2017-11-21 UTC
# 00:26 Loqi [Gargron] Author deletes toot -> "Delete" goes out to followers + anyone who reblogged -> servers of those who reblogged forward the "Delete" to all of their followers, reblogs are deleted and "Undo->Announce" is also distributed to all of their followers
I... # 00:31 cwebber2 it turns out that revocation is the hardest problem in computer science bengo joined the channel
# 00:43 cwebber2 bengo: btw not sure if you read above but (too late for this revision of AP) but I think I have a backwards-compatible version of this forwarding mechanism and many other things which I think would be much nicer # 00:43 cwebber2 and that would be to have collections of actors themselves have an inbox # 00:43 cwebber2 in this case, you just send to the collection's inbox and let it do the work # 00:44 bengo That sounds good. # 00:44 cwebber2 the nice thing is that we could implement this and just have servers that see the inbox property do that, otherwise do the annoying forwarding thing # 00:44 bengo In distbin it never checks if anything is an 'Actor' # 00:44 bengo it only worries about whether things have the right properties # 00:44 bengo Sometimse it verifies things are Activities or Links. # 00:45 bengo Now I have it in TypeScript so a lot of that is in the type system # 00:45 bengo I will publish ts libs for activitystreams and activitypub once things are more robust # 00:48 bengo lol I have a lot of OneOrMore<LDValue<ActualType>> # 00:48 cwebber2 after this week wraps up I'm working on a hybrid graph/object store system for AP :P # 00:49 cwebber2 though puckipedia tells me Kroeg is already such a system ;) # 00:49 bengo wait wait wait. I still haven't seen ANYONE do an interoped reply # 00:49 bengo can we PLZ get that working first lol # 00:50 cwebber2 well I have that running in pubstrate and so does puckipedia in kroeg # 00:50 bengo And it works between the two? # 00:50 cwebber2 but I haven't tested against theirs because I don't do http sigs yet :D # 00:51 bengo Just saying I think we have like 3 things now that interops with itself but not each other. # 00:51 cwebber2 bengo: nothing running publicly, though I can spin it up # 00:51 bengo And no public URLs to prove it # 00:51 bengo So I think that will be a big milestone. # 00:52 bengo puckipedia I've taken care of some issues you interopping with Kroeg. But will you meet me in the middle by doing your own testing? I'll spend some more hours on it over Thanksgiving break. # 00:54 bengo seems like you're saying it's there for the reason I worried about # 00:55 bengo Like when I reply to Alyssa I get no say about whether she will share my message with whoever she wants to deem her follower # 00:55 bengo let me re-read this part of spec # 00:55 cwebber2 bengo: you get no say about this in general, since anyone can copy information around, no matter what the DRM people will tell you :) # 00:55 cwebber2 but yes we need to be careful about how we advise Alyssa to forward! # 00:56 bengo snark aside, it's a design decision # 00:56 cwebber2 note that the inbox route handles this a lot more nicely # 00:56 bengo Like... if we just removed inbox forwarding what would we lose? Is it at risk anyway? # 00:57 bengo It would just be left to F1 and B how to share data in this case, which I think is great. A is not encouraged to share everything # 00:57 cwebber2 bengo: this is a real actual problem encountered in practice tho # 00:57 cwebber2 people complained about this in statusnet and pump.io and mastodon all the ttime # 00:57 cwebber2 our solution was based on some adjustments from how diaspora does it # 00:57 cwebber2 which is to have the toplevel of a thread do all delivery # 00:57 bengo Ah and I realized # 00:58 bengo I guess B can just *not* cc the followers collection # 00:58 bengo as a way of opting out # 00:58 bengo OK then my concern is assuaged and I am thankful for your patience. # 00:59 bengo do you have a trimmed list of things at risk other than the zerobin from last WG