@lispi314 "obsession with trees instead of graphs" something something models predispose implementation, once multics debuted the idea of the hierarchical filesystem it was only a matter of time before the implications of that decision spilled out to other areas too (the past half century and a bit of computing orthodoxy really being highly elaborated and selective multics fanfiction)
@affine@allison@lispi314 humans and their societies tend to centralize. you can think of it via adam smith (as it being rational because specialization) or via donnela meadows (it being an unfortunate result of perverse incentives or winner reinforces winner games) but its an inevitable recurrence.
i think peer stuff has been solved as much as it can but the social issues are relatively insurmountable.
@icedquinn@allison@lispi314 File organisation is one thing, but the techno-feudalism model of lords owning server farms, lending fiefdoms to admins who then populate them with subjects (like now on fedi, but the node-user relationship baked into Xanadu addressing also smells of it) is another. But idk, maybe it's also inevitable due to material conditions of the internet (and doing p2p properly is hard)
@affine@allison@lispi314 it often comes back to tying things to key pairs directly or some uuid which is backed by an attestation from some keys.
i made some posts earlier in the year about trustify as a middleware for handling key tests because "do we trust this key" turns out to be an entire rabbit hole of suck. this is to say nothing of people tend to lose them.
@icedquinn@allison@lispi314 Social issues can explain things like spam filtering by domain (which is a kludge, but still beats being flooded by random user ids on nostr) but there's nothing inevitable about tying a user's identity to the admin's property, or tying the instance's identity to a rented domain. It was just easier to implement it that way. And despite that, fedi is substantially better (for me, at least) than centralized services (including technically-it-could-be-possible-to-host-an-independent-implementation-trust-me-bro-its-coming-soon bsky) so I think it's still worthwhile to identify what parts could be improved.
@affine@allison@lispi314 fwiw dns itself doesn't give a shit. being a tree of rent-seekers is a social issue. nothing at all prevents people from using namecoin-esque solutions to distribute records. nothing at all stops people from having socialist TLDs. nothing actually stops you from using a UID or pubkey as a strong name and then having dns servers just cache attestations.
it doesn't make lines go up so it doesn't get done.
@lispi314@allison@affine > from insufficiently polished software, not the tradeoffs inherent to the schema
they are the same thing. i can give you a discord link and in one click it just works. the same cannot really be done for p2p. it can kind of be done for federated things.
deltachat has tried to help this by having part of *their* website that just acts as an onboarding URL for you to share contact details with. which is nice, though its still basically putting someone back in as "the instance."
@icedquinn@allison@affine Peer stuff has major usability issues (from insufficiently polished software, not the tradeoffs inherent to the schema) and tends to get abandoned so fast it rapidly suffers from bitrot if it was ever properly finished to start with.
@lispi314@affine@allison it'd be entirely possible to finagle a pubkey or instance name in to a URL for a name. though with p2p these key names can get kind of long and crappy. but in their case you're sort of interacting with copy/paste bits in contrast to just telling someone your aol user name is coolblob420 and everyone just understands.
qr code scanning helps i guess. people are more used to that form of copy/paste which again has worked out well for deltachat.
@lispi314@allison@affine > Which I specifically outlined as out of scope. which is itself irrelevant because those difficulties are literally a major irreducible pain point around the UX.
i've personally onboarded people to p2p systems and issues like key loss and somebody's laptop being off at the time breaking the network were adoption killers. they would heavily push back for the centralized service that, to them, "just works."
there's a reason amazon filed patents for "one click" buying. UX matters. critical UX failures can't be "out of scoped," they have to be "resolved."
@icedquinn@allison@affine Not what I meant. Consider even just how much the Briar desktop client lags behind compared to the mobile one (for an example that isn't actually abandoned).
And browsers are *built* for the web, obviously things that rely on it will work with browsers.
I do not consider delays and difficulties in obtaining distributed data to be the fault of implementations, they are difficulties inherent to the schema. Which I specifically outlined as out of scope.
@lispi314@affine@allison for the rest of retroshare and toxx they had no issues at all figuring out how to work the software. it was problems inherent in the p2p model that continuously blew everything up.