#social 2017-09-08
2017-09-08 UTC
KevinMarks and meltheadorable joined the channel
# 
Gargron cwebber2: is there a json-ld attribute/context for "license"?
# Gargron i don't "need" it but i want to research my options for something
# saranix you could add a
{type:Link, rel:license} (http://microformats.org/wiki/rel-license)
# saranix chortles at the thought that # of followers is somehow considered prized information worth verifying "I need to know how many people in your hierarchy so I know how important you are" *animal sound*
melody and dmitriz joined the channel
# melody hi! i'm here
# saranix hello
# melody that's about accurate, i'm working on joining the SocialCG still -- my company technically has a business interest so I need to wait for some signoff before I can sign the CLA but I don't anticipate any difficulties
# melody is there anything i should be familiarizing myself with in order to be a good participant?
# saranix there's a lot of ground to cover but there is also a lot of cruft - IOW there's a ton of stuff to review but also tons of no longer relevant junk in between that would waste your time
# saranix will you be writing an implementation?
# melody eventually, maybe? for the social network i have in mind, my primary goals are anti-harassment and abuse prevention, everything else is secondary or tertiary, including federation -- i would REALLY like to make it work, and AP looks a lot more promising than ostatus did, which I was going to need to break or not use at all
# melody but i'm not sure how far off i am from an implementation yet, i'm still doing design research
# saranix neat
# melody i'm also not sure which places AP might be insufficient or where i'll need to break compatibility, if anywhere -- i read through everything here https://www.w3.org/TR/activitypub/#Overview but i'm still digesting
# melody presently i'm reading the open issues here https://github.com/swicg/general/ -- i can't remember how i found the github but i have a bunch of tabs open
# saranix also https://github.com/w3c/activitypub/issues
# saranix yes, many tabs is a frequent problem for me
# melody i will open all those tabs when i finish closing all these, thanks :)
# melody i wanna say -- it's really encouraging to see so many people taking anti-abuse seriously, i wish it wasn't so surprising but i'm used to it really being an afterthought for most people
# xmpp-social [ajordan] Lol I gotta say good night too but welcome melody! \o/
xmpp-social, KevinMarks, cdchapman and KevinMarks_ joined the channel
# Gargron melody: hi! abuse prevention was also one of mastodon's starting goals, have you tried it? maybe you don't need to develop from scratch - or maybe you see something to be improved upon further?
KevinMarks joined the channel
# melody Gargron: I have an account on a mastodon instance, but don't really use it. I've been eyeing it closely and I've given a lot of consideration to whether or not it would be possible to use it as a base for an implementation, but I'm honestly not sure yet.
# melody I am still working out which features I really need, which ones are possible, at which layer they need to be handled, and how badly I'd need to break compatibility to do it. Then I need to determine which assumptions are just rooted too deep in existing implementations to excise, if any.
# melody I do think there are some areas for improvement though for sure, the most obvious one at a more cursory glance is that the content warning system is handled exactly backwards from what I would want
# Gargron exactly backwards? please elaborate
# melody My instance needs to know which content I don't want to see so that it can filter it before it gets to me, creator-provided content warnings can supplement that for non-textual content, but a creator can't reliably know what the entire potential audience will find sensitive
# melody So sensitivity needs to be determined per-recipient, not per-sender
# Gargron are you looking to employ machine learning as a sort of spam-filter?
# Gargron but with personally undesired content instead of spam
# Gargron or are you looking more for defining some keywords in preferences and having toots that match those be filtered out from home/notifications?
# melody Nothing that sophisticated to start, I don't think it's necessary. I'm coming from a tumblr which has a lot of weird cultural and ecosystem stuff that has inspired some of my thinking on this -- the biggest being tumblr savior/xkit blacklist -- which are just lists of words to hide posts if they contain them
# melody It's important that it be server-side because the largest problem with relying on a browser extension for this behavior is it's not portable to other clients when you aren't accessing from your desktop browser
# Gargron right, server-side keyword muting. i see. it's been a feature request for some time
# Gargron it's possible & wanted but bothersome to make the UI for
# Gargron presumably you want to define muted keywords individually and decide whether that one's gone from home, notifications, or both.. or some such
# melody I also have sort of two different levels of sensitivity, which I've termed "muting" and "muffling" respectively -- if something is muted, it is filtered completely, I don't even want to know that you've filtered it, I don't want the option to opt back into it. If something is muffled it defaults to collapsed and tells me why, and I can choose whether to engage it or not in the moment.
# melody For muting, clients should never get the post at all when they request posts for me, the server should filter them out of the collection so that clients that don't understand muting won't show me the content anyway. "Muffling" would I guess need to be an advisory flag of some kind
# melody to tell clients the content is sensitive and should not be displayed by default
# melody I also recognize sort of three to four separate states that a post can be in, with respect to content warnings: 1) No content warnings apply at all [safe], 2) The creator opted out of providing content warnings [risky], 3) The creator provided content warnings [filterable], 4) The creator provided some content warnings, but they are innacurate/incomplete [risky]
# melody as an end user, I want the ability to distinguish between cases 1 and 2 and choose not to see posts for which case 2 applies
# melody based on the level of risk i'm willing to expose myself to
# melody I also want moderators to be able to add moderation metadata to posts that can disambiguate 3 and 4 but that gets complicated with federation
# melody With tumblr, this is sort of a culturally enabled thing because of the way tags are transient to a specific reshare -- people in your follow network can be semi-reliable in ensuring that reshared posts have enough tag metadata to be useful for your keyword blacklists even if the original creator didn't care, but it's a weak solution
# melody that's a lot of words
# melody some of the plausibility of how workable some of this stuff is becomes really dependent on other pieces of what I want to do that are probably not obvious tackling it only from this angle
dmitriz joined the channel
cdchapman, KevinMarks_ and dmitriz joined the channel
# 
cwebber2 btw did you notice Gargron https://tools.ietf.org/html/draft-cavage-http-signatures-07#section-1.2
# Gargron i think it was there for a while
# cwebber2 btw, some conversation happening right now on mastodon about adding a server-wide actor for example server-wide blocks of an actor without revealing who's doing the blocking https://mastodon.social/users/Gargron/statuses/17471620
KevinMarks joined the channel
# Gargron cwebber2: anyhow, how would you imagine custom emojis represented in AP?
howl joined the channel