meeting? # 00:59 bengo Also puckipedia does Kroeg *really* implement all of the features? If so, wow! great job # 01:00 bengo cwebber2 ok great then wtf am I not outside lol # 01:01 cwebber2 it's not even on AP itself since it's something I left open to not forget to add to the wiki # 01:01 bengo cwebber2 do you have any public AP-speaking URLs I can test against? # 01:03 cwebber2 bengo: aside from the one in the test suite, no... but I can put up mine if you really want # 01:03 cwebber2 bengo: can I promise that I'll have it up and I'll actually be using it 1 month from today? :) # 01:04 bengo I mean it's only useful to me insofar as we actually get a repeatable working public federation demo. # 01:04 bengo distbin.com is public to anyone so it should be easy to use to test with or demo ad infinitum # 01:04 bengo I just want the other side # 01:04 bengo But yeah I'll hold you to it # 01:04 bengo Also happy to schedule some time to just do it synchronously on a call or something # 01:04 cwebber2 bengo: kroeg and smilodon and mastodon have been testing against each other # 01:05 bengo ya I guess me too :) # 01:05 bengo I know I can clone mastodon... that's what I should just do nicolasparada joined the channel
# 03:08 rowan hey anybody got any idea why posts from smilodon aren't showing? it gets the count from the OrderedCollection but not the posts themselves. sample here: https://pastebin.com/hTPUZGFv # 03:13 bengo I have some work to do # 03:15 rowan i know that feel timbl joined the channel
# 03:40 bengo who has mastodon profilse I can follow? :) # 03:41 rowan @rowan@toot.cafe # 04:07 cwebber2 bengo: looks like it's down at the moment, but totally follow @cwebber@octodon.social :) # 04:07 bengo I'm upgrading to 2.0 cdchapman joined the channel
# 04:31 rowan in a dev instance, what's the best way to clear mastodon caches? htrobinson, xmpp-social, rowan_, rowan, DenSchub and timbl joined the channel
# 14:54 puckipedia rowan: hm, how do you mean 'cache', do you mean the local representatino of the user? # 14:54 rowan puckipedia: yeah # 14:55 rowan tryna avoid adding a million users to test things, y'know # 14:57 puckipedia hm, I thought there'd be a button for that # 15:00 puckipedia okay so what works: bundle exec rails console # 15:00 puckipedia Account.find_by(username: 'test').destroy! # 15:00 rowan ahhh you're a hero # 15:02 puckipedia I just quickly looked at if .destroy! would work, then used Account.find_by # 15:02 rowan i'll learn rails someday :) eprodrom, cdchapman and bengo joined the channel
# 17:07 ajordan started off as a pump.io fork but then I just gave distopico write access to upstream so # 17:34 eprodrom That's good to hear # 17:37 cwebber2 eprodrom: ready to merge your impl reports whenever they're ready! :) # 17:37 puckipedia oh wait it is tuesday # 17:44 xmpp-social [ajordan] aaronpk: yikes? Question mark? # 17:45 eprodrom cwebber2: just added the issues # 17:45 eprodrom Last possible second # 17:45 eprodrom cwebber2: I implemented http signatures last night # 17:45 eprodrom It was less hard than I'd thought it would be # 17:46 puckipedia now implement the Digest: header :P # 17:46 eprodrom And the NodeJS library for http sigs handles most of the header-ordering hassle # 17:46 eprodrom That said, I haven't tested interop with any other server, so ??? # 17:47 eprodrom Argh my Docker build failed # 17:47 eprodrom I have to say, I really love Gitlab's CI/CD pipelines # 17:48 eprodrom I have not yet gotten it to push to production, but... soon # 17:49 eprodrom I knew that would be the case # 17:51 eprodrom Yeah that's fine # 17:51 eprodrom ajordan: didn't you do an IR for pump.io? # 17:52 eprodrom Now or never, man # 17:52 eprodrom It only takes a few minutes # 17:52 ajordan I don't think it even checks any of the AP boxes, just AS2 # 17:52 rhiaro side not: not never. Can keep 'em coming in during PR # 17:53 rowan fwiw mine's very immature and it's in # 17:53 rhiaro and for all time so long as cwebber2 is running the server # 