#social 2023-09-13
2023-09-13 UTC
ajordan, capjamesg, timbl, omz13, lanodan, Raito_Bezarius, benaiah, ramsey, englishm, rhiaro, nhk-ryo, mro, vt, tantek, pfefferle, pchampin and bumblefudge__ joined the channel; nhk-ryo and vt left the channel
# 
bumblefudge__ also activitystream stuff not included
# bumblefudge__ present+
bumblefudge___, bumblefudge_ and by_caballero joined the channel
# by_caballero q+
# by_caballero q-
# by_caballero q+ to mention the branching/dependency structure of the yaml
by_caballero joined the channel; by_caballero left the channel
# by_caballero thanks for that! i was so confused
vt, pfefferle, dmitriz, eprodrom, tantek, sandro and Zakim joined the channel; vt left the channel
# pchampin Zakim, start meeting
RRSAgent joined the channel
# 
RRSAgent logging to https://www.w3.org/2023/09/13-social-irc
# pchampin meeting: SocialWeb data portability - TPAC 2023
# eprodrom scribe: eprodrom
# eprodrom chair: dmitriz
# eprodrom dmitriz: this topic is very popular. A criticial feature of interoperability is data portability between ActivityPub instances
# eprodrom dmitriz: microblogging, forums
# eprodrom dmitriz: migrate accounts (actor profiles), migrate content
# eprodrom dmitriz: including social graph (followers, following), events, content
# eprodrom q+
# eprodrom dmitriz: talk about in-progress specifications
# dmitriz present+
# eprodrom present+
# pfefferle present+
# pchampin present+
# dmitriz scribe+
# 
tantek Social Web Incubation Community Group (AKA SocialCG / SWICG) https://www.w3.org/wiki/SocialCG. Chat log: https://chat.indieweb.org/social/ currently W3C TPAC session: https://www.w3.org/events/meetings/46ce9082-710a-4fe2-8cf8-5dcdc207c877/
# dmitriz eprodrom: One agenda item I'd like to add is - I did a portability report https://w3c.github.io/activitypub/data-portability-report.html
# eprodrom eprodrom: propose reviewing https://w3c.github.io/activitypub/data-portability-report.html
n8s joined the channel
# dmitriz eprodrom: the goal of this report was not to propose any new data portability systems, but to cover the state of data portability in the AP world, today in 2023
# dmitriz ... in particular, this would be guidance for developers and end users, to understand what their options are
bumblefudge__ joined the channel
# bumblefudge__ https://w3c.github.io/activitypub/
# dmitriz ... first section, I gave a probably exhaustive catalogue of what 'Your Data' might mean
# dmitriz ... in two major sections. When we talk about a federated system, there is the data that exists on your own account server,
# dmitriz ... but then of course there's the data on other servers
# dmitriz ... so, on your own servers, there's your actor id / identity, various URIs, actor profile properties (name, avatar, links, etc)
# dmitriz ... and the profile URI (which can be distinct from actor id)
# dmitriz ... you have the Outbox (a collection of all activities the actor has done), and also, as part of that, are reactions (each activity or created content can have a creation of Replies, Likes, and Shares)
# dmitriz ... the Inbox is everything that the actor has received through subscriptions or direct messaging. There are uploaded files. Followers and Following collections
# dmitriz ... there's the Blocks collection (of blocked users), and also a Public Key that's used for HTTP Signature requests.
# dmitriz ... what is distributed on other servers: the user's Actor ID is in others' Followers/Following lists.
# dmitriz ... there are links in various @mentions etc. Object IDs that track back to their canonical location
# dmitriz ... as well as the URIs for uploaded files
# dmitriz ... so the upshot here is -- we have a LOT of distributed data, that either lives on the account's server, or distributed among other servers
# dmitriz ... in terms of data portability, we have two main mechanisms today for doing Data Portability
# dmitriz ... one is Domain-Based Data Portability (when you own your own domain), which lets you transfer implementations, hosting services, etc.
# dmitriz ... and the second one is Mastodon's (& others) Move action
# dmitriz ... which is a technique of moving accounts and activities to other implementations or instances
# dmitriz ... the analogy I always use is - similar to WordPress
# pfefferle nice example :)
# dmitriz ... you can Export your content, comments, accounts, etc, and then Import it on another implementation or another domain, etc.
