Media PDS / Service

[iame.li]

Thanks @thisismissem.social for doing the writeup! I think this is a really solid overview of what a sophisticated and feature-rich VOD atproto server could look like. I’ve played around with the term “VPDS” before (the V is for video), though I think it’s an open question whether this kind of media server would be the source of truth for the video content (like a PDS) or just a supplemental set of tools (like Bluesky’s existing video upload infra.)

To me, the most interesting and urgent piece of the design: what does the VOD primitive record look like? If you’re looking at a big list of videos on “ATProto YouTube”, what are you looking at? If video-on-atproto developers could come to a rough consensus on this, it opens up the design space a lot; that’d allow backend and frontend developers to get to work on implementations at the same time. From the other thread:

@iame.li

Past that, I think we could do a lot if we just agreed on a VOD lexicon, with the Bluesky three-minute clips providing a starting point. That’d allow users to define their content libraries in their repos and could decouple the content with the distribution; you could imagine a front-end that renders the content library with support for Bunny, Cloudflare, etc, with an Internet Archive link as a fallback. You can even go a step further with something like a Bittorrent magnet URL if you’re not looking to monetize the distribution.

I (and Streamplace) would be interested in doing a few community consensus/work sessions to see if we could throw together something usable for this. Open questions for such a design, off the top of my head:

1. How is video content uniquely identified? I don’t think “take the CID of the input video” is a very good idea; who wants to keep the input video around after processing? (Streamplace has a very opinionated answer to this based on segmentation but I wouldn’t necessarily want to lock the entire ecosystem into Streamplace segmentation.)
2. Realistically we’re probably not going to ship fully-atproto-backed infrastructure for this stuff in the next few months, in time for (say) AtmosphereConf. Should our design accommodate a middle ground whereby videos can be backed by a variety of CDN providers and URLs, to accommodate videos on things like Bunny Stream or Vimeo?
3. What’s the moderation story look like?

If folks are interested I could set up some recurring meetings to start to flesh this stuff out, kind of like a Community Specification process but I’m not sure I’d want to be that formal.