Hey everyone,
This isn’t an announcement, just wanted peoples thoughts on this.
I think everyone knows searching the fediverse can be better. Googling doesn’t work too well, etc. So I wanted to do my part and help out.
Indexing all posts, etc is quite a lot to handle, so I wanted to start small and just focus on video search. I’ve started indexing videos from Peertube and other video websites. (Even YouTube but this could be removed to just focus on independent sites)
I know Peertube has their own search engine for videos. I will be reaching out to them. Compared to my site I’m planning it’ll have other video sources and be easier to use.
So that leads to feedback from you guys.
- What do you think about indexing videos posted on the fediverse and other independent platforms?
- Are there similar services?
- Am I just wasting my time?
I found FediSearch, and also this post basically saying that a fediverse search engine would just be used as a tool by trolls.
It’s worth noting that since FedSearch, Mastodon has actually natively implemented opt-in search on posts.
That’s a good point. But those people can be banned? I guess Reddit handles this by moderation and archiving old posts.
Yes, but moderation teams on the fediverse are very small, and by nature of it, can make hundreds of account of different servers all trailing that would need to be individually sought out and banned.
It is a game of cat & 100 mice
People will take the harassment off site especially if they are dedicated enough or use it to scrape for potential personal info to publicly release.
How is that different from Reddit? If trolls want to search and scrape and find information on people, they’re going to. You can’t put your information on the open Internet and not appreciate there’s always a danger of that.
There is more effort barrier if the trolls have to do all the scraping and sorting themselves than just popping a term that is a right wing lightning rod into search and getting a list of targets.
That post wasn’t claiming that a search engine would only be used by trolls; it was explaining that they shut down their project because a chunk of the fediverse thinks that and complain about any search engine projects. Discoverability is one of the network’s biggest challenges and a search engine could really help with that.
Yes, not only used by trolls, but would be a tool that could be leveraged by trolls. And I think the fediverse makes it easier to establish instances for marginalized groups, but also has more admins that just don’t want trolls because nobody here is making $ off them like the corporate socials are. I think if adding search that is going to try and vacuum up everyone’s posts in the fediverse and make them easily sortable/targetable without instance admins permission, then that isn’t cool. If someone is running a general instance that covers nothing that a troll could latch onto and wants the instance catalogued and searchable then that’s fine by me. I don’t think boys should be doing that to the fediverse as a whole without admin permission though.
I don’t think an admin’s permission has anything to do with it. If you post publicly on the fediverse, your posts are public. You should have the option to opt out of any indexing (just like you do for the rest of the open web). But saying its ok for you to read this post if it happens to come across your feed but you shouldn’t be allowed to find it via a search is ridiculous. Users get to make the choice with each post whether its public or not, but they don’t get to control how people consume those public posts.
Reading a post and having a bot thrashing a server indexing everything are 2 different things. If a user used the site like that they would be throttled and if repeated afterwards, banned. It is also one thing to read/interact with a site as that adds value to the site as a whole. A bot that just mass hits links cataloging everything is just a strain on the server an Admin needs to support, with no upside for the instance, as it’s a bot ingesting and no real interaction actually took place.
and having a bot thrashing a server indexing everything
This is a completely separate argument and one that we already have mechanisms for. Servers can use status codes and headers to warn about rate limits and block offenders.
It is also one thing to read/interact with a site as that adds value to the site as a whole
A search index adds value as well; that’s why this keeps coming up. And, again, there are existing mechanisms to handle this. A
robots.txtfile can indicate you don’t want to be crawled and offenders can be IP blockedShould a dedicated search not use/index ActivityPub instead of the html interface?
If so, instances can simply defederate from search engine instances. So the point you are trying to make still holds.
Well, please make sure it respects post privacy at least but also realize that on the microblogging side of the fediverse, they may not take kindly to this prospect at all. People who start these kinds of projects are often harassed or at least receive passive hostility. Making it opt in instead out of opt out in some capacity is best.