# dmitriz ... you can do something similar on a Mastodon server on the Fediverse.
# capjamesg present+
# dmitriz ... but the overall pattern is important. Fairly basic, but critical
# dmitriz ... there are a number of limitations with Domain-Based portability. 1) you have to own the domain, and 2) You have to run your hosting server on the fediverse
# dmitriz ... there are a number of hosting services that offer this, but you do need to use one that lets you map a domain onto the hosting system
# dmitriz ... this is a fairly high barrier to most users
# dmitriz ... the other challenge is -- we don't have a standard format for Exports / backups
# dmitriz ... we have a start on this on Mastodon - it outputs a single user's data. But there is not Import on the other side, ironically
# dmitriz ... we have an open Github issue on this
# tantek q+ to note challenges with AP/AS @-mention interop/portability https://github.com/snarfed/bridgy-fed/issues/493 and consider BAF for backup format
# dmitriz ... lastly, different implementations may use different URI patterns for Activities and content. For example, Mastodon's profile is at /users/username, other impls may use other patterns
# dmitriz ... so backing up and restoring, might not map correctly
# capjamesg Not AP-specific, but micro.blog has experimented with a "blog archive format" for blog content that has a HTML format with microformats. Not directly applicable, but may be interesting: https://indieweb.org/blog_archive_format.
# dmitriz q?
# eprodrom ack eprodrom
# dmitriz tantek: no questions about technique per se,
# bumblefudge__ apologies for screensharing the wrong report!
# dmitriz tantek: I appreciate the framing of the scope of the document. I think that literally what you said belongs in the abstract of the doc
# dmitriz ... so like, just copy/paste what was scribed, into abstract! :)
# dmitriz ... I appreciate you mentioned the React mentions in particular, interop wise
# dmitriz ... to be specific, the way Mastodon and Friendica do @-mentions, Bridgyfed has tried the examples in the spec & some variants, and has been unable to get @-mentions to work
# dmitriz ... for example on Bridgy, when I @-mention you, it's not able to generate a correct activity or object to notify you, specifically
# dmitriz ... I'm not sure if that's in scope, but it's one of the issues
# capjamesg Good point, Tantek.
pfefferl_ joined the channel
# dmitriz eprodrom: yeah, I think the general idea with portability, is that the @-mention stays, links to the old domain, but that hopefully redirects you, etc.
# capjamesg Yeah we should decouple domain portability vs. data / instance portability.
# dmitriz tantek: agreed, yeah, and there's still various challenges there
# dmitriz tantek: lastly, you mentioned there's no official backup format. There's a challenge there with blogs in general
# dmitriz ... there's a default meme in general "Just use RSS", but that doesn't always work
# dmitriz ... so there's one specific format that (?) Reese worked on - blog archive format
# dmitriz ... would that suit the needs for account backup & restore?
# dmitriz q+
# dmitriz ack tantek
# 
Zakim tantek, you wanted to note challenges with AP/AS @-mention interop/portability https://github.com/snarfed/bridgy-fed/issues/493 and consider BAF for backup format
