#social 2019-05-29
2019-05-29 UTC
timbl, Guest84, up201705417, biodan and tenma joined the channel
# 
jaywink[m] trwnh (IRC): sure, though I find it difficult to understand how "public" could mean "followers only" so yeah definitely there should be a way for for example mastodon to define this without having to abuse `as:public` - since by no definition of the world can public mean "only a defined list of public". Every definition of "public" in for example https://en.wiktionary.org/wiki/public indicates not limited to a group of people but open to an unknown
# jaywink[m] amount of people in the larger community.
# jaywink[m] what that larger community is is the thing that is impossible to define by the exact nature of larger communities - they're unknown in size
# 
trwnh i didn't say public = followers-only. i said that there should be levels between the two. consider "logged-in users only", or "password protected", or "friend-of-a-friend"
# jaywink[m] if you share content with an unknown number of people then one cannot realistically be disappointed if a random person you don't like sees it or a search engine indexes it
# trwnh e.g. semi-public
# jaywink[m] I remember we had this discussion in diaspora years ago and attempts to introduce a "logged in" visibility was rejected due to the simple fact that the software cannot make any guarantees for you - it gives a false promise of privacy. but I also understand people might want that (and actually have the same visibility in socialhome, my own software ;))
# jaywink[m] I definitely agree to your earlier comment re visibility field, just pointing out my opinion of guarantees of "unknown number of people" visibility and the problems it will inevitably also introduce
vitalyster joined the channel
# trwnh this'd probably be easier with ocap, heh. imagine a token that can be delegated once -- that's FOAF semantics.
# trwnh or a token that is granted to any actor that requests it -- that's "logged in users" right there
# trwnh really, i think the fundamental issue here is that once the data leaves your server, it's effectively a question of trust in whether the other system will respect your rules
# trwnh and partially the issue with federation when not accompanied by something else like ocap
# trwnh it would also be easier to do access control if you moved from a push model to a pull model, but that's not a direction i think most people want to go bc of the centralization dynamic it introduces
# trwnh although... if done like zot's openwebauth then it doesn't have to introduce centralization dynamics. it just makes caching less desirable.
# rialtate[m] The idea that any software system could keep something perfectly private under any conditions is completely nonsense. And users will always be a weak link. You can't stop leaking of a token? Well netflix can't stop people from sharing passwords to their entire account. There is no such thing as security, only trust, and that will always be in the hands of some human somewhere along the lines.
# trwnh ^ exactly
# trwnh in fact i think OWA tokens are single-use anyway? they basically just establish a cross-domain session
# trwnh that's good enough for most purposes, right?
# jaywink[m] that probably works for non-public (in every sense limited and known who shared with) content, but as soon as something is shared with an unknown number of people it becomes a matter of "I'll tell you my rules, please respect them"
vitalyster and xmpp-social joined the channel; vitalyster left the channel
# fr33domlover o/
# fr33domlover the AP spec says 'followers' is for actors; what if I want to follow something that isn't an actor? What should I name the 'followers' collection?
# fr33domlover I mean, the spec says 'followers' is for use by actors, but I have non-actor objects that can be followed
# jaywink[m] create another collection for them if you really need one?
# jaywink[m] `followed_notes` etc :)
# fr33domlover jaywink[m], well I'm listing that collection under the non actor object, e.g. an issue or merge request may have a followers collection. I'd love to just use the 'followers' property because this is what it's for, subscribing to updates etc., and since there's no Actor type I imagine the RDF domain of 'followers' has to be 'Object', BUT the AP spec says: 'followers' is a collection of actors following
# fr33domlover this *actor*. So it implies the domain in actors and I'm unsure whether to use 'followers' in non actor objects, or use a different name
# cjslep[m] Additionally, isn't the `stream` property a list of other collections? Or am I thinking of it incorrectly
# fr33domlover cjslep[m], I already checked that stream stuff and decided it's a weird unspecified thing and there's no benefit using an object inside an object, and I'd need to choose a property either way
# fr33domlover cjslep[m], what's the rdfs:domain of 'followers' in your AS2 OWL file?
# fr33domlover I couldn't find the OWL in go-fed
# fr33domlover (But I admit I didn't try grep -r ^_^)
Guest84, saranix, timbl, Chocobozzz, Chocoboz1z, tantek, Loqi, tigerfinch, brion, pmbauer, fr33domlover, bitbear, pdurbin, melody, trwnh, mattl, rZr, ma1uta, aaronpk, downey, englishm, KjetilK, dregin, gobengo-irc-bot, stevenroose, DenSchub, ichoquo0Aigh9ie, dansup, puck, ajordan, bigbluehat, xmpp-social, rhiaro, martijnvdven, csarven, ben_thatmustbeme, sknebel, erincandescent, Gargron, fiatjaf, lain_soykaf, er1n, jdormit, raucao, nightpool, dlongley, up201705417, nicolas-constant[m], Rixon[m], EricDrechsel[m], natmac[m], rigelk[m]1, paulus[m], nolep[m], ircer[m], josef[m], grag[m], schmittlauch[m], jeeg[m], finn, cesar[m], Po-chiangChao[m], ciet[m], benaiah, dag[m], etjet[m], cawal[m], iarp[m], alicedebob[m], rigelk[m], heluecht[m], labs[m], jaywink[m], cjslep[m], jdormit[m], nightpool[m], cybrematrix, rialtate[m], xkr47, dzho, ahihi, rektide, Cult, edsu, cm, biodan and tenma joined the channel
# cjslep[m] fr33domlover (IRC): Unsure if you are wanting domain or range, but here is a link to domain (range is below): https://github.com/go-fed/activity/blob/master/astool/activitystreams.jsonld#L5143
# fr33domlover cjslep, thanks! Hmm what if someone uses a custom actor type? Then things would fail
# fr33domlover But yeah it means non-actors can't have followers
timbl joined the channel
# cjslep[m] Just means the code wouldn't compile, people would need to code generate a fork.
vitalyster, Loqi, dzho, rektide, xkr47, edsu, Cult, ahihi, nicolas-constant[m], labs[m], dag[m], rigelk[m], ciet[m], dlongley, ircer[m], Guest84, nightpool[m], jdormit[m], cesar[m], rialtate[m], schmittlauch[m], alicedebob[m], rigelk[m]1, jeeg[m], cybrematrix and jaywink[m] joined the channel; vitalyster left the channel