CMaB beta

Sun, 26 Jul 2026 17:24:14 +0000 from YunoHost Streams

@Mike Macgirvin

Hello Mike,
An update from the test pair, and a correction of something I told you earlier.

First: your id fix works. My test instance runs the dev-branch build of 18 July (commit b001d8f49), and the Accept and the reciprocal Follow carry distinct ids; the Accept's suffix is derived from the original Follow URI. That part is settled.

The Accept is still dropped, and I can now say where. To answer your question about the Pixelfed side first: there are no failed, silenced or pending jobs — there never are. Two POSTs arrive, both 200. The reciprocal Follow runs the full chain and creates the follower relation. The Accept runs InboxValidator and ActivityHandler and then stops. No database effect, no log line; laravel.log is not even created.

I ruled out timing. By setting force_queue_threshold to 0 I separated the two deliveries by ten minutes, so the Accept arrived entirely alone. Same result. The failure signature is identical at 72 ms, 114 ms and ten minutes of separation.

I then read the Pixelfed code (0.12.7; file hashes verified against the released tag). Its Accept validator requires the Follow to be present as an inlined object with id, type, actor and object, and requires that inlined object's object to equal the Accept's actor. If the check fails, the inbox handler returns without logging anything. I ran that validator directly against the exact signed payload my instance had sent, preserved verbatim in the streams debug log. It fails on those four object fields and on nothing else. The did:key gateway actor and the id form both pass.

The observation that matters most came from re-reading our thread. The 16 July follow from badmin (my Pixelfed) to @yorg@macgirvin.com completed normally – same direction, same Pixelfed instance, same account as the failing cases. The job chain I reported then (a single InboxValidator, then ActivityHandler and FollowPipeline creating the follower relation) is exactly the success path of Pixelfed's Accept handler as it reads in the code. The difference is in the payload. Your Accept carried the Follow inlined – id, type, actor and object all present, and the inlined object's object equal to your actor. The Accept from my streams build carries the object as a bare string: the Follow's id, quoted verbatim, nothing more.

So the working case and the failing case differ in the form of the Accept's object, not in the direction of the handshake, and not in the actor id form. macgirvin.com produced an inlined object on 16 July; streams at b001d8f49 produces a string. Since you mentioned this part of the code is the same in forte and streams, I would be interested to know whether the two actually differ here, or whether something else accounts for it.

Two further observations on the Pixelfed side, for completeness. The Follow id it issues contains a fragment (#follow/…), and that fragment does not appear in the actor document a fetch of that id would return. And its follow_requests table stores an empty activity column. So it neither retains nor can dereference the id it issued itself.

On the specification: I have checked, and as far as I can tell both forms are permitted –ActivityStreams allows Object or Link for the property, and ActivityPub §7.5 speaks of an Accept referencing the Follow as its object.

Full payloads, timelines and log excerpts are available, and the test pair stays up. I am happy to run any test you name against it.

All the best,
Poul
Admin of streams.flamenco.info

Mike Macgirvin

Sun, 26 Jul 2026 21:30:09 +0000

Should have it sorted shortly.

Mike Macgirvin

Wed, 05 Aug 2026 21:03:57 +0000

I pushed up a fix for that particular issue 10 days back (I had been testing in forte rather than streams - and there actually are some minor differences), but subsequently got side-tracked.

Just tried this again and I seem to be able to connect bi-directionally in forte but streams is still getting stuck somewhere. As usual, pixelfed accepts the delivery and provides no indication of anything wrong, but the Accept still isn't getting through.

Meanwhile in forte, I can receive pixelfed posts but replies to it are dropped. I can also receive posts (with images) from forte, but the images are corrupted. So the issues run much deeper than just making connections. Unless somebody steps in who has the ability to do a deep dive into the code of both projects, the only way out I can see is for me to run a pixelfed server locally. (And a gts server, and a lemmy server, and a nodebb server, and a pleroma server, etc... )

Anyway, here's what the streams Accept message looks like now - and it still is getting rejected.

