#social 2017-06-16

2017-06-16 UTC
timbl joined the channel
#
@JefferyRPowers The latest De La Crème Marketing Daily! http://paper.li/JefferyRPowers/1425169611?edition_id=7c6ef500-522a-11e7-b4c4-0cc47a0d1605 Thanks to @SocialWebWG @sejournal #seo #socialmediamarketing (twitter.com/_/status/875509937807536132)
timbl, cristomc, ben_thatmustbeme, thebaer, ajordan and cwebber2 joined the channel
#
puckipedia https://github.com/puckipedia/Kroeg I published the code for my ActivityPub server finally :)
#
Loqi [puckipedia] Kroeg: An ActivityPub server in C#. Also does OStatus for fun.
#
aaronpk nice. do you have it running somewhere?
#
puckipedia it kinda runs at lol.puckipedia.com while I work on it
#
puckipedia (currently: turning it on)
#
puckipedia it should be online (https://lol.puckipedia.com/auth/login) - but there's a stupid bug somewhere with inbox and IDs exisitng already
#
puckipedia (it tries to add duplicate ids)
#
cwebber2 puckipedia: nice :D
#
puckipedia okay, I think I have oauth2 added now
cristomc joined the channel
#
puckipedia yup OAuth2 is there now too
#
puckipedia so I should now support fragments in IDs too (getting e.g. example.com/test#object will probably get example.com/text, with #object embedded) - makes https://github.com/w3c/activitypub/issues/224#issuecomment-302737846 work
#
Loqi [cwebber] @jaywink So another option is, maybe, to use fragment identifiers. Here's an example: ``` {"@context": "https://www.w3.org/ns/activitystreams", "type": "Create", "id": "https://social.example/alyssa/posts/49e2d03d-b53a-4c4c-a95c-94a6abf45a19...
#
cwebber2 puckipedia: cool :)
#
puckipedia I also registered #kroeg on freenode for this project
#
puckipedia (also, Kroeg is dutch for pub :P)
#
cwebber2 puckipedia: so you're using oauth2 for client to server I'm guessing? are you also using that workflow for server to server or do you have any authentication setup for that?
#
cwebber2 looks like I'll be at TPAC; who else is going from here I wonder?
#
puckipedia cwebber2: there's currently no server->server auth
#
puckipedia I was working on implementing JWS (per oshepherd's idea)
#
cwebber2 nods...
#
puckipedia also I have to do something for remote retrieval of entities (currently it's unauthed too)
#
cwebber2 nods
#
puckipedia ooh I just realise that I misread how LD signatures would be used
#
puckipedia I assumed the client was in control of the private key, not the server
#
cwebber2 puckipedia: yeah it would be that the client's in control for LD signatures
#
puckipedia yeah. that means e.g. wrapping an object with a Create (in client->server) can't be done, afaik?
#
cwebber2 oops
#
cwebber2 sorry I meant
#
cwebber2 puckipedia: yeah it would be that the server's in control for LD signatures
#
puckipedia right, woops :P
#
cwebber2 puckipedia: there's a possible workflow where the client does have its own private key kind of vaguely outlined in the spec, so that the client could be authenticated and authorized via the public key part of that pair, but you wouldn't use LD signatures with it
#
cwebber2 you'd use http signatures
#
cwebber2 since it's really more for each http transaction, rather than something that "sticks around" on the object
#
puckipedia provideClientKey and signClientKey, right?
#
puckipedia I looked at HTTP signatures and I'm not sure how you would communicate "this key is signed by the actor's key"
#
cwebber2 puckipedia: I think I know how to do it but I need to prove it via implementation ;P
#
puckipedia oh fun
#
cwebber2 you basically would have the server sign your client key
#
cwebber2 and present that signature authorizing access to the remote server you're accessing
#
cwebber2 there may need to be some vocabulary defined to clearly express that though.
KevinMarks joined the channel
#
puckipedia ah, the signature part isn't in the http signature spec?
#
cwebber2 puckipedia: well presenting it as basically a certificate isn't iirc
#
puckipedia right, explains a lot
#
puckipedia ... { "type": ["Create", "Post"], "id": "https://example.com/thing", "object": "https://example.com/thing", "content": "Hello.", "actor": "https://example.com/me" }
#
puckipedia I'm scared that this might be valid
#
cwebber2 puckipedia: oh I was about to correct you :)
#
cwebber2 puckipedia: hehhhhhhhhhhhh well
#
cwebber2 yikes
#
puckipedia yes
#
cwebber2 puckipedia: that's some epic vocabulary trolling right there...
#
puckipedia ..... I think this might be valid
#
puckipedia I'm scared
#
cwebber2 well, it's not invalid technically, but whether it'll do what you want, I dunno...
#
cwebber2 depends on a server to server basis :)
#
cwebber2 I'd have to test with pubstrate to see what it does
#
cwebber2 hilarious though
#
cwebber2 I think really realistically
#
puckipedia yeah. I was thinking of ways to avoid having different IDs for activities and their objects
#
puckipedia and honestly I am scared by what I made
#
cwebber2 so I think it would render in pubstrate, though the server would accept it, but it might not do exactly waht you want
#
puckipedia I use a bunch of First() for the Type
#
puckipedia so it'd store the object as a Create in the database
#
puckipedia and it'd probably indicate it's an Activity, which means it'll probably stack overflow the HTML renderer
#
puckipedia my AS2 renderer though, will be just fine (it only unflattens each object once)
#
cwebber2 I don't think Pubstrate would have a stack overflow because of how I'm doing rendering but I should probably try :)
#
puckipedia so if you e.g. have a CollectionPage with 10 objects, and each is attributedTo the same person
#
puckipedia it'll only render the person's object for the first object
#
cwebber2 hilariously Pubstrate might do the right thing here: in rendering on a feed, the outer Create would win, but then it would try rendering the inner object, and it would find the Post type, and it would render that
#
cwebber2 however if you have a system where in your db one type "wins"
#
cwebber2 then yeah it would just break it
#
cwebber2 puckipedia: but if you wanted to break it with your stack overflow problem
#
cwebber2 you could have done the same thing with still "type": "Create" right?
