#social 2015-10-22

2015-10-22 UTC
nicolagreco, jasnell, kevinmarks, bblfish, aaronpk, JonathanC, the_frey, ShaneHudson, the_frey_, peacekeeper1, peacekeeper2, jaywink and tilgovi joined the channel
#
jasnell https://github.com/jasnell/w3c-socialwg-activitystreams/pull/220
#
jasnell cwebber2: ^^
#
aaronpk "816 additions, 720 deletions not shown"
the_frey joined the channel
#
cwebber2 jasnell: will look
the_frey joined the channel
#
ben_thatmustbeme cwebber2: while i'm thinking of it (somewhat on a lack of sleep today) why do you have published and updated when an object is deleted? why would you care? http://w3c-social.github.io/activitypump/#h-delete-activity
#
cwebber2 ben_thatmustbeme: good question! tbh, I kind of hate this, and I have a direction I'd like to go to get rid of it, but let me explain the current logic.
#
cwebber2 ben_thatmustbeme: if I get a notice from rhiaro saying "rhiaro favorited this image by ben_thatmustbeme of a kitten typing on a keyboard"
#
cwebber2 that's great, right? everyone loves kittens!
#
cwebber2 but instead, what if it was this
#
cwebber2 ben_thatmustbeme: I get a notice from rhiaro saying "rhiaro favorited this note by ben_thatmustbeme saying "I am against Godwin's Law because I support the Nazi party""
#
cwebber2 wow! did you really say that? i'd better check, right?
#
cwebber2 so my server should verify that you actually posted this
#
cwebber2 all sorts of nodes might point to other nodes
#
cwebber2 and so the document requires to avoid spooofing attacks
#
cwebber2 that you verify that the other objects on other peoples' servers really came from them
#
aaronpk whaaaat does that have to do with published and updated?
#
cwebber2 aaronpk: I'm getting to it.
#
cwebber2 this is a prereq for the current situation.
#
cwebber2 also I never believe rhiaro or ben_thatmustbeme would ever do these things, other than the kitten one :)
#
cwebber2 I was intentonally picking an example of something that clearly wouldn't have happened, and throwing in a godwin's law example, hope that's clear
#
cwebber2 so
#
aaronpk (this is why i like the microformats/webmention approach where we don't send notification text around and everything is always referenced by URLs)
#
cwebber2 right
#
cwebber2 well what if my thing points to your old activity, and it isn't there?
#
cwebber2 and it was never there either
#
ben_thatmustbeme then its treated as invalid
#
ben_thatmustbeme if you liked something that never was there, it can just be dismissed
#
aaronpk you can tell if something was there before based on whether the site returns 404 or 410 for the URL
#
ben_thatmustbeme if you like something that is old and deleted.... its treated as invalid and can be dismissed
#
ben_thatmustbeme as to the case you were getting to, i see where you are going, you can see when the post existed, that it changed at some point, things like that
#
cwebber2 right, so, tsyesika can clarify on this better than I can
#
cwebber2 but the validation step requires that things stick around permanently
#
rhiaro_ I like the approach of just referencing things by URLs - if the receiver needs to verify it anyway, the sender doesn't need to ship the whole lot in the first place
#
cwebber2 except that I *HATE* this approach.
#
cwebber2 especially
#
cwebber2 because it means that you can't implement transient communication in the spec
#
cwebber2 a log that a communication occurred is around forever
#
cwebber2 no plausible deniability
#
ben_thatmustbeme but edited is basically pointless as you have to treat the non-authoritative data (godwin's law line) as possibly incorrect / false
#
aaronpk if you don't like URLs, then you are basically creating a message passing platform that has nothign to do with the web, and can be accomplished with TCP and crypto signing
#
ben_thatmustbeme cwebber2: there is nothing saying that you can't delete it after you have given it to the target or have some sort of handshake there
#
aaronpk or just use jabber
#
cwebber2 well I think crypto signing probably should be in there for the transient communication sake
#
cwebber2 honestly, I think dereferencing nodes is better usually
#
cwebber2 I do agree there.
#
ben_thatmustbeme also, this would probably be a really GOOD thing for corporate social networking where they don't want plausible deniability as to what you said to someone at another company
#
cwebber2 activitypump's predecessor, the pump api, has the option to support either dereferencing
#
cwebber2 or including in line
#
cwebber2 except
#
cwebber2 it means that activities recursively bloat.
#
ben_thatmustbeme there has to be some trust of the source if you don't derefence
#
cwebber2 ben_thatmustbeme: that's true
#
cwebber2 I think for server to server, it's necessary to dereference anyway, so maybe doing the recursive bloated approach is not so helpful anyhow
#
cwebber2 client to server is different
#
cwebber2 it makes sense to give the client a filled out payload with all that data for display purposes
#
cwebber2 but you trust the server you're retrieving it form, so
#
cwebber2 from
#
cwebber2 anyway, this is an annoying part of the spec IMO
#
cwebber2 ben_thatmustbeme: I'm open to removing the "keep a shell around" part
#
cwebber2 but I think tsyesika and evan would need to be part of that conversation
#
cwebber2 it's my least favorite part of the spec.
#
cwebber2 ben_thatmustbeme: glad you're reading it and giving feedback though, that's great :)
#
ben_thatmustbeme well depends on how you mean client to server. client to your own server (how micropub works) you can trust because you have oauthed to your own server
#
cwebber2 ben_thatmustbeme: right
#
cwebber2 that's what I mean.
#
ben_thatmustbeme if you are doing client to someone else's server you can only trust data it reports about its own data
#
cwebber2 right.
#
ben_thatmustbeme its things like this where i still say client - server and server to server are different things
#
cwebber2 ben_thatmustbeme: sure
#
ben_thatmustbeme there is certainly overlap
#
cwebber2 ben_thatmustbeme: activitypump only specifies client to your own server
#
cwebber2 for that reason
#
ben_thatmustbeme okay, other question, as probably the one to use media the most (you as in mediagoblin... the royal you :P) do you see the necessity of the mediaType property?
#
ben_thatmustbeme seems to me like its better to fetch the content and assume that
#
cwebber2 ben_thatmustbeme: I *absolutely* see the value of the mediaType property! I think it's pretty critical for us being able to represent things properly.
#
cwebber2 MediaGoblin gives different presentation displays depending on what kind of media is being shown
#
ben_thatmustbeme but again, it could be given incorrectly
#
cwebber2 a document PDF loads a totally different display than an audio display
#
ben_thatmustbeme i'm talking server to server
#
cwebber2 ben_thatmustbeme: sure, and that'd be the fault of the remote server
#
cwebber2 anyway, yes, absolutely, I'm glad that property is there
#
cwebber2 it fits well with MediaGoblin's design
#
cwebber2 in this case, the worst thing that will happen is the remote media won't display right for the remote server's media
#
cwebber2 which is too bad, they won't get to show off their stuff
peacekeeper1, the_frey, bblfish, nicolagreco, elf-pavlik and kevinmarks joined the channel
#
aaronpk standards: https://medium.com/@ade3/the-zombie-mobile-b03932ac971d#.15kcnjt39
#
cwebber2 aaronpk: heh :)
kevinmarks, bblfish, the_frey and the_frey_ joined the channel