Finding other users with PODs

Hi! I’m pretty new here, but I was wondering: How do users with PODs find other users? For example, on YouTube, it might suggest another video based on the one you just watched, and ones you have watched in the past (among other things). The only reason it can do that is that YouTube knows the location of every video on its site, and so can find a match. With PODs, I can envision a situation where every user has videos on some random site, so unless there is a massive web crawl to find videos, it would be impossible to find them. One solution I thought of was to have a registry of all users, but that wouldn’t help much with decentralization. This feature is really nice for learning new things and is one thing that big companies have an edge on. Do any of you know a possible solution to this?

2 Likes

I don’t know if there is a good decentralized approach to suggesting new content. I think discovering relevant without some kind of centralized suggestion service wouldn’t work well (but there may be a way I don’t know of yet). I am currently tending to platforms like peertube and mastodon, maybe with a bit more Solid style.

Peertube consists of many servers (“peertube instances”) which communicate with each other about new videos. Each of these instances provides video upload, search, suggestions and other features on its own. But to be able to work with a huge amount of data, the instances can follow other instances and retrieve meta data about new uploads. Mastodon is a social network like Twitter which is also based on this federated instances model (and they actually both use the AcitivityPub standard, so they could communicate with each other).

I think this already covers much of what solid aims to provide: One can choose where to host self made videos (ie similar to a solid pod you can choose on which peertube instance you upload it). I am not sure about peertube, but Mastodon is built several standards to ensure that other apps could use the same architecture. This means other apps could evolve without the need to get a big user base first (similar to what solid tries to achieve with linked data and pods). And everyone with enough resources like a company can create an instance of peertube/mastodon with customized settings itself (similar to solid pods).

I could imagine that videos could be stored in ones own pod instead of a peertube instance. So peertube only knows the metadata and links to your pod provider for watching the video. This would give you more control about the video, but I am not sure if I like this approach better (it seems much less reliable as pods could do whatever they want with the video after uploading the meta data to a peertube instance). Another way pods could be integrated is to store a user profile on your pod rather than the peertube instance. The watch history, likes, comments, etc could be stored there. All of these could be used to create a recommendation profile which is sent to a peertube instance to get recomendations for this profile. The advantage would be, that the peertube instance first of all doesn’t store my activity data and furthermore only gets to know an abstracted recommendation profile instead of all my activity.

So that’s my view on social platforms like youtube where the discovery of relevant content is pretty crucial. I’d highly appreciate it if someone wants to add more thoughts on this and shares another view on this.

4 Likes

The devs behind Beakerbrowser have also been seeking a solution for making a social graph out of in-browser content created by users. Recently they announced “Unwalled Garden” as a project to work on schemas similar to RSS for this purpose. Even if the content is decentralized, a subscription-based central aggregation of such would be handy for filtering and recommendation engines.

Asked why create a new schema and not use, for example, RDF, the lead dev (Paul) mentioned human readability and coding simplicity as reasons.

https://Unwalled.garden/docs/why-not-rdf

1 Like

IIRC one of Paul’s objections is that using RDF is hard, particularly wrt ontologies (undeniable I think). He favours using software to overcome the issues which ontologies are intended to solve (such as determining semantics) although I suspect his perspective is from that of most current web developers - so relatively simple application logic and data handling.

We had a discussion with him on Solid chat a while ago (one year ish maybe), of anyone is interested. I can dig it out of you can’t find it.

I’ve read his intro post about Unwalled garden and I’m afraid I don’t get it. It seems too simple to me, but I shall follow with interest.

Edit: reading the Why not RDF? link is useful for us I think. How could we make using RDF as simple and readable as Paul’s examples? I put the same question to the SAFE dev forum and was reminded that Josh has already had a go at precisely this:

2 Likes

Not sure I want a solution.

One of the things I like about Solid is that my friends and family are exactly that, friends and family.

I don’t want anything spamming me to say ‘Do you know such and such’. If I know them, I can add them. I don’t know anyone who has thousands of friends in real life, or thousands of people I want to have daily updates from, so why would I want that when I’m sat in front of my computer or on my phone? I don’t.

If I want to know what some politician thinks about something, I read an online newspaper. I don’t need to follow their every word every day.

If I want to know the POD address of friends, I’ll just ask them. Same with family members. Job done!

I definitely agree with you! I think that unless someone registers to be part of user recommendations (or any recommendations), they should be under the hood. I’m sure many people want to go both ways – some not detectable, and others looking for people.