Private Lexicons

Hi,

I’ve read a lot about permissioned/private data, and for me it has always been about keeping the contents of a post private/shared with a pre-approved group of people.

I’m wondering if anyone has explored using lexicons in a permissioned/private manner.

How could this be achieved without severing interoperability? Has anyone given this any thought?

There’s definitely been a lot of thought.

Northsky sounds like maybe the closest thing to what you’re describing.

There’s lots of thoughts in this thread:

Also check out the permissioned data diary from the Bluesky folks that is slowly coming out with more posts:

2 Likes

Ohh I missed so many posts on that leaflet, guess I will have a place to dive back in, also appreciate you pointing me to Evelyns post! Cheers!

1 Like

Just a heads up, some stuff apparently changed with Northsky since that article was written:

Just in case it takes you a bit to get to the bottom of that post. :smiley:

2 Likes

All the proposals also use Lexicons. The Lexicons have to be “known” – the Lexicons themselves aren’t permissioned or private (although some approaches have additional records added to carry permission info).

3 Likes

Yep, in each case it has to be a unique lexicon due to the permissible metadata. I was thinking idly how to handle arbitrary lexicons and I guess you could just fire it right at stratos but then it doesn’t know the permission of it so maybe adding a verifiable “addendum” to the lexicon could work rather than making lexicon devs adjust or get it verified or something.

1 Like

You can add fields to lexicons that are just ignored so yes @evelyn.northsky.team you should totally follow that pattern if you want arbitrary lexicons to work with your system.

Someone still has to build an interface to view and render them.

A private microblog post is going to be different than an event.

1 Like

just off the top of my head, I think it makes a lot of sense for app data to be stored in the same lexicon regardless of what kind of repo it’s in, actually; feels like the shape+authN of the data and the authZ of that data should be decoupled if possible?

2 Likes