#social 2015-05-08

2015-05-08 UTC
bblfish, tantek, LCyrin, tilgovi, Arnaud, the_frey and shepazu_ joined the channel
#
melvster http://indiewebify.me/ seems to be trending
#
ben_thatmustbeme it ended up on hackernews
#
aaronpk mostly positive response, which is surprising for hackernews
#
melvster i think we need to come to a consensus on how following works, in order to implement swat0
#
melvster i dont think swat0 should be an approved use case until that is known
#
melvster or maybe split it up more
#
aaronpk the act of following is not part of swat0 intentionally
#
melvster well maybe this is one reason that almost no one has managed to implement it, in 5 years, let alone test interop
#
aaronpk the story is not about following
#
aaronpk i agree we need an approved story about following though
#
melvster ill be interested to see what people implement when this spec goes to CR
#
ben_thatmustbeme we should hope we get implementations long before then.
#
aaronpk indeed
#
melvster agree, but as I say, that's been hoped for, for 5 years already, it would be easier to do if the dependencies were also approved use cases
#
ben_thatmustbeme its on my list of things, is to start adding m-card to my micropub endpoint, and start being able to 'follow' that way, hopefully i can use that list as input to readers *hint hint aaronpk/monocle*
#
aaronpk also note that there is only one "following" relationship in swat0, and it's actually called "subscribed"
#
aaronpk "Evan, who is subscribed to Dave"
#
aaronpk which implies that Dave doesn't even need to know that Evan is subscribed to him
#
melvster i think it would help if that was explained to implementers
#
melvster how to do that
#
aaronpk i think the point is that swat0 works even if the "subscribed" is something dumb like a script polls Dave's site and sends an SMS to Evan on new posts
#
melvster or if there's another spec that describes it
#
melvster "Evan, who is subscribed to Dave" -- i dont unambiguously know what this means
#
melvster i would not know how to test this
#
aaronpk i *think* that's why it's in a parenthetical
#
melvster many of the other use cases are less ambiguous and I think could be more easily implemented
#
aaronpk we'd have to ask tantek and evanpro about the original intent
#
aaronpk many of the other use cases are equally ambiguous ;)
#
melvster it's encouraging for people trying to implement the spec to start with some use cases that are more self contained, imho
#
aaronpk yes, swat0 has always been a more advanced one
#
aaronpk that's why it's the last one I wrote up
#
melvster i agree, but it's still the *only* approved one at the moment, so I think it's where many people will start
#
aaronpk nah, people will start wherever they want
#
aaronpk or at least should ;)
#
melvster ok, well that wasnt clear, I thought swat0 was the focus
#
melvster why not approve other stories then?
#
aaronpk i actually don't know what the timeline or goal is for approving other stories
#
melvster i think it would be useful for an approved story for following to come before swat0
#
aaronpk i agree
#
aaronpk and several other simpler ones
#
melvster great
#
aaronpk yes, swat0 was by no means meant to be the simplest one, sorry if that was unclear
#
melvster no need to apologize, it just seemed logical to start with the approved ones, because then there is a chance to test interop
#
melvster the other ones are interesting too
#
rhiaro cwebber2: resume
#
Loqi Aaronpk made 1 edit to [[Socialwg/2015-05-12]] https://www.w3.org/wiki/index.php?diff=84065&oldid=0
#
cwebber2 hi
#
aaronpk hello
#
cwebber2 I was about to link the user story example I hacked up, and now I think I made a mistake in it :)
#
cwebber2 checks
#
rhiaro cwebber2: is there an example somewhere of what a republished activity looks like?
#
rhiaro (for everyone else's context):
#
rhiaro <rhiaro> so when A comments on B's post, A's server notifies B's server, so that B's server can propagate the activity to the receivers of the original post
#
rhiaro <rhiaro> Does that mean B republishes the *same* activity A published, or does B make a new one?
