Skip to main content
Should developers publish federation protocols that aren't ActivityPub or ActivityPub extensions?

  • Strong yes (25%, 66 votes)
  • Qualified yes (37%, 96 votes)
  • Qualified no (28%, 74 votes)
  • Strong no (12%, 32 votes)
256 voters. Poll end: 2 years ago

Evan Prodromou reshared this.

[old] stefan 🙌
[old] stefan 🙌  
•
In other realms I've seen a great way to influence the most popular tool is to build a brand new tool and show how that new world could work. Then some of the best parts find their way back in to the main tool (if the devs are responsive enough)
Evan Prodromou
Evan Prodromou  
•
@stefanhayden fine, but what if the demo confuses people about the more popular tool, which is orders of magnitude less popular than its bad commercial equivalents?
@[old] stefan 🙌
[old] stefan 🙌
[old] stefan 🙌  
•
personally I find it rare the new thing gets that much traction. I'm looking towards things like coffeeScript and Deno in the javascript echo system. Both have influenced javascript long before they really had wide spread adoption or a chance to confuse novices.
Andreas Rilinger
Andreas Rilinger  
•
Not sure. If it‘s better than ActivityPub and gets a chance to get used especially by mastodon I guess. If nobody uses it, it could be irrelevant.
Evan Prodromou
Evan Prodromou  
•
@andreas it could be confusing and waste a lot of the community's effort and focus at a critical time, too.
@Andreas Rilinger
Olivia Vespera
Olivia Vespera  
•
I don't think i understand the question. what can they make?
Evan Prodromou
Evan Prodromou  
•
@OliviaVespera federation protocols defines how different programs talk to each other. ActivityPub is the protocol that Mastodon uses.

I'm asking, should anyone be working on or promoting protocols that aren't closely-related to ActivityPub right now?
@Olivia Vespera
Olivia Vespera
Olivia Vespera  
•
ah I see. I don't know if there are drawbacks to activitypub at the moment compared to the others.
Bob Wyman
Bob Wyman  
•
Developers of federation systems, who believe that ActivityPub cannot address their needs, should explain why that is and give others the opportunity to propose conforming solutions.

The primary ethical imperative should be to avoid publishing a new protocol unless it is certain that ActivityPub cannot be reasonably extended to address the unmet need. In the unlikely case that a new protocol is appropriate, all should be done to ensure maximum reasonable interoperability.
Johannes Ernst
Johannes Ernst  
•
@bobwyman Have you talked to the BlueSky folks? 🙂
@Bob Wyman
Evan Prodromou
Evan Prodromou  
•
@J12t@bobwyman BlueSky is the prime example right now
@Johannes Ernst @Bob Wyman
Bob Wyman
Bob Wyman  
•
@J12t There is some stuff I like about BlueSky. For instance, their approach to allowing posts to be "labeled" and thus enabling all sorts of downstream filtering. I'd like to see content labeling done by an arbitrarily large number of services. However, labelling isn't something that requires a new protocol. I think it would be relatively easy to add to a Mastodon-like system (although it might introduce a bit of latency.)
@Johannes Ernst
Johannes Ernst
Johannes Ernst  
•
@bobwyman I like a bunch, too. I wonder whether somebody brave could come up with a gradual migration protocol -- possibly practically impossible, but both content storage and identity layer are interesting. E.g. because it appears the cost of the whole "fediverse" at Twitter-scale appears to be potentially much lower with the BlueSky approach. (haven't done the math, just a hunch)
@Bob Wyman
​
​  
•
Hot take here: ActivityPub is kind of a "worse is better" protocol and has quite a lot of flaws, which is why I'm excited about alternative federated or distributed protocols like https://spritelyproject.org/

However, most of my development work (Gitea federation, ForgeFed, FEPs) are all for ActivityPub and extensions, because that's where the ecosystem momentum is right now.

Spritely

spritelyproject.org
networkException
networkException  
•
ActivityPub extensions yes, if they are clearly scoped (like https://forgefed.org). Non ActivityPub federation protocols also yes if they solve a different problem (like https://matrix.org)

For anything else improvements to ActivityPub or its implementations should be considered first to strengthen the ecosystem rather than causing further fragmentation

