@lucy@grillchen@sun I guess I can't see posts from brotka.st, but it is P2P fedi and designed to avoid the problems of servers being heavyweight and difficult to maintain, the handshake problem of post distribution, the liability/moderation problems, the imbalance between users and admins (specifically things like users having less control than admins about what sort of thing they see and some of the privacy issues), the bandwidth problems for content distribution, the reliance on DNS.
It is also a thing that I just avoided putting "PissDB" into and decided to call it "PiDB".
@lucy@grillchen@sun No, it is not like either of those. It'd probably be closer to SSB if you removed a lot of the brain damage and came up with a clever way to bridge to fedi and you didn't hate the clearnet.
@sun@grillchen@lucy It's Go. I actually would have written it in C but it is easier to port Go programs from Plan 9 to Linux than it is to do the same for C programs and I've been doing the development on Plan 9; the binaries running on FSE/Screamshitter/Apotheosis/etc. were just built from the same mkfile on the Plan 9 machine. (It'd be "typedef PiDB struct" in C, and I'd be using lower-case letters because fuck CamelCase.)
(I think I'm already pushing it by calling the internal K/V store "the ID Index Orthogonality Table". If I am too whimsical the code will read like Honk's and I like Honk but it is sometimes not as easy to read when all of the words rhyme and differ in only one letter. There was a conlang where the guy had made related words share prefixes, and it turns out that this was a terrible thing: if you make the first syllable of your words for "fork" and "knife" and "spoon" the same, it makes it less fault-tolerant. No one hears "fork" and thinks you might have said "knife", they sound and look very different from all of the other words that might be used in their place in a given context. It's fine that a fork on the table and a fort on a hill sound similar enough to be easily confused because you do not usually use them in the same context, same reason "hashi" is "bridge" and "chopsticks" in Japanese. Like words have a local optimum for avoiding ambiguity. I think Honk's readability kind of suffers from zonking its donks throughout the codebase.)
@p@grillchen@lucy I am trying to add some special features to my implementation, and pleroma but pleroma is difficult because it involves refactoring. when I get my implementation of certain "features" going I'd like to contribute it to yours
@p@grillchen@lucy@sun >if I can get Pleroma to feed the media directly If not making a custom uploader module, you could implement a basic subset of S3 protocol or IPFS-compatible uploader gateway.
@sun@grillchen@lucy I would be delighted. One of the things I have slated is getting Pleroma to fetch objects from alternative sources (in this case, a locally running Revolver) so as to allow the network to distribute posts rather than making every server fetch the same thing from the origin and flooding smaller boxes or failing to fetch posts when an instance is down; same thing with media, which could be a big deal (and if I can get Pleroma to feed the media directly to Revolver, even better). Once it's easy to fetch even when the other instance is down, it should be possible to make object expiration cheaper.
@vic@grillchen@p@lucy@sun I found it absolutely sucking if you're the only one seeding the file, might as well give a direct HTTP link at this point. Same can be said for plain old torrents, but unlike those IPFS is pretty much expected to be accessed from the browser. That said, I meant just piggybacking of those third-party uploader APIs pleromer already supports.
@mint@grillchen@p@lucy@vic you still need an http gateway to access it on the web, there is an IPFS-JS library but its been like almost 8 years and it still can't locate nodes from other nodes, you have to give it a list.
@sun@grillchen@p@lucy@vic Didn't Brave or some other Chrome reskin support IPFS natively or did it just do a redirect to some gateway upon opening ipfs: URI?
@i@grillchen@p@lucy@vic@sun The RFC itself might've had some good intentions, I'm mostly pissed off at the fact I can't just torsocks curl (or whatever using libcurl) anymore with the only alternative being fucking around with envvars that might or might not be respected. https://github.com/curl/curl/pull/11236 - even the patch got rejected with no other proposed solutions I know of.
> I'd like to wish ass cancer to whoever made RFC 7686
Internet Engineering Task Force (IETF) J. Appelbaum Request for Comments: 7686 The Tor Project, Inc. Category: Standards Track A. Muffett ISSN: 2070-1721 Facebook October 2015
Appelbaum is a creepy asshole; I don't recognize Muffett's name but it does say "Facebook" under it and that's enough.
I do not understand the point of forcing that behavior. I get the rationale but it seems pretty pointless.
@syzygy@grillchen@lucy@mint@sun@vic I may or may not have mentioned him but maradydd and some others talked about it at length. I don't know if the website is still up; I did read the IRC conversation where he tries to get everyone to cancel a guy for being ex-military ("how many innocent iraqis did you kill" / "i feel so unsafe right now you guys" / etc.) and he's kind of a sociopath.
@vic@mint@grillchen@lucy@sun IPFS's codebase is a turd on fire and their network is a shitshow but being drop-in compatible in a limited context wouldn't be a bad thing.
Oh that is always fun. Is it because of the custom TLD or is it because of some aspect of proof of ownership, or something else entirely different like the SOCKS debacle of a few months ago?
@RedTechEngineer@grillchen@lucy@sun There's stuff I miss about Limbo when writing Go (language support for a linked list of tuples would be nice sometimes) but I like to be able to hand people one executable and they can run it.