I disagree. Post privacy sure, but the internet is by definition public. Anything you put out there can be used for pretty much everything, the original rules of the internet apply. I’d be happy to see an easy opt out on the engine to remove yourself, but if everything is opt in it’ll never get off the ground.
deleted by creator
again it’s not going to servers and scraping data, it would be sitting somewhere receiving public data that is pushed out. There’s no malicious getting around privacy settings, if it’s pushed out then it’s free game. I agree about post privacy, but again activitypub already takes care of that
As the fediverse is almost exclusively run by volunteers that are paying server bills and being admins, I could see some larger instances not taking kindly to this, especially depending on how much stress it would be putting on some already at capacity servers.
Ideally, OP’s crawlers will just come from their own instance that other instance owners can defederate from if they want to opt out.
Yeah that would be the case.
That’s a good idea. Listen to public data being broadcasted out, then you aren’t worrying people with scraping or anything. It would only be from go live onward, but you would just be listening to the protocol.
For that to happen on an instance organically users would need to visit all these instances/communities. To speed that up you would need a bot to do all.that “seeding” for you. That brings you full circle to the server resources on bigger instances.
This seems like an opt-in, not an opt-out activity.
How much bandwidth do you suppose a crawler would use? I’d guess very little
I was thinking more in terms of resources (number of spider threads X posts/communities/users being indexed) that would be now dedicated to a bot, not so much network traffic that is probably tiny if not downloading images.
Right, it would be an initial hit but if the bot was properly built it wouldn’t need to do full reindexing very often. I’m no expert but I think it could be done in a way that there is no noticeable spike in traffic or anything
That’s the thing, it would need to be done in chunks and have its revisits scheduled if you want to do a complete indexing of an instance. And for a large instance that’s a lot of DB thrashing if you aren’t spacing that out, or just sampling like “top 10 posts” or something, but that kind of data is going to make a useless search engine depending on the goal of the search engine. If you wanted to just catalog the daily top posts of the fediverse that might work, but if you want to catalog everything it’s going to take a lot of resources and a long time to make sure you’re not hammering people’s servers.
It will be very little if not downloading full html pages.
That’s not how the fediverse functions and approaching it that way is a problem waiting to happen. I’m stating so as a warning to be mindful of the culture of the way the fediverse itself functions. This is not Reddit, we share the fediverse with other software with different uses and features and we need to be mindful of that especially when building these kinds of tools. Making it opt out not only places a burden on smaller instances but presents a potential harassment risk for instances with vulnerable people on other fediverse platforms. As well, it is contrary to the entire way specific other activitypub instances operate. The fediverse is like a city we share with others, if Lemmy is not mindful of that city’s culture then people will promptly give them the boot.
I’m not saying user by user opt in either, but instance by instance. Lemmy needs a tool of archiving especially. There is already cultural clashes I see occurring with the rest of the fediverse. Post like these of potential tools when it seems like the creator doesn’t know the messy history behind previous projects like them in the fediverse make me fearful of the clashes coming to fruition.
Well that’s why I’m asking for input. And I won’t launch this on every instance without letting them know. Baby steps.
Mastodon since 4.2 version supports allows its users to opt into appearing into search results. Just respect this flag with Mastodon users, and you will be fine, IMHO
My matrix is open if you want/are actually interested in doing this in a way that won’t make the rest of the fediverse flip shit. I support this tools creation especially for lemmy, but if it isn’t done the right way it’ll be received poorly. Making it behave differently on lemmy compared to other software as well might be an idea too.
But ActivityPub already publishes all of the data out. I don’t think this is going out to servers asking for data, it’s listening to public data being broadcasted out. If people are broadcasting over activitypub then they’re okay with it being shared.
If they don’t want it shared then they don’t have to publish ActivityPub to anyone. They can defederate from the search federation. Those tools already exist.