#
rhiaro <paroneayea> rhiaro: B publishes a new activity, referencing the old one
#
cwebber2 rhiaro: I made a mistake in my user story example, hold on I'm updating it
#
cwebber2 hilariously I included the object only, and not the activity ;p
#
rhiaro haha
bblfish_ joined the channel
#
cwebber2 argh I wish pypump had references to the original json objects it was passed in
#
cwebber2 and I wish Tsyesika was around :)
#
cwebber2 rhiaro: okay, so
#
cwebber2 wow
#
cwebber2 I was about to paste you a real example of a share from pump.io
#
cwebber2 it's... amazingly verbose.
#
cwebber2 http://pamrel.lu/55258/
#
aaronpk holy
#
melvster cwebber2: that's just a single share, or is it a bunch of stuff?
#
cwebber2 that's a single share. It repeats a lot of data.
#
cwebber2 I don't think it needs to do this for activitypump
#
cwebber2 but that's the reality in pump.io, it seems
#
aaronpk pump.io is not AS2.0 right? i keep getting confused
#
cwebber2 it's as 1.0
#
cwebber2 I was just referencing it to see "oh, I should see what a real life example is before fixing my current example"
#
cwebber2 in the user stories doc
#
rhiaro ohhh so I don't actually need to understand that
#
cwebber2 no you don't :)
#
rhiaro I was just about to ask if you could break that down
#
rhiaro but, never mind :)
#
cwebber2 I'm going to give you the actually simple version
#
rhiaro :D
#
rhiaro simple++
#
Loqi simple has 1 karma
#
rhiaro cwebber2++
#
Loqi cwebber2 has 30 karma
#
cwebber2 note, a lot of this is like, repeating content
#
cwebber2 where the content could just be an id
#
rhiaro ok
#
cwebber2 you don't need to inline the entirety of the debian actor
#
cwebber2 ;P
#
rhiaro yeah
#
melvster what's this?
#
melvster 'displayName': 'Debian Women MiniDebConf Bucharest next 16-17 '
#
melvster 'May',
#
melvster is that JSON?
#
aaronpk jusssst ignore that ;)
#
Loqi Aaronpk made 1 edit to [[Socialwg/2015-05-12]] https://www.w3.org/wiki/index.php?diff=84066&oldid=84065
#
cwebber2 melvster: this is a python de-serialization of the json.
#
melvster ah
#
cwebber2 I pretty printed it, but pretty printed it in python
#
cwebber2 I figured nobody wanted to see a completely one-line version of that.
#
cwebber2 here you go rhiaro: https://github.com/w3c-social/activitypump/blob/master/userstories/reading-a-users-recent-posts.md
#
cwebber2 I hope this helps
#
cwebber2 that's a simpler explaination
#
cwebber2 which has an inReplyTo
#
cwebber2 note, there may still be things missing, I wrote this as an example late at night, and obviously I already had the activity missing. But I think it's right
#
rhiaro thanks cwebber2
#
rhiaro takes a look
#
melvster looks like quite a complicated way to do a share, i dont really like the acct: uris because that dereferences to a whole other JSON language to support, which doesnt interoperate with AS2
#
melvster having said that, it's possible to write a translator in many cases
#
cwebber2 melvster: I think we agreed in the group that the acct: stuff wouldn't be supported
#
cwebber2 in the socialwg stuff
#
cwebber2 so a lot of this stuff can change from pump api -> activitypump
#
cwebber2 I'd also like to be less "duplicate all the data recursively in a payload" and more "linked data implies that the other end has that data anyway" where possible!
#
cwebber2 I think one reason pump copies stuff recursively so much is to save the other server round trips in fetching it, or to preserve the old state of things, or something
#
cwebber2 maybe evan can fill in more on why.
