Considering the dial one turns when building social software, between emphasis on people and emphasis on the spaces they occupy.
Compare being "in" a channel or rooms vs. "following" an individual. Systems focus on one or the other (yet always support both: DMs and privmsg; hashtags and groups)
Anyway this all needs to be rebuilt atop matrix.org or something because content addressing. Seriously it's 2017 and we're back at "overloaded server, can't load". Not even like zooko's triangle.
@Gankro chained history graph => same data no matter where you pull it from (bonus: these days p2p in the browser works too)
@graydon does that force some kind of transitivity of federation?
e.g. if neutral.site communicates with nazis.site, do I have to transmit nazi.site's content to transmit netural.site's?
@Gankro Depends on the granularity of chaining. But yes it has potential to cause some "carrying other people's traffic" (that's at least part of the point!)
You can also make mutant forms that cut off the fabric at a given topology or trust threshold, like shallow git clones or required signature-trust proximity or whatever.
@Gankro My general feeling on this is that we should build for the "friend-of-friend" threshold (and not friend-of-friend-of-friend): i.e. assuming each person has order-100 friends, you handle order-10,000 people's traffic, which is probably fine on a modern machine. There'll be plenty of redundancy, and not _much_ room for abuse.
@Gankro Clearly I haven't explained of my _other_ Clever Design Intuition for Social Software, which is to cap graph-degree for each person / each room at some reasonably low number. Set a speed limit / fixed amount of human oversight on propagation. No super-propagators.
@graydon so what you're saying is that you would ban @sarahjeong for Literally Destroying Mastadon 🙃
@Gankro Certainly not! Backpressure / fanout-caps / rate limiting is completely different than banning. Different purpose, different people effected, different consequences. A maxed-out account is a successful account, not one you want to ban!
@graydon oh god people are going to build account multiplexers to get around this aren't they
@Gankro every problem in social software comes face to face with the turing test and/or identity farming at some point.
@graydon but jokes aside, I have a vague discontent towards the idea that participation in a social network should require a "good" computer/network. Also p2p seems like a nightmare given the way phone networks currently operate?
@graydon (to say nothing of battery concerns)
@Gankro Sometimes. Native phone apps (including web pages on phones -- https://wicg.github.io/netinfo/) can change behaviour when they're on a fast/cheap network like local wifi and/or mains power.
When on both, your phone might as well mesh. Agreed that when only on battery and lousy 3G it should drop to dumb client mode.
(Laptops and desktops more-likely to default-mesh)
@graydon *thinks about how those features are pretty much only used for fingerprinting*
@graydon well, hmm... the fact that some accounts are massively followed isn't inherently problematic.
It's accounts that follow a lot that are problematic to this metric. The most extreme examples I have in my twitter follows are people with ~2000 follows, but that's fine as long as those are scarce...
Although it sucks for them and their computer?
@Gankro Like in your own friends-of-friends 10k set, there might be a few total assholes but likely no more than you can manually blacklist and/or figure out which friend-with-bad-judgment they're all coming from?
@Gankro (Perhaps as a guideline: 10k is somewhere between "large highschool" and "small town"; like .. a still-tractable unit of social reasoning)
@graydon content addressing ~= magnet links for toots?