Matrix.org

Matrix is an open standard for interoperable, decentralised, real-time communication
Matrix.org
Alan Langford
Alan Langford  
•
Publishing is part of innovation. Should they be widely adopted? That's for the W3C to decide.
Christina Sørensen
Christina Sørensen  
•
should is one of those words that I generally disagree with
Evan Prodromou
Evan Prodromou  
•
@cafkafk ooh then there's an RFC that you're really not going to like
@Christina Sørensen
Rigo Wenning
Rigo Wenning  
•
ok, if @w3c is defunct, there should be some publication. But if now everyone is doing their own ActivityPub in their own corner, will that be interoperable?
@World Wide Web Consortium
Leonardo
Leonardo  
•
not sure what's the point of this. The internet is already full of competing implementations of communication protocols. In the fediverse, ActivityPub is the widely adopted solution now, but it was born just 8 years later the Diaspora* federation protocol, which was (and still is) very not compatible. If a new protocol emerges, and its specifics are robust and open, and it fits as a good solution for a scenario that let its adoption grow, it is more a benefit than a problem in my opinion
Evan Prodromou
Evan Prodromou  
•
@leonardo it's a poll. The point is asking what you think.
@Leonardo
Leonardo
Leonardo  
•
yeah sorry😁, i mean, i don't think a poll is the appropriate tool to discuss the question. I find the replies people gave more insightful
Evan Prodromou
Evan Prodromou  
•
@leonardo

So you're saying the poll stimulated interesting conversation?
@Leonardo
Leonardo
Leonardo  
•
i'm saying that the question was stimulating, and it would probably be interesting even without a poll attached on it. Please don't try to twist my words just to prove a point, i'm not here to argue and i'm sorry if i gave the opposite impression, it was not my intention.
Evan Prodromou
Evan Prodromou  
•
@leonardo oh, great!

I do a *lot* of polls. I love the thought process they call up in me, and I love the way conversations flow out of them. And I love finding out if my opinion is mainstream or if I'm an outlier.

As a #FuzzyLogic fan and implementer, I try to represent shades of truth as best I can in the limited number of options.

If you don't like polls, that's OK! You can follow other people and have a different conversation somewhere else. My stream isn't the only game in town.
#FuzzyLogic @Leonardo
Leonardo
Leonardo  
•
thanks for explaining your rationale, it's way more clear to me now 😀
Zash
Zash  
•
Can haz moar #XMPP and XMPP Extensions‽
#xmpp
Stephen Kellat
Stephen Kellat  
•
Qualified no as I would love to see some of the capabilities of pump.io migrate into ActivityPub. I have yet to see an OpenFarmGame or ih8.it pop up on an ActivityPub backend.
sofia ☮️🏴
sofia ☮️🏴  
•
qualified: only if they want to 😉.
Julien Genestoux
Julien Genestoux  
•
So disappointing :(
We will end up with more new "standards"
Jon
Jon  
•
Strong yes. ActivityPub was developed years ago and the people involved didn't really take harassment or disinfo into account. New systems on new protocols can always provide a compatibility layer to interact with ActivityPub.
Evan Prodromou
Evan Prodromou  
•
@jdp23 Tell me more about what the people involved with ActivityPub considered and didn't consider.
@Jon
Jon
Jon  
•
Fair enough!

Can you point me to the analysis of ActivityPub from disinfo and anti-harassment perspectives?

And who were the disinfo and anti-harassment experts involved in developing ActivityPub?
Evan Prodromou
Evan Prodromou  
•
@jdp23 that is not a good use of my one wild and precious life.

I'm pretty sure minutes from the meetings are up on the web.
@Jon
Jon
Jon  
•
Searching https://www.w3.org/wiki/SocialCG , disinfo never appears in the minutes.

Harassment is mentioned a few times but there is nothing I would call a serious analysis -- for example you say "I'm really sympathetic to desire to avoid harassment ... but systems need to acknowledge that there will be bad actors" which while true is only the very first step.

SocialCG - W3C Wiki

www.w3.org
Oblomov
Oblomov  
•
Qualified no: it's always better to work extending what currently existing than writing something completely different and potentially incompatible, unless there is a significant gain in doing something completely different and potentially incompatible.
aeva
aeva  
•
if someone thinks they have a better idea they're welcome to try, but realistically I think most implementors would end up also implementing activitypub anyway if whatever hot new thing didn't have any support for it
Alexander Bochmann
Alexander Bochmann  
•
Mastodon migrated from OStatus to its ActivityPub dialect, supporting both for some time, and took much of the Fediverse with it.

As long as there's a way to have a transition period or even just a lossy fallback, new protocols can maybe solve problems that an ActivityPub extension can't.
dria :firefox: :redhat:
dria :firefox: :redhat:  
•
you need an "I don't understand the question" option :)
Evan Prodromou
Evan Prodromou  
•
@dria I think that's the "don't answer and go do something else" option.
@dria :firefox: :redhat:
Ryan Makes, Dreamscaper
Ryan Makes, Dreamscaper  
•
I'd like to see a list of active protocol with multiple clients.

git counts, right?
Evan Light
Evan Light  
•
um... I'm not sure this should be a poll. Also, I need so much more context here before voting. Also, that's why I can't see this as a poll.
Evan Prodromou
Evan Prodromou  
•
@elight

1) Then don't answer and don't post a poll on the topic. Problem solved!

