so gladden nodes are set to replicate the entire content space. I think peerbit still needs some additional features to make this more feasible for us to actually (soft)-re-neable sharding. This link has a good explanation with visuals too: https://peerbit.org/#/topics/sharding/sharding?id=future-work-and-improvements

Also I just got that after last peer who saw certain thread goes offline, thread practically becomes inaccessible. So centralized archive would still be handy. And normal webarchive could serve as such, if Gladden had public node (basically nginx reverse-proxying node). Except random people would delete content from it lol as if that's their own node (fixable I guess) + abuses for bad content possible (spin up .onion domain) + ddoses + spam. What are you planning to do with spam though? And wipes. Captchas unlikely would help imo

>>5315b41d ❤️

>Also I just got that after last peer who saw certain thread goes offline, thread practically becomes inaccessible.

Ok I wrote that before updating thread to read last response. So if all peers copy all info this should not be possible, but eventually will be as the total size of data grows and replication level will be lowered.

>>8c1ea5be

>Also I just got that after last peer who saw certain thread goes offline

Could you explain this more?

>if Gladden had public node (basically nginx reverse-proxying node).

Yeah I also am planning to have a "gateway mode" where you can set permissions for non-localhost users, that way you can disable moderation, adding or removing boards, (or even disable posting files if you want), like you said yeah.

>>69aef6cb There is a configurable setting for "min replicas" so there should be some redundancy for a given piece of data. If min replicas is set to 5, then the sharding algorithm will adjust to try to ensure 5 peers have a copy of that data, so that any 1 peer going offline won't cause the data to become inaccessible.

>Could you explain this more?

I mean last peer who fetched data from thread in situation with replication limited only to the data peer ever accessed with search queries aka by naturally browsing - someday all the copies will be lost if all peers will permanently go offline. So it would be great to have centralized archive somewhere or just a few super-seeds which would function as backup servers practically and try to grab all the content to replicate it.

By the way, having gateway version could lower potential amount of peers = of available space for replicas, as it's far more convenient to use web version (also in that case, user would trade smaller attack surface (running node + browser vs just browser) for potential logs + more attack actors (gateway owner and everybody with access to gateway server)) + optionally privacy if gateway is under reverse proxy waf like cloudflare. Or enlarge it lol if gateway would popularize the node version somehow.

>>0b0779c7 As far as I know there are at least two "always online" seedbox nodes at present. But they might not necessarily be seeding all boards, though since files are currently in a shared pool they will. >>74837bc1 Yeah lol... exactly. Opening a gateway is another level of pandoras box, and double-edged sword. But not something really preventable in practice so something we have to accept the existence of. I believe one peer actually is running a gateway but I don't know the address.

>>8c1ea5be

>What are you planning to do with spam though?

This is the elephant in the room and the main architectural challenge of this project. If we want to be permission-less then from what I have reasoned out so far, the ideal solution seems to be "filters" where posts are checked with various means like regex, locally running AI models, various logical rules, etc, to detect them as spam. But there is a potential arms race here. Sure you can detect "asgdfdw4r4rfsfr" style spam simply maybe even with regex, but if you have ChatGPT generating paragraphs and posting them over and over then you also need an AI running to "turing test" those posts and.. what is the end result of that lol? AI's chatting with each other and trying to pass each other's filters? Some sort of sheathed-sword detante? It's an open question :) On the other hand, people can also build more highly permissioned systems on top of this base, which interact with various external systems, including captcha, things like that. for example already there is some form of stock validation to check that a Post Subject isn't 10 MB of text, to check that the deleter of a post is in the list of trusted moderators, and things along those lines. To avoid the asymmmetry of one single spam post having to be detected by 100 different peers redundantly, the task of managing spam control could be offloaded to a trusted peer who detects and deletes spam... Much like organisms have varying levels of "immune system" (eg. skin, bark, white blood cells, antiobiotic compounds, liver, plate armor, ideology) to protect them from things like germs, pests, weather conditions, arrows, propaganda, etc, seems like entities in a p2p system might need such things as well.

>>65e7f5c2 I believe it is possible to detect AI generated texts with other AI or even (under)paid human accessors, to some extent. After AI will take thread context to generate appropriate wipe text virtually same as human, only hidden boards could be a wipe-safe harbor - unless all board names will be bruteforced to find survivors. Computational captcha could increase attack cost, but if actor is willing to use AI to generate wipe text, that would not be effective. There is no final solution, but at least we can harden our armor and prepare good defense beforehand. By which I mean example code for automated moderation (fetching all new posts for board, process each one by one, do something, send API request to delete post optionally), AI or heuristics or regexp, any automated algorithm would require that.