17:53 rhiaro I mean, now is the important time but having more for REC doesn't hurt # 17:53 puckipedia showing that there's people working on it # 17:54 csarven will submit an AP report.. if not within the next days, he permits all to vegan-egg his flat # 17:55 cwebber2 *dusted with a slurry of nutritional yeast, tofu, and black salt* # 17:55 eprodrom cwebber2: whoa # 17:55 eprodrom You just closed my authorization question # 17:56 eprodrom OK, so, I'm looking further # 17:56 eprodrom We're using the Web Payments vocabulary for public keys? # 17:57 rhiaro ughhh I don't know if I can stay that long :s # 17:58 cwebber2 let's try to get to spec votes within the hour or at least pre-emptively get spec votes if we get to the point where some people have to leave tantek joined the channel
# 17:59 csarven would like to preemptively vote if that's okay. On another call :S RRSAgent joined the channel
# 18:01 rhiaro ARGH i'm on a different laptop and everything is b0rked # 18:06 eprodrom I think by definition no # 18:07 eprodrom ajordan: pretty sure that's me # 18:07 rhiaro does not have access to all of her usual resources # 18:07 eprodrom ajordan: updated # 18:08 Loqi eprodrom has 51 karma in this channel (52 overall) # 18:08 rhiaro ajordan I know but i don't know if it's still right # 18:08 rhiaro webex is telling me it's wrong but also this phone seems to be glitchy # 18:09 Zakim Present: tantek, cwebber, npdoty, annbass, hadleybeeman, snarfed, torgo, rhiaro, ajordan, eprodrom, aaronpk, bengo, sandro, get-webmention-url@1.0.0, +, tsyesika, csarven # 18:09 Zakim sees on irc: RRSAgent, tantek, bengo, cdchapman, eprodrom, DenSchub, xmpp-social, htrobinson, dlongley, JanKusanagi, dlehn, bwn, sandro, rhiaro, distopico, Loqi, sknebel, ajordan, # 18:09 Zakim ... bigbluehat, surinna, bitbear, howl, dwhly, tsyesika, nightpool, trackbot, puckipedia # 18:09 Zakim ... csarven, flackr, KjetilK, hadleybeeman, Zakim, Chocobozzz, aaronpk, er1n, cwebber2, jungkees, raucao, saranix, erincandescent, jet, ben_thatmustbeme, Gargron, melody, mattl, # 18:10 Zakim Present: tantek, cwebber, npdoty, annbass, rhiaro, ajordan, eprodrom, aaronpk, bengo, sandro, tsyesika, csarven # 18:10 Zakim On IRC I see RRSAgent, tantek, bengo, cdchapman, eprodrom, DenSchub, xmpp-social, htrobinson, dlongley, JanKusanagi, dlehn, bwn, sandro, rhiaro, distopico, Loqi, sknebel, ajordan, # 18:10 Zakim ... csarven, KjetilK, hadleybeeman, Zakim, Chocobozzz, aaronpk, er1n, cwebber2, jungkees, raucao, saranix, erincandescent, jet, ben_thatmustbeme, Gargron, melody, mattl, # 18:10 Zakim ... bigbluehat, surinna, bitbear, howl, dwhly, tsyesika, nightpool, trackbot, puckipedia # 18:10 sandro (chair does "Present-" on all the folks who are listed as Present by Zakim from TPAC, but are not actually on today's call) # 18:11 Zakim Present: tantek, cwebber, rhiaro, ajordan, eprodrom, aaronpk, bengo, sandro, tsyesika, csarven # 18:11 Zakim On IRC I see RRSAgent, tantek, bengo, cdchapman, eprodrom, DenSchub, xmpp-social, htrobinson, dlongley, JanKusanagi, dlehn, bwn, sandro, rhiaro, distopico, Loqi, sknebel, ajordan, # 18:11 Zakim ... bigbluehat, surinna, bitbear, howl, dwhly, tsyesika, nightpool, trackbot, puckipedia # 18:11 Zakim ... csarven, KjetilK, hadleybeeman, Zakim, Chocobozzz, aaronpk, er1n, cwebber2, jungkees, raucao, saranix, erincandescent, jet, ben_thatmustbeme, Gargron, melody, mattl, # 18:11 rhiaro when the phone thinks the code is right it then asks me for a pin # 18:11 eprodrom rhiaro: you are having a hard time with this # 18:12 rhiaro "you have not set up a pin for your account. Thanks for using webex. Fin." # 18:12 hadleybeeman waves from another meeting. Have a good call! # 18:12 eprodrom rhiaro: I think you're IRC-only this week # 18:15 sandro ajordan, you here but need to leave at the hour? # 18:16 sandro aaronpk: I made it through the issues, all editorial # 18:17 sandro aaronpk: Commenter said he's suprised Link header is used in Request. Typically in response. # 18:17 sandro aaronpk: I