# capjamesg I like the plain text, structured, ZIP-based format of the blog archive format.
# dmitriz tantek: it also includes replies, responses, etc
# dmitriz ... its' very close to modern AS2 profile
# capjamesg https://wordpress.com/support/export/
# capjamesg ref ^
# dmitriz dmitriz: also worth looking at WordPress export format
# capjamesg IIRC WP format was a lot of XML?
# pfefferl_ it is simply an RSS file with some namespace extensions
# dmitriz tantek: I think it informed the blog archive format, too? I think it does more / is superset
# capjamesg More semantics are included in blog archive format.
# eprodrom q+
# bumblefudge__ presumably wordpress has its own actor/identity model baked in?
# dmitriz q-
# dmitriz @bumble - yeah
# bumblefudge__ -1 for XML :D
# dmitriz eprodrom: I want to mention the fact that - I believe that Mastodon uses is a collection of AS2 data
# dmitriz ... and since that's the native format for AP, it makes a good candidate
# dmitriz ... so it's a good starting point
# capjamesg Worth discussing re: AS-encoded data for an archive format.
# dmitriz ... I'd like to talk about this second data portability technique that is used on the Fediverse right now, and that's the Move action
# dmitriz ... it's a mechanism that's used for - primarily used to move from one Mastodon instance to another. However, there are other implementations that support the technique
# dmitriz ... it is limited in scope, but has some good outcomes that cover data portability
# dmitriz ... the mechanism is relatively simple: a user has an existing Actor at username@oldexample, then creates a new actor at username@newexample
# dmitriz ... then add an 'alsoKnownAs' property, points at the old actor
# tantek FYI micro.blog has some support for Move: https://www.manton.org/2022/12/02/moving-from-mastodon.html
# dmitriz ... which denotes "I'm wiling to accept Move requests from old actor"
# bumblefudge__ wonders if maybe forum-based (lemmy/kbin/discourse/etc) content export should be considered separately as a corner case for export, and/or an extension/separate FEP? https://codeberg.org/fediverse/fep/src/branch/main/fep/d36d/fep-d36d.md
# dmitriz ... similarly, adds a 'movedTo' property to the old actor profile
# bumblefudge__ maybe we come back to that at end of session if time allows, not urgent/top-of-mind for me
# dmitriz ... then finally, initiates a Move activity, from the old account to the new account
# tantek FYI2 and Bridgy Fed is working on Move support: https://github.com/snarfed/bridgy-fed/issues/330
# dmitriz ... this goes out to all followers, who check & validate, then unfollow the old account, and follow the new account
# dmitriz ... once this happens, the old account's profile URI _will_ automatically redirect to the new account's URI
# dmitriz ... there is not an automated mechanism for moving the Following list. But, mastodon allows downloading the old following list, and importing it into the new account
# dmitriz ... it's a manual step, to download a CSV file & re-import it
# dmitriz ... the results are - we have the redirect between profile URIs, the follwers list is close to what was at the old account. the following list does the export/import thing.
# dmitriz ... the old account's Following list is empty, and the network has everything updated
# dmitriz ... this technique is primarily moving the social graph. It does not cover the content, uploaded files -- those remain at the old URLs at their old IDs
# dmitriz ... if the old account's server is down, it is no longer possible to move to another account
# dmitriz ... and it does not work if the old server's account is blocked by followers
# capjamesg Interesting.
# dmitriz ... so a common occurrence -- a server gets massively de-federated, users then try to move from that server, but are unable to, because the old server is not accepted (the Move actions bounce)
# capjamesg I haven't read many docs about that. Worth highlighting more.
# dmitriz ... so, if it gets de-federated due to a small percentage of bad actors, but it traps everyone else
timbl joined the channel
# bumblefudge__ defederation is the chokepoint of today's fediverse and breaks all kinds of UX and features (not just this)
# capjamesg That feels *really* important to address.
# bumblefudge__ ^^
# dmitriz ... I'll pause here. that covers the primary mechanisms for data portability on the Fediverse right now
# capjamesg q+
# dmitriz ... I know that Firefish and Friendica have somewhat enhanced mechanisms for this Move practice that may move over some content, or mirror some content. I haven't tested it out, but would love to track these
# eprodrom ack eprodrom
# bumblefudge__ q?
# eprodrom capjamesg: point made with regard to people being isolated warrants a great discussion.
# eprodrom capjamesg: being left in that predicament is difficult for the average user
# dmitriz q+
# dmitriz ack capj
# capjamesg q-
# eprodrom tantek: should we track this problem with an issue?
# capjamesg Thank you, kindly!
# eprodrom eprodrom: yes
# eprodrom bumblefudge__: there are feps to address
# dmitriz ack dmitriz
# bumblefudge__ https://codeberg.org/fediverse/fep
# eprodrom dmitriz: feps = Federation Extension Proposal
# bumblefudge__ ^ Extensions to the Spec (defining behaviors and/or data models not defined in the core spec)
# eprodrom dmitriz: there are 2 interesting feps - identity proofs, signed objects impact migration
# eprodrom dmitriz: main challenge if old server is dead or uncooperative or defederated enough that on-line dataportability will not work