Btw, if moderator starts to delete excessive amount of posts (due to their political position or anything), and also delete posts with information that moderator X is now deleting more - there would be no way for user to know about what they're missing, except randomly disable moderation subscriptions. I mean option to still download deleted data but hide it by default + show on hover would help with censorship discovery. Or maybe that situation will never happen and I'm making it up. But in a future it can be useful to know what exactly is being deleted and from which channel deletion is broadcasted

>>cd3994cc ah yea that SURE would be usable to detect if antispam is deleting something it should not

By the way, back in the day I myself wanted to create true p2p imageboard, I was thinking that someone surely will post CP, then some peer will download it while seeing index page or catalog, and share it because that's just how p2p works, then FBI opens CP thread, sees peers IP and basically farms frags indefinitely as long as someone uses it. So, I thought, I had to integrate anonymization system to guarantee safety and that no ones goes to jail because of my flawed software. If it will be tor, then whole thing has to be executable program, so I rejected that as doomed to be dead in modern world of websites (I haven't thought of gateways...). Then I encountered WASM client for Nym mixnet, which could be used for p2p communication anonymously within just the browser, solving the adoption issue. And then I gave up. So I'm glad that someone actually did it, even with anonymity delegated to users, and who knows what gladden can become. Even street walls are governed space of communication. Gladden can be governed only by advanced wipers lol

Also, I think that, maybe in a single year, a lot of people on the internet will talk to AI without knowing it's AI. LLMs are already human-plausible enough with some preprompt, and only limited context window, which forces LLM to forget earlier context and drive thought away - is indicating inhuman source of the text. Context size gets rapidly increased though, and eventually, half of social media will be talking to AI web agents promoting some agenda to humans. But imageboards are first on this way, because they profile-less and because they have less antifraud to bypass (I mean it's easier to be blocked on social media for using automation or multiaccounting). Not in a form of wipe, but in a form of meaningful conversation. And maybe I personally am ready to accept that, after all I need to have a good insightful conversation, and there's little difference between alive low-quality poster and AI trying it's best on the topic. Except when AI is creating threads and seeding discussion on themes it's programmed to, that would be an awful end to free speech

>>85b487e8

>only hidden boards could be a wipe-safe harbor - unless all board names will be bruteforced to find survivors.

It's also possible to implement passwords for boards. Boards themselves are barebones right now (just being an address) actually the boards can also be fleshed out with different configurable properties, with configuration settings subject to the same type of moderation scheme as post deletion currently is. Users are free to alter their client or access control logic for this as well, so if you have lots of configuration divergence it could lead to "canonic splintering" of a given board's state type of thing, Even now the semi-hardcoded (in config/validation.json) max filesize is 4MB, but this could be made a board-specific setting (maybe would make more sense if each board has its own files pool in that case). However if you edit this yourself, you will be able to post larger files than 4MB, but other users won't consider those files valid unless they also increase their own allowed limit. >>cd3994cc

>But in a future it can be useful to know what exactly is being deleted and from which channel deletion is broadcasted

This can be logged and possibly stored somewhere. Under the hood Peerbit does mark items that have been explicitly deleted and doesn't just forget about them (there are two things it can do.. "delete" (erase and it was deleted), and "prune" (just erase the data)), but this isn't currently hooked up to the querying logic or anything. A way around that is to store it ourselves or at least log it in the console, or it's also possible to add this custom behavior into the Peerbit aspect of the code itself. >>40183b23 I'm gladdened. Hopefully we can figure out some sort of more simple anonymity solution that works nicely for our case as well so it's easier for people to do.

>Gladden can be governed only by advanced wipers lol

hehe :) >>86a58f5e Yeah... it's also definitely something it seems we have to accept the existence of, and in some cases I think this has already happened. But in a way it's just another form of life sort of thing, humans never truly existed in a vaccum in the sense of, we've always been influenced by and part of and embedded in nature, AI's or self-sustaining automated conversation/digital information transfer is maybe another network to be woven into the biosphere. I sometimes think about how much meaning is being transmitted by birds chirping and insects buzzing, maybe not as intuitively comprehensible to humans as the ones making the sounds themselves.

>>cc7a3ae5

>It's also possible to implement passwords for boards

Long enough board names are essentially passwords imo

>if you have lots of configuration divergence it could lead to "canonic splintering" of a given board's state type of thing

Anarchic standards generation, that would be cool

>but other users won't consider those files valid unless they also increase their own allowed limit

Maybe show file metadata even if file fails vaildation, so people would have reason to enlarge max filesize in their config

>Hopefully we can figure out some sort of more simple anonymity solution that works nicely for our case as well so it's easier for people to do.

I think just --proxy parameter for node would be good enough (nym and tor have local socks5 proxy ports) + people can use whatever connection chaining or mixnet system they're using for visiting gateways. Besides, data replication already messes with identification of original poster. By the way, is peerbit vulnerable for Sybil p2p attack (which is peer topology isolation)? I mean how isolation-proof it's peer discovery works, we have some hardcoded kickstart-peers to start discovery?

>I sometimes think about how much meaning is being transmitted by birds chirping

They could have some sort of consciousness https://www.cell.com/trends/cognitive-sciences/fulltext/S1364-6613(20)30192-3 + https://fcmconference.org/img/CambridgeDeclarationOnConsciousness.pdf Anyway, AI is directly affecting intellectual side of our existence, which was something unprecedented.

>>b68a576d

>I think just --proxy parameter for node would be good enough (nym and tor have local socks5 proxy ports) + people can use whatever connection chaining or mixnet system they're using for visiting gateways.

This would be nice. I'm looking into it recently, there should be some sort of setting in the libp2p node to dial things through a given proxy port...

>I mean how isolation-proof it's peer discovery works, we have some hardcoded kickstart-peers to start discovery?

Yeah there are public bootstrap nodes that we use currently. There is also a "direct dial" option on the /home page but its a little tricky to use and you need to have the ports opened/forwarded properly for the node being dialed. But you could use this to connect to "non-public" bootstrap nodes as well and possibly have subnetworks of peers.