#
Tsyesika i believe that is the reason to also
#
Tsyesika when there is an object pump provides the entire thing
#
Tsyesika so you don't need to do many requests
#
Tsyesika this isn't confirmed to me by evan just my assumption from using the API
#
cwebber2 anyway, dereferencing of URIs are possible in activitypump
#
cwebber2 so you shouldn't have to produce such bloated things
#
cwebber2 thanks for confirming, Tsyesika :)
#
aaronpk that sounds like it works fine if you assume the API exists in a closed system ;)
#
aaronpk as soon as you move across servers, you can't necessarily trust the data
#
aaronpk i want to avoid putting things like that in the spec, in order to be able to support the federated use case
#
aaronpk i know we're not supposed to solve federation with v1 necessarily, but I want to make sure we don't prohibit it
#
cwebber2 aaronpk: I'm confused as to what you're arguing :)
#
aaronpk let me rephrase
#
aaronpk I don't want the API to require that objects from other domains is sent in a payload
#
aaronpk or at the very least, it should require that a dereferencable URL is sent
#
aaronpk e.g., if you see something in my stream that says "ben liked aaron's photo", you might get all the details about ben and ben's like, but you need to be able to go look at ben's URL to verify it
tilgovi joined the channel
#
rhiaro elf-pavlik: http://indiewebcamp.com/wiki/index.php?diff=19095&oldid=18096&rcid=19102
bengo joined the channel
#
rhiaro !tell elf-pavlik: http://indiewebcamp.com/wiki/index.php?diff=19095&oldid=18096&rcid=19102
#
Loqi Ok, I'll tell them that when I see them next
#
cwebber2 aaronpk: yes I think that's what I was referring to
#
cwebber2 maybe I should have said referenceable rather than dereferencable? I'm not sure on the term ordering
#
cwebber2 but seems we agree nonetheless
#
aaronpk i think so yes, just making sure that was explicit for the logs :)
#
aaronpk pretty sure i have this private webmention thing figured out
#
rhiaro ^ so this allows some degree of access control for static sites (elf-pavlik)
#
aaronpk it's definitely possible
#
aaronpk it's just a matter of writing it down
#
aaronpk i have two options in mind, going to write them both down to compare
bengo, LCyrin, jaywink, bblfish, rhiaro_, KevinMarks_, mattl_, LynnCyrin, ben_thatmust, cwebber2, bblfish_, rhiaro, ElijahLynn, cwebber2`, wilkie_, Loqi, KevinMarks, wilkie and tilgovi_ joined the channel
#
oshepherd cwebber2`: Was there actually consensus to not support acct: URIs?
#
oshepherd Evan, myself and actual deployments in the wild would seem to disagree with that
#
oshepherd aaronpk: Your client should trust *your* server, no?
#
cwebber2 oshepherd: so, there was agreement to go the "follow your nose" route with http URIs at the second face to face
#
cwebber2 however, evan opened an issue on activitypump to include them
#
cwebber2 so I'm not sure what's going on
#
oshepherd I think its' one thing to say "For HTTP(s) URIs, follow your nose" and another to say "We aren't going to support the account naming format that the vast majority of distributed social systems use today"
#
oshepherd One thing that early ActivityPump drafts mandated was that for any object with an acct: /ID/ the object references must also contain a self: URI
#
oshepherd This reduces the need to do any "out-of-band" dereferences to "user types name into UI"
#
aaronpk oshepherd: sure but i was talking about my server reading a stream from someone else's server
#
aaronpk i may be thinking a step ahead
#
aaronpk still need to get this clearer in my head
#
oshepherd aaronpk: In the case of pump.io, all server-to-server interaction is push. But yes, in the case of reading from another server, be conservative in what you accept
#
aaronpk i'm thinking of the case when pump.io is sending push data to another server
tilgovi joined the channel
#
cwebber2 let's add an activitypump "strict" mode where it's http urls only ;)
#
oshepherd aaronpk: In that case it doesn't matter if it's verbose or not; you have to be conservative anyway
#
aaronpk https ;)
#
oshepherd cwebber2: As opposed to?
#
cwebber2 oshepherd: as opposed to allowing acct: uris
#
oshepherd cwebber2: How does that interoperate?
#
cwebber2 well, I was making a joke.
#
oshepherd (Answer: It doesn't!)