# eprodrom dmitriz: to address these challenges, alternate technique to do this cryptographically
# eprodrom dmitriz: signatures prove equivalence
# bumblefudge__ (insofar as signature can be checked by discovering key material for that actorID...)
# eprodrom dmitriz: use a key or DID, sign old account, set up movedTo and alsoKnownAs, sign new account
# bumblefudge__ (and insofar as new server lets you BYO key :D )
# eprodrom dmitriz: can claim a continuation of identity
# eprodrom dmitriz: can perform most of the activity based on equivalents
# eprodrom q+
# eprodrom dmitriz: only possible if user has key control, or old server lets you export your keys
# eprodrom dmitriz: warning and red flags on exporting private keys
# dmitriz scribe+
# eprodrom ack eprodrom
# dmitriz eprodrom: I do want to note that although the technique - the FEP covers doing the signature, it does not cover the Move protocol
# dmitriz ... the implication is - you can USE it for Move activity, but doesn't describe the full procedure
# dmitriz ... so it needs more work / additional specifications of that
# dmitriz ack tantek
# capjamesg Thank you dmitriz for highlighting the FEP.
# eprodrom tantek: fascinating proposal. Uncovers a lot of really good use cases. Should be documented independent of a solution
# bumblefudge__ dmitri: 3 diff failure cases: old server down, old server uncooperative, or older server defederated enough to not be cooperated with
# capjamesg Have we discussed what happens if a domain relapses and is taken over?
# dmitriz ohhh good point (re read-only servers)
# capjamesg (maybe this has never happened, but just thinking about a fail case)
# bumblefudge__ ooooh read-only server is a corner case i'd not thought of
# eprodrom tantek: dead server, uncooperative server, read-only server, defederation
# dmitriz @capjames - also good variation (I think that's a variant of "old server is down" or uncooperative), but it's a different flavor - a possibly actively hostile server
# eprodrom tantek: do you mean one or both servers are defederated (to v from, from v to, both)
# capjamesg dmitriz Yeah. Is there a case where a server becomes "untrusted" due to malicious activity?
# eprodrom tantek: read-only: archive server like Wayback Machine backup
# dmitriz @capjamesg - definitely!
# dmitriz eprodrom: that IS a technique people use for blogs, websites, etc. of scraping the Wayback Machine, and using that to recover data
# eprodrom tantek: useful to document these approaches
# eprodrom eprodrom+
# dmitriz +1 to documenting the use cases!
# capjamesg +1 Tantek.
# capjamesg ++ DID is key to bsky.
# eprodrom tantek: Blue Sky considered data portability from scratch uses a DID
# eprodrom tantek: supposedly portable across servers, without actual proof since no other servers
# dmitriz good call (doing an A/B analysis on Blue Sky's identity approach and this FEP)
# dmitriz q?
# eprodrom tantek: is there a possibility of a bridge?
# bumblefudge__ q+
# eprodrom q+
# eprodrom bumblefudge__: the way Blue Sky uses DIDs is something we have been following, also Nostr. we == DID people
# capjamesg ref: https://atproto.com/specs/did
# capjamesg > The AT Protocol uses Decentralized Identifiers (DIDs) as persistent, long-term account identifiers
# eprodrom bumblefudge__: BlueSky and Nostr are private key == identity, retrofit compatibility onto URI-identity
# dmitriz q?
# dmitriz ack bumbl
# eprodrom bumblefudge__: identity proofs FEP tries to be DID neutral, should not matter what DID method you use, can't think of DID methods won't work, including Blue Sky DID Web, DID PKH (Ethereum wallet style), if delegate to wallet
# bumblefudge__ q-
# dmitriz eprodrom: we haven't taken note of it during this conversation, but I do want to talk about it quickly, which is - the topology of the Fediverse today
# dmitriz ... in terms of the locus of control
# dmitriz ... I think there is a structure/topology that is common on the Indieweb, in which a single implementation supports a single user, on a single domain
