Post

OK, I wrote up a whole blog post with some more detailed thoughts on this. Curious if anything here resonates with other folks. Thanks to @iame.li, @dholms.at, and @bnewbold.net for helpful pointers and discussions! pckt.blog/b/bits-of-en...

deniability is worth keeping but i like the rest of this

Why do you think deniability is useful?

The more relevant questions are when is *non-deniability* useful? For whom? That capability is NOT a baseline expectation elsewhere online. The web + screenshots usually don't offer durable, attributable *proof* that someone possesses an authentic copy of what you once put online

Normal, non-deniable crypto provides two features here: 1. It keeps things simple. Better for builders, easier to understand. “Everything works the same except for the ACLs” 2. It enables redistribution in positive cases, if we can design redistribution systems that maintain the ACLs.

Don’t discount the latter point — redistribution is what lets AT scale as well as it does. I’m surprised that people want to bake in a scaling limit that will force private things to be small.

If privacy is a spectrum, I tend to be most interested in use cases which operate at interpersonal scopes Aside from rent-seeking, what use cases would you find valuable for anything like "broadcast-scale privacy"? The very idea sounds like an oxymoron, so what am I missing?

I have to keep reminding myself that you + I are engaged in different activities here My goal is to identify the sweet spot in Daniel's design -- where his chosen constraints make things easier vs. harder Whereas you're sketching out a new design with different core assumptions than he chose

NB: My idiosyncratic way of looking at atproto's architectural choices is more data-centric than protocol-centric From this perspective, protocols are "just" an enabling sync tool which lets me build new UXes from my copies of richly-linked lexicon-formatted data

Within that ecosystem, I view as peers anyone else who's thinking about how to cleanly collect + organize the data we each store on our own PDSes Thus the most interesting scenarios tend to become coordination problems where "everyone" uses the same lexicon to store their own records