#
oshepherd :)
#
cwebber2 hence the winky face
#
oshepherd But anyway, as I originally drafted the spec WebFinger or any other such name resolution system was confined to a very tiny corner (to the point that, if you are doing your own implementation for you only, it's probably unnecessary)
#
cwebber2 oshepherd: maybe you should weigh in here:
#
cwebber2 https://github.com/w3c-social/activitypump/issues/20
bengo joined the channel
#
Loqi Aaronpk made 1 edit to [[Socialwg/2015-03-17]] https://www.w3.org/wiki/index.php?diff=84067&oldid=83908
#
aaronpk i cannot for the life of me find a documented reference to that resolution :(
#
cwebber2 aaronpk: I'm not sure the minutes of the second f2f got published :\
#
aaronpk not even in IRC tho
#
aaronpk unless my find isn't working
#
aaronpk ahh found it
#
aaronpk http://socialwg.indiewebcamp.com/irc/social/2015-03-18/line/1426710400225
#
Loqi Rhiaro made 1 edit to [[Socialwg/2015-03-18-minutes]] https://www.w3.org/wiki/index.php?diff=84068&oldid=0
#
Loqi Rhiaro made 2 edits to [[Socialwg]] https://www.w3.org/wiki/index.php?diff=84070&oldid=84012
#
oshepherd posted long comment https://github.com/w3c-social/activitypump/issues/20#issuecomment-100385194
#
cwebber2 it's a good comment, oshepherd
#
cwebber2 I do agree with your corporate rollout point.
#
melvster im not sure it's a great idea to prioritize a new uri scheme (acct:) on top of http: and it even has a dependency on yet another JSON serialization (JRD) which must be supported and parsed, with some strange rules such as forcing results to be in order of 'preference' ... I think it's adding a big new dependency to the WG which may put the deliverables at risk
#
melvster at the same time, there's no reason to exclude it I dont think
#
melvster these points were brought up during webfinger standardization
#
melvster however i do think complete interoperability with acct: uris is an *impossibility* -- id be very happy to be proved wrong
#
melvster technical reasons: it doesnt handle numbers, all uri lists are ordered specifically in order of 'preference' (no other serialization does that), literal properties can contain at most one value (other serializations can contain many) ... there's a few other things that came up when I tried to write a convertor
tantek joined the channel
#
melvster this is the web, we really should be using HTTP when we can
#
melvster both indieweb and SoLiD do that, and what I saw of mediagoblin I think that could be the case too
bengo, bengo_ and bblfish joined the channel
#
oshepherd melvster: Uh, prioritize? Who said prioritize?
#
oshepherd And what do you mean complete interoperability is an "impossibility"?
#
oshepherd Are you thinking we are proposing to encode the entirty of people's profiles in a JRD document? (Nobody is)
#
melvster good
#
oshepherd And, as a *pure consumer* of webfinger, why would any of the properties stuff matter? Surely the preference order actually makes things easier? (Grab the first link)
#
melvster i mean interop between JRD and AS2
#
melvster ie json ld
bengo joined the channel
#
melvster writing a two way bridge i found impossible, id love to see how it could be done tho
#
oshepherd In the early ActivityPump drafts there was *one single defined use* of JRD/WebFinger: If a user types in "name@domain.com" to a "username" box and clicks follow/login/whatever, then you do webfinger query https://domain.com/.well-known/webfinger?q=acct:name@domain.com[&rel=profile-whatever], grab the first rel=profile-whatever link which either matches "application/activitystreams+json" or has unspecified type, and then follow that link and proceed as if
#
oshepherd you'd been given a https URI
#
melvster let's see if it is needed for the use cases
#
melvster that's encouraging to hear
#
oshepherd It's simply a "human name/canonical name" to "URI" mapping
#
melvster i think jrd only breaks at the edges, for some stuff it will interop fine, with a library, lets see if we can get interop working tho
#
oshepherd For all I care you'd do .well-known/asprofile?q=acct:name@domain.com and get a HTTP redirect, but I'd rather we avoided *yet* *another* bikeshed on how to turn name@domain into domain.com/~name or whatever