# capjamesg Good assessment re: IndieWeb.
# dmitriz ... all under control of that single user. On the Fediverse today (so, Mastodon, Pleroma, Firefish, etc), we have a different common topology
# dmitriz ... which is - hundreds/thousands/more of users, who use a single domain, they have a weak affinity of that domain (picked it from an arbitrary list),
# pchampin waves at sandro :)
# dmitriz ... with a volunteer admin, not a paid service
# capjamesg Agreed.
# dmitriz ... this is not universal, but very typical. Because of the low affinity between user and domain service, the requirement of portability (of moving from server to server), it's very common
# dmitriz ... so in Mastodon world, it's very common to talk to people who have moved 5-6+ times
# dmitriz ... so the needs for portability is extra high
# capjamesg Agreed.
# dmitriz q?
# capjamesg Blog posts!
# dmitriz ack eprod
# eprodrom ack eprodrom
# dmitriz ack sandro
# dmitriz ohhh man, "backups you don't test are useless" --- +1 !!!
# eprodrom sandro: a user requirement. backups that you don't test are useless. Need to know that social backups are working, without breaking or undoable effects
# eprodrom sandro: common example is for users on social services to have backup accounts on same service
# bumblefudge__ sandro, WILL YOU END FINSTA
# dmitriz q+
# capjamesg Mirroring feels complicated to the average user?
# eprodrom sandro: excellent for peace of mind
# eprodrom dmitriz: what additions will we need to add to the core data model
# eprodrom eprodrom+
# eprodrom q+
# dmitriz q-
# dmitriz ack eprod
# dmitriz eprodrom: so, I think one mechanism that Sandro alluded to here - having the ability to back up the content that's created. Preferably "hot backups"
# dmitriz ... of activities, uploaded files, social graph
# dmitriz ... and I think Sandro was suggesting having an "alt" account type of hot backup. The other thing is of course, a static backup on an external storage
# dmitriz ... and having the location of that backup travel with activities & files as they go around the network
# dmitriz ... so that if the content is inacessible, you can try it at this alternate location
# capjamesg q+
# eprodrom q-
# dmitriz ack capj
# bumblefudge__ q+
# eprodrom capjamesg: writing blog post about all thought processes on this issue, let's collaborate on proposed additions
# dmitriz ack bumbl
# eprodrom q+
# capjamesg ++
# eprodrom bumblefudge__: we should start FEPs based on use cases, not solutions
# dmitriz q+ to ask where user stories should live
# capjamesg Absolutely.
# eprodrom +1 to use case first dev
# dmitriz q-
# capjamesg We should be use-case drvien.
# capjamesg *driven
# bumblefudge__ q-
# capjamesg q+
# dmitriz eprodrom: with 5 mins left in the session, I'd like to make a possibly controversial proposal -- just as we created a task force around Testing, would it make sense to take on this fairly large chunk of functionality, around Data Portability, as a task force of the Social CG?
# dmitriz ack eprod
# dmitriz ack capj
# eprodrom capjamesg: good to have a lead spoc for data portability
# capjamesg I'd start with a prior art review too.
# capjamesg Blog archive format, WP format, others of which we may not be aware but are used (Squarespace maybe etc.)?
# eprodrom tantek: excellent focus area for attracting group attention; broaden to include identity/account, data/posts, social graph
# dmitriz @capjamesg -- agreed, yeah
# eprodrom tantek: should handle both of those examples
# dmitriz q?
# dmitriz ack tantek
# capjamesg Why thank you :)
# eprodrom q+
# capjamesg gives thanks to Dmitri for chairing the session.
# eprodrom q-
# eprodrom tantek: backups should be in scope for data portability
# dmitriz whoops didn't mean to --
# dmitriz @capjamesg ++
# capjamesg welcomes rhiaro
# dmitriz eprodrom: next step would be to find a volunteer lead
# capjamesg A discussion for the mailing list
# eprodrom RRSAgent, make minutes
# RRSAgent I have made the request to generate https://www.w3.org/2023/09/13-social-minutes.html eprodrom
# eprodrom Zakim: end meeting
# eprodrom Zakim, end meeting
Zakim left the channel
# RRSAgent I have made the request to generate https://www.w3.org/2023/09/13-social-minutes.html Zakim
# capjamesg Thank you, everyone!
tantek and dmitriz joined the channel
RRSAgent left the channel