#
cwebber2 with it posting at itself
#
puckipedia true
#
cwebber2 if I'm understanding how you're describing your code right
#
cwebber2 sounds like you might want a recursion limit
#
cwebber2 in your renderer/flattener
#
puckipedia it's just the static frontend, so I might fix that there
#
cwebber2 nods
#
ben_thatmustbeme testing a recursive object might be a good thing to put in the test suite
#
puckipedia also maybe the AP spec should just plain out disallow someone from Creating another Activity?
#
ben_thatmustbeme make sure that implementations don't break on things like that
#
puckipedia (at least, if I understand the spec correctly that is not supposed to happen)
#
puckipedia same with liking activities, as those are technically objects too
#
cwebber2 ben_thatmustbeme: might be hard to test where it would and wouldn't break things; eg in puckipedia's example it seems like it would store / etc fine, and even render as AS2 fine
#
cwebber2 but rendering as HTML it would break
#
cwebber2 which is technically out of the concern of the AP spec
#
cwebber2 but
#
cwebber2 might still be worth exploring
#
cwebber2 as a test anyway
#
cwebber2 a bonus "don't fall on this particular sword" check
#
puckipedia well, the HTML renderer would be a static client
#
puckipedia so that might fall under 'testing clients'
#
puckipedia hm. another fun test might be: server->server federating a create with an object with an attributedTo with id=a, and the create has an actor with id=a but different values
#
ben_thatmustbeme yeah, thats what i mean, its good to give people things that are easy to get tripped on
#
puckipedia maybe fragments in IDs is fun as a test case
#
puckipedia I retrieve the object by GETting without fragment
#
puckipedia then I store that object flattened in the DB, and try to extract the ID *with* fragment
#
puckipedia so e.g. { "id": "example.com", "object": { "id": "example.com#object" } } works
KevinMarks_ joined the channel
#
jaywink I think K I wass
#
jaywink Argh sorry
#
jaywink Anyway, I was mentioned at some point. Sorry, on mobile client which means typos. Fragment ids didnt mean like a very compatible idea. I'd that something @cwebber2 you would actually recommend?
#
jaywink I'd == is
#
cwebber2 jaywink: what's incompatible about fragment ids?
#
puckipedia I think it's actually a really nice way to fix the issue of having sub-objects
#
cwebber2 I think that the way puckipedia is doing it is probably right
#
cwebber2 fetch the page, then search for the fragment within the document
#
puckipedia yeah, I get it for mostly-cheap because of the way I store entities in the DB
#
cwebber2 puckipedia: how are you storing them?
#
puckipedia as you would in a nosql db, so I remove all subobjects and store them separately
#
puckipedia e.g. {"id": "example.com/create", "type": "Create", "object": {"id": "example.com/note", "type": "Note", "content": "hi"}}
#
puckipedia is stored as "object": "example.com/note", and example.com/note is stored separately
#
cwebber2 puckipedia: yep that's what I'm doing as well
#
puckipedia and when rendering as JSON I just say "render with max depth 3"
#
puckipedia currently this is just a hard-coded list of id-having properties, but once I have JSON-LD I can store the proper context and separate them by which property has a type mapping of @id
#
cwebber2 yep that's also what I'm doing puckipedia
#
cwebber2 nice to hear we're on the same page
#
cwebber2 puckipedia: one thing I haven't solved though
#
cwebber2 is whether not to copy parent @context onto embedded objects when you "extract" them
#
cwebber2 you could expand then compact the whole thing to deal with it I guess but that seems a bit wasteful maybe
#
puckipedia I do not do json-ld @context yet
#
cwebber2 I'm sure people who know json-ld better than I do have better answers about this
#
puckipedia was thinking of a few interesting ways to fix this
#
puckipedia e.g. having a database table that is just a mapping from short IDs to context URIs
#
puckipedia and then having those short IDs be global, and just import the contexts that each item uses
#
cwebber2 puckipedia: the usual json-ld way is to expand an incoming document, then compact it to your local @context, and use that @context everywhere
#
cwebber2 but it's slightly trickier when @language values are in the context
#
cwebber2 I'm not sure what to do there, but I guess I haven't played with it enough
#
puckipedia hmmm
#
jaywink Cwebber2: only a comment I guess re interoperability. Whoever implements AP fully server to server will for me imho show the way. Still waiting for someone to do that.
#
cwebber2 jaywink: :)
#
puckipedia looks at their ActivityPub server
#
jaywink puckipedia, link?
#
puckipedia https://github.com/puckipedia/Kroeg
#
Loqi [puckipedia] Kroeg: An ActivityPub server in C#. Also does OStatus for fun.
#
jaywink Ok. How does kroeg do delivery server to server?
#
puckipedia https://github.com/puckipedia/Kroeg/blob/master/Kroeg.Server/BackgroundTasks/DeliverToActivityPubTask.cs it sends the plain object
#
puckipedia I shooould probably change that to send an unflattened object
#
puckipedia fixed
#
puckipedia it now unflattens to a depth of 3 before sending the delivery
#
puckipedia going to bed now, bye
#
jaywink ? yeah.