2) You should probably start reading then.

3) That probably is, yes.
@Evan Light
Stíofán MacThomáis
Stíofán MacThomáis  
•
If open source is to survive and thrive, then yes, everything should be challenged - survival of the fittest, etc...
Evan Prodromou
Evan Prodromou  
•
So. these are interesting results!

Right now, I'm a strong no. Anything anyone publishes before the end of the year is going to be, at best, a spoiler and a distraction.

Worst case, if it picks up steam, it could split the interest in federation, and really inhibit further growth.
Evan Prodromou
Evan Prodromou  
•
Metcalfe's law says the value of a communications medium goes up with the square of the number of participants.

If the total number of people interested in federated social networks is N, a single pool of people using compatible protocols -- AP, right now -- generates ~N^2 value.

If you split that in 2 even groups, the value of each group is ~N^2/4, and total value is N^2/2.

Every time we split the fediverse, we drive down value significantly.

So, if you care about federation, don't split.
Evan Prodromou
Evan Prodromou  
•
Long-term, sure, go for it. If your protocol is 10 times better than ActivityPub for some reason, it will get some headway.

But don't be a spoiler during this crucial period.
Evan Prodromou
Evan Prodromou  
•
So, anyway: short term, strong no. Long term, qualified no.
Markus Feilner :verified:
Markus Feilner :verified:  
•
I disagree... publish early publish often has always been the strength of our communities. I know what I am talking about, been with linux since 1994 and I've seen a lot, from browser wars to realplayer...
Evan Prodromou
Evan Prodromou  
•
@voxpelli it's great! Bridges are good, interoperability is great. The existing Indie Web stack is great, RSS and RSSCloud is great, existing protocols like OStatus are great. Keep using them, keep building bridges.

But!

I think publishing a *new* protocol in November 2022 or launching a new network is pretty reckless, and I hope whoever thinks of doing that would think hard about whether they're providing a net good for the fediverse and for the world.
@Pelle Wessman
Kevin Marks
Kevin Marks  
•
@voxpelli you didn't say "new" in the poll though, so it was effectively asking other existing protocols to keep quiet
@Pelle Wessman
Evan Prodromou
Evan Prodromou  
•
@KevinMarks@voxpelli

I guess my idea of "publishing a protocol" is that you're releasing a new spec and not revising or updating an existing spec.

You're right, "publishing a protocol" could mean the second, too.
@Pelle Wessman @Kevin Marks
Evan Prodromou
Evan Prodromou  
•
@KevinMarks@voxpelli I don't see much if any harm in updating a protocol spec right now.
@Pelle Wessman @Kevin Marks
Evan Prodromou
Evan Prodromou  
•
@KevinMarks@voxpelli

So, now that I'm rereading, I think there's another potential source of confusion.

In the Web world, we sometimes use "publish" to mean "support" or "implement", especially when we're talking about feed formats.

Like, "Does that site publish a WebSockets interface?"

Or, "My blog engine publishes RSS and Atom."

"This server publishes SMTP and SSH."

That's not what I meant, so I should have picked a clearer term, like "publish a new spec".
@Pelle Wessman @Kevin Marks
Pelle Wessman
Pelle Wessman  
•
@KevinMarks Or even more accurate: Publish a new competing ecosystem

And it’s complex as eg. Mastodon isn’t a pure ActivityPub actor as far as I understand, it keeps the WebFinger part from its OStatus roots and augments that on top of ActivityPub.

Likewise both OStatus and the IndieWeb shares the use of PubSubHubbub/WebSub, so it’s hard to even know where one ecosystem starts and where another ends.
@Kevin Marks
Oblomov
Oblomov  
•
@KevinMarks
I'd say the new was implicit in the publishing. Why would you publish an old protocol?
@evan@voxpelli
@Evan Prodromou @Pelle Wessman @Kevin Marks
Pelle Wessman
Pelle Wessman  
•
@KevinMarks@oblomov Well, the IndieWeb like OStatus and to a degree Mastodon consists of many protocols and more may very well be added / extended / iterated upon, so they would be new, but part of something old and existing
@Oblomov @Kevin Marks
Evan Prodromou
Evan Prodromou  
•
@voxpelli@KevinMarks@oblomov so, there you go! An exception that proves the rule.
@Oblomov @Pelle Wessman @Kevin Marks
Pelle Wessman
Pelle Wessman  
•
100%, we need to prioritize interoperability
Funkatron (Ed Finkler)
Funkatron (Ed Finkler)  
•
@voxpelli oh hey didn’t we work on some Ostatus stuff r somethin?
@Pelle Wessman
Funkatron (Ed Finkler)
Funkatron (Ed Finkler)  
•
@voxpelli I liked that. I enjoy data transformation 🤷‍♀️. I recall working on it for a mobile client.
@Pelle Wessman
Gabriel Bauman
Gabriel Bauman  
•
ActivityPub has a ton of momentum and adoption right now thanks to Mastodon, and I don't think a dev publishing a paper or a toy implementation threatens that. People should publish early and often.