replied I didn't know why it wouldn't be, and this is how PuSH has worked for a long time # 18:18 eprodrom Not listed there I don't think # 18:20 sandro sandro: as I recall, he explained yes, you can use Link headers on requests. It's not specified in the RFC, which means both. rowan joined the channel
# 18:20 eprodrom This is surprisingly complicated # 18:21 eprodrom " This specification updates the "Message Headers" registry entry for "Link" in HTTP [RFC3864] to refer to this document." # 18:22 sandro sandro: me neither. Do we use the "target IRI" terminology # 18:24 sandro aaronpk: Did something relevant in the link registry change? # 18:25 sandro tantek: Let's not change the reference if that doesn't solve the issue # 18:29 sandro sandro: That list is not supposed to be exhaustive. Link header isn't listed either? # 18:30 sandro aaronpk: It's not listed among response headers # 18:31 sandro sandro: Yes, I did, and he said it was fine. (like 7 years ago) # 18:31 eprodrom I think we can actually summon @mnot to the issue if we want # 18:32 sandro tantek: Do we need to add a note to the spec? # 18:32 tantek PROPOSED: resolve 142 with comment that LINK header is not disallowed by 7231 or 5988, and informally, the author of 5988 (Web Linking) intended it to be allowed. # 18:32 sandro eprodrom: No, it's fine, given this discussion # 18:33 sandro tantek: And we can ask the commenter if they want a clarifying note added to the spec # 18:33 tantek RESOLVED: resolve 142 with comment that LINK header is not disallowed by 7231 or 5988, and informally, the author of 5988 (Web Linking) intended it to be allowed. # 18:33 Loqi [aaronpk] #138 different hub for same topic if denied # 18:34 sandro aaronpk: This is about Location header on request # 18:35 sandro .. I think this is something WebSub is defining # 18:35 sandro .. using the Location header to say your subscription was denied, but you can try using this other URL # 18:35 eprodrom julien-aaronpk.json # 18:36 eprodrom julien-eprodrom.json # 18:36 sandro eprodrom: I think the idea would be, you ask for one, a capabilities URL, and you get a different one # 18:37 sandro sandro: or is this conneg? or no, that would be Content Location # 18:37 sandro eprodrom: This was thrown in a long time ago. I've never seen this implemented. # 18:38 sandro aaronpk: julien says he knows this was used once but he's not seen it in ages # 18:38 sandro sandro: so this is for access controlled URLs? # 18:39 sandro tantek: we don't know if anyone has implemented this # 18:40 sandro aaronpk: It's a MAY, not in conformance criteria # 18:41 cwebber2 notes that they would like a hard cut off on this conversation at :45 # 18:41 cwebber2 15 minutes of people present to discuss going to PR is not a lot of time # 18:41 sandro "Hubs may provide an additional HTTP [RFC7231] Location header (as described in section 7.1.2 of Hypertext Transfer Protocol [RFC7231]) to indicate that the subscriber may retry subscribing to a different hub.topic. This allows for limited distribution to specific groups or users in the context of social web applications." # 18:42 sandro sandro: Can we try hard to find out if anyone cares? # 18:42 sandro sandro: it's kind of a back-door MUST on the subscriber library, as worded. # 18:43 sandro sandro: I think it'll be easier to drop if we can find evidence no one has implemted it # 18:43 ajordan sandro: it seems backwards-compatible to me despite the quasi-MUST? # 18:44 sandro sandro: aaronpk can you ask everyone who has filed an impl report # 18:45 sandro tantek: Okay, that's the best we can do for now # 18:46 sandro .. relaxed from SHOULD to MAY on Liked property # 18:46 Zakim sees ajordan, eprodrom on the speaker queue # 18:48 sandro .. also Forwarding mechanism, said you need to remove bcc # 18:50 sandro eprodrom: the only change that's a bit challenging is the "Liked" property, which was a SHOULD in one spot and MAY in the more detailed spot, so we just resolved it to a MAY. It was an error in previous