That’s not how the fediverse functions
That is how the fediverse functions. Instances send posts to anyone who request it, unless a block is in place. ActivityPub is opt-out and the web has always worked this way.
be mindful of the culture
There is no “the culture” on the fediverse. Your talking about a subgroup, which has a different opinion from other subgroups. They don’t get to define “culture” on the fediverse.
A good search engine would be quite important. One thing that annoyed me back on the site that should not be named was that their search engine was completely useless - It was not even capable to find posts where I entered verbatim text of.
Having a good search engine that can actually find a post I was looking for would be a major plus for the fediverse.
Yep, the idea is to simulate the type of results you get from Google. People trust Lemmy answers more than spam sites now a days.
Why wouldnt people want do have search engine? Without it Fediverse stands no chance against non-free internet. Everything posted here would be much more valuable if it was searchable. Now comment posted once is viewed only until post gets less popular. Any other site of this kind displays answers decades old. Privacy isnt issue as everything posted here is available to everyone on internet.
I mean they are posting on the public internet, they should know that it can be read by anyone. I like the idea of users opting out.
If you give users the choice to opt out, all the privacy-focussing communities won’t be searchable.
What if someone who opts out posts a comment and someone who opts in answers?
The Fediverse is public, a search engine doesn’t show anything that isn’t already open for anyone to see.Good point. They should know they are making public comments. If you want it private then send a private message.
You should federate the search engine so that folks can defed from the search as desired.
But then we would need a search engine for all the search engines…
Everything you post on the Fediverse is public.
If you don’t want to show up in an internet search, post your stuff on your private server and only give access to the people you want to invite.Yep
Is this something you can point yacy at?
I heard it’s not optimized well but I’ll take a look at it.
I’d be interested to know what “not optimized” means.
I heard it will send out requests as fast as possible essentially creating a DOS attack but I could be wrong. Also the UI can use some improvement.
There is already Sepiasearch specifically for Peertube.
It’s limited to only Peertube and it’s not the most intuitive. I want to work with them on expanding this.
I love the idea, especially from a technical standpoint!
How big is the fediverse today? How many posts are there? What kind of algorithms atmre you using to store the results? Do you scan sites and then their connected sites or do you have a premade list?
More technical information please 😊!
The fediverse is a few thousand servers, from Mastodon, Lemmy, etc. Can’t say the amount of posts but there are a lot.
So on the more technical side, I plan on using a light weight fast search engine called Sonic (It’s written in rust). I have already used it in other projects and it can handle billions of messages / posts. But it has a cost it doesn’t have faceted search, like for example if you want to exclude certain texts from the results. I think this is a fair trade off. The other solution would be to use something more mature like ElasticSearch but it’ll be expensive (I’m assuming not much money will be made from this and I’m talking about donations)
For scanning sites there are premade lists to start with and it’ll be possible to scan new sites from other instances if found. So a bit of both.
I support bigger picture. Rather than an independent site, wouldn’t it be more practical to work with current fediverse app developers for lemmy, mastodon, etc to integrate search engine within the app?
I’m reaching out to see their thoughts. But there are limitations to what they can index.
The fedi elite will snipe you from a bell tower.
I don’t know anything about the technical side of this. But I would (possibly naively) think that it would be simpler to have a filter that you could automatically apply to sift bog-standard search engine results for Fediverse instances? Like adding “site:uk” to the end of a normal search, except that your filter term would check a list of Fediverse instances to return the relevant results.
And make it an app/add-on so that people can use it with their usual search strategies.
People been doing that now but it comes with limitations. Thanks for the input.
Just, for the love of god, make sure to make it opt-in. Don’t scan people’s posts without their consent.
Also, if you’re going with this, make sure to respect people’s requests for removal, e.g. deleting their posts from the engine when they delete them from their instances. Otherwise you’d get in trouble with the EU regarding GDPR…
Your public posts are not protected from GDPR from indexing…