It looks like the biggest adoption spoilers might to be federation wars over moderation policies, followed by inter-instance latency.

Basically, mods have to be made irrelevant; users must be empowered to set boundaries that work for them, are transparent to others, and are enforced broadly.
Evan Prodromou
Evan Prodromou  
•
@gabrielbauman What about a project funded by a big company, like Blue Sky?
@Gabriel Bauman
Gabriel Bauman
Gabriel Bauman  
•
Say they release tomorrow. Mastodon still trending, people still joining because now that's where their friends are if they're not on Twitter. The head start is real, and switching fatigue works in our favor.

Devs/enthusiasts will read and experiment with the new thing. Unlikely that the bulk of the Twitter diaspora will care. ActivityPub can ignore/embrace/extend concepts or be replaced itself if X is really so great.
Gabriel Bauman
Gabriel Bauman  
•
if the worst happens and we get significant fragmentation, maybe Jabber-style gateways can work. Hopefully that doesn't happen.
Evan Prodromou
Evan Prodromou  
•
@gabrielbauman one way it doesn't happen is for people to sit on their hands and not publish their snowflake protocols.
@Gabriel Bauman
Gabriel Bauman
Gabriel Bauman  
•
Hahaha true enough! I still don't think there's a need to feel threatened by the mere possibility of competition.

ActivityPub has been building and working for years, is a recognized standard, and has unbelievable momentum. It's not going anywhere as long as it keeps evolving to meet the challenges the network faces. The ocap work in particular is quite promising.

Other things may yet show some merit. They'll have a hard road and AP can hopefully assimilate any good ideas worth grabbing.
Evan Prodromou
Evan Prodromou  
•
For those not familiar with Metcalfe's Law:

https://en.wikipedia.org/wiki/Metcalfe%27s_law

empirical rule that that the value of a telecommunications network is proportional to the square of its number of users

Contributors to Wikimedia projects (Wikimedia Foundation, Inc.)
Nicolas Pettiaux
Nicolas Pettiaux  
•
i contest this law. Value is a human perception. It has no physical ground. It is by definition subjective. One does not necessarily want to be in the network of anyone else
Nicolas Pettiaux
Nicolas Pettiaux  
•
perhaps, but why would be this law correct ? It is a perceived value in an attention based economy that, for me, is completely artificial
Kevin Marks
Kevin Marks  
•
Metcalfe's law applies more to the value to spammers than to individuals. A small world network is much nicer for humans than a universally connected one.
Mark Atwood
Mark Atwood  
•
Or at least, implement AS as well as your own protocol
Alexandre Franke
Alexandre Franke  
•
Is @matrix in scope for your question too? It beats the fediverse by a large margin when it comes to Metcalfe’s law.
@The Matrix.org Foundation
    n8   Doesn't follow you 🇺🇦
    n8   Doesn't follow you 🇺🇦  
•
Is that a statement about protocols, though? It seems more like it speaks to instances and the patterns of instance-admins refusing to federate with other instances.

Clearly there are some sharply different philosophies on who/when/how federating is done; it may remain to be seen which among those are destructive to the overall value of the network.
Vincent St. Pierre :mw:
Vincent St. Pierre :mw:  
•
I'm learning a lot about ActivityPub by the replies. Thank you, Evan, for promoting the discussion.
aeva
aeva  
•
If it is of any reassurance I have absolutely no plans right now to develope any federated protocols.
Evan Prodromou
Evan Prodromou  
•
@aeva well, at least there's two of us
@aeva
ocdtrekkie
ocdtrekkie  
•
Defederation is already doing this. As it is, a bunch of small instances are trying to get people to leave my instance because they plan to block it, and the moderator behavior is the top reason I'm seeing people defend remaining on Twitter.

If a couple dozen decent size instances defed dotsocial, we'll basically have split the fediverse down the middle.
@reiver ⊼ (Charles) :batman:
@reiver ⊼ (Charles) :batman:  
•
Then what is the process for getting the people who want to create or support alternative federation protocols — to do that within the framework of ActivityPub?

Assuming you want to try to bring that "26% Strong yes" and "38% Qualified yes" to support your preference.
Evan Prodromou
Evan Prodromou  
•
@reiver that's an interesting assumption! I need to think about that.
@@reiver ⊼ (Charles) :batman:
sjvn
sjvn  
•
Agreed. ActivityPub is already a W3C standard. The last thing we need is something new that's not an open standard.
⇧