version. # 18:51 sandro tantek: good to clarify in change long, that you made it consistent # 18:53 sandro sandro: column for number-of-yes's which turns green when it's 2 or more # 18:55 eprodrom +1 let's vote # 18:56 Loqi [Christopher Allan Webber] ActivityPub # 18:59 ajordan alright, gotta go, sorry! thanks all, and woot! we did it # 18:59 sandro tantek: so, we'll want to closely track issues that might come up until then # 19:00 Loqi [aaronpk] #146 At risk: limiting the use of HTML <link> to the HTML <head> # 19:00 sandro aaronpk: We didnt have an issue to track At-Risk # 19:01 sandro tantek: Do any implementations parse it outside the head? # 19:02 sandro sandro: shocked we let an At Risk feature get published in a PR # 19:03 sandro tantek: Any evidence of subscriber implementations? # 19:04 sandro tantek: what do webmention and micropub say about this? # 19:05 sandro sandro: We have to find either (1) a major publuisher putting it outside the head, OR (1) a major subscriber that doesn't look outside the head. # 19:07 sandro aaronpk: There is one HTML tag discovery for subscribers, and it's in the head # 19:07 sandro .. there is NOT a separate test for it showing up in the body # 19:09 sandro aaronpk: webmention also lacks a test for link outside head # 19:09 sandro tantek: procedurally, I feel like because no one commented on this, it should be dropped. # 19:10 sandro sandro: agreed, but what if there is a major subscriber / library which doesn't look outside head # 19:12 sandro sandro: so it may be that everyone publishes in the head, so subscribers might just be looking in the head # 19:12 sandro aaronpk: This is HTML, but *most* implementations are for RSS and/or xml/atom, so this doesn't even come up # 19:15 sandro aaronpk: Three reported looking in HTML. I'll ask them if they look only in head. # 19:15 sandro tantek: Any other proposals? Anyone want to keep the restriction-to-the-head ? # 19:16 tantek PROPOSED: Resolve 146 as if the at-risk feature was dropped in CR to PR per no data on any known implementations of it. # 19:16 sandro +1 contingent on the three respsonses aaronpk said he'd get. If they have a problem then we re-consider # 19:18 sandro eprodrom: I was surprised to see HTML allowed link elements in the body # 19:18 sandro sandro: true, most publishers probably wont use the <body> # 19:19 tantek RESOLVED: Resolve 146 as if the at-risk feature was dropped in CR to PR per no data on any known implementations of it, contingent on aaronpk asking for three responses regarding it. # 19:19 Loqi [aaronpk] #143 hub vs subscriber in section 8 # 19:20 sandro aaronpk: Kind of outside of scope of spec, because it's about publish-hub relationship. # 19:21 sandro aaronpk: they'll have to on expiry, and not until then # 19:22 sandro sandro: So, take this out? Since it's out-of-scope # 19:22 sandro aaronpk: I guess it's in there because the headers may be different than the subscriber originally found. # 19:23 sandro aaronpk: yeah, except the spec says the subscriber MUST NOT use the link headers to identify subscription (must use capability URLs) # 19:24 sandro sandro: let's take out the sentence, more confusing than helpful # 19:25 sandro aaronpk: remove sentence about hub/discover, and change to the topic-url-may-be-different # 19:26 sandro tantek: Is it a ref to something elsewhere in spec? # 19:26 sandro aaronpk: Important part is to warn subscribers that topic may be different # 19:28 sandro ajordan: The topic might still be useful, for dereferencing # 19:28 sandro sandro: Let's rephrase it as "topic url may be different", and see if commenter and julien are okay. this isn't normative, # 19:29 sandro aaronpk: right, this sentence is just providing a (confusing) example of WHY it might change # 19:30 sandro aaronpk: these were the tough ones, lots of easier ones # 19:31 sandro PROPOSED: delete problematic sentence "A hub MAY use discovery from time to time to detect changes in a topic's canonical URL and Hub URLs." and instead explain topic URL might change # 19:33 sandro RESOLVED: delete problematic sentence "A hub MAY use discovery from time to time to detect changes in a topic's canonical URL and Hub URLs." and instead explain topic URL might change # 19:34 eprodrom That was a long one # 19:34 Loqi sandro has 51 karma in this channel (58 overall) # 19:34 Zakim As of this point the attendees have been tantek, cwebber, npdoty, annbass, hadleybeeman, snarfed, torgo, rhiaro, ajordan, eprodrom, aaronpk, bengo, sandro, # 19:34 Zakim ... get-webmention-url@1.0.0, +, tsyesika, csarven # 19:34 tantek so much for my attempt at fixing the present+ # 19:34 Zakim leaving. As of this point the attendees have been tantek, cwebber, npdoty, annbass, hadleybeeman, snarfed, torgo, rhiaro, ajordan, eprodrom, aaronpk, bengo, sandro, # 19:34 Zakim ... get-webmention-url@1.0.0, +, tsyesika, csarven # 19:35 tantek there, next time Zakim should start with clean present list # 19:35 aaronpk now to go braindump everything before I forget it # 19:37 eprodrom cwebber2: you still around # 19:37 eprodrom I have an auth question # 19:37 eprodrom Generating a different pubkey for every user seems :P # 19:38 eprodrom Is there a reason not to use just one key for all the actors? # 19:48 puckipedia eprodrom: imagine if one pubkey got leaked timbl and bengo joined the channel
# 20:15 eprodrom puckipedia: oh no # 20:16 eprodrom A public key? Getting leaked? # 20:16 eprodrom That'd be horrible # 20:16 eprodrom puckipedia: I think you mean what if a private key gets leaked # 20:17 eprodrom Which is a good point, except if you keep all your private keys in the same store, they're all equally vulnerable anyway # 20:17 eprodrom But I get the point # 20:18 eprodrom I don't think it's worth the overhead, though. Generating and storing keys is expensive # 20:18 puckipedia in the future I want the user to be able to supply their own keys # 20:18 eprodrom That sounds nice # 20:18 puckipedia so instead of locally signing, it will return a special status code plus the objects to sign # 20:19 eprodrom So the client signs? # 20:19 puckipedia imagine a decentral ActivityPub :P # 20:29 aaronpk Wait what isn't already decentral about activitypub? # 20:30 puckipedia aaronpk: I mean, decentral object store etc as well # 21:32 cwebber2 "you don't have to generate one key for every user but it's probably a good idea" rowan joined the channel
# 23:09 cwebber2 I'm glad I did the activitypub.rocks page in Haunt; hacking sxml as a templating language is a delight :) # 23:09 ajordan that was one of the most cwebber2 things I've ever heard # 23:10 puckipedia and again, \o/ to all involved rowan joined the channel
# 23:20 cwebber2 I added code so it would red if < 2 on any of them but since that didn't happen :) htrob joined the channel
Gargron i think i summed it up here https://github.com/tootsuite/mastodon/issues/5761#issuecomment-345875480
aaronpk ajordan: I went through all the open issues from last week and there's only a few that I need group input on. managed to make it through most on my own https://github.com/w3c/websub/issues
cwebber2 eprodrom: check it https://activitypub.rocks/implementation-report/
Loqi Strugee made 1 edit to [[Socialwg/2017-11-21]] https://www.w3.org/wiki/index.php?diff=105249&oldid=105223
RRSAgent logging to https://www.w3.org/2017/11/21-social-irc
tantek PROPOSED approve last week minutes: https://www.w3.org/wiki/Socialwg/2017-11-14-minutes
ajordan rhiaro not sure the instructions on https://www.w3.org/2015/10/social-wg-telecon have worked for me for months
sandro aaronpk, skimming https://tools.ietf.org/html/rfc8288#page-24 I don't see any problems
@jaywink @DeadSuperHero Well, it's a must if you want to do reply forwarding to all participants.. AP left it out of the spec but when Mastodon did their implementation they realized the forwarding case is impossible without signatures. Shame it's not in spec.
See https://www.w3.org/TR/activitypub/#inbox-forwarding (twitter.com/_/status/933074905536630784)