{
  "@context":"https://www.w3.org/ns/activitystreams",
  "id":"https://dev.macgirvin.com/.well-known/apgateway/did:key:z6MkiD7Bm4CjNC4L7ENvZauQvoowdNEyYRpQxqgsnjnzcxZg/follow/10/a9465\
1147a6591838bdfcf2260039905",
  "type":"Accept",
  "actor":"https://dev.macgirvin.com/.well-known/apgateway/did:key:z6MkiD7Bm4CjNC4L7ENvZauQvoowdNEyYRpQxqgsnjnzcxZg/actor",
  "object":{
    "@context":"https://www.w3.org/ns/activitystreams",
    "id":"https://pixelfed.com/users/macgirvin#follow/990709023791263135",
    "type":"Follow",
    "actor":"https://pixelfed.com/users/macgirvin",
    "object":"https://dev.macgirvin.com/.well-known/apgateway/did:key:z6MkiD7Bm4CjNC4L7ENvZauQvoowdNEyYRpQxqgsnjnzcxZg/actor"
  },
  "to":[
    "https://pixelfed.com/macgirvin"
  ],
  "cc":[

  ],
  "proof":{
    "type":"DataIntegrityProof",
    "cryptosuite":"eddsa-jcs-2022",
    "created":"2026-08-05T20:16:09Z",
    "verificationMethod":"did:key:z6MkiD7Bm4CjNC4L7ENvZauQvoowdNEyYRpQxqgsnjnzcxZg",
    "proofPurpose":"assertionMethod",
    "proofValue":"zuem6d3zF9QEHbc2SuFgrowj1LqwFv5wUh81NvtiYQ5HL2ERz5w2NcHM2cXgqrZsTprSNRu85vRzmjLmG2U7S1K9"
  }
}

I'll try a few more things but I'm really running out of ideas.

CMaB beta

Tue, 21 Jul 2026 14:13:01 +0000 from YunoHost Streams

Admin wrote the following post Tue, 21 Jul 2026 14:12:58 +0000

Dette er en DM fra Pixelfed til Streams

@Admin

Mike Macgirvin

Tue, 21 Jul 2026 19:31:58 +0000

Just so you're aware, a "top-level" DM to a group channel creates a post to the group. This is not a bug - it's the only way that private groups can work at all.

CMaB beta

Mon, 13 Jul 2026 16:38:45 +0000 from YunoHost Streams

@Mike Macgirvin

Apologies for the direct mention — the fediversity.site support group appears
to be gone and the Codeberg issue tracker shows "Access denied", so I wasn't
sure where else to bring this. Please redirect me if there's a better venue.

**Follow requests from Pixelfed never complete — Accept delivered (HTTP 200)
but apparently discarded on their end**

Setup: streams 26.5.3 (streams.flamenco.info, YunoHost) ↔ Pixelfed 0.12.7
(gram.social).

When a Pixelfed user follows any channel on my instance, the Follow arrives,
the connection is approved, and streams delivers the Accept — their inbox
returns 200. But Pixelfed never registers it: the UI stays on "Follow
Requested" forever, no error on either side. A control test with Mastodon
using the same channel works perfectly, so the Accept itself seems fine.
This is not group-related; a plain Person test channel behaves identically.

Timeline from the debug log (2026-07-10):

09:41:22 Inbox.php:70:post: inbox_ua: (Pixelfed/0.12.7; +https://gram.social)
[connection approved manually]
09:41:38 Queue.php:417:deliver: sending draft-cavage signature to
https://gram.social/users/ballhaus/inbox
09:41:38 Queue.php:496:deliver: queue post returned 200

The Accept's actor (and the Follow's object) use the FEP-ef61 form
(https://…/.well-known/apgateway/did:key:…/actor). My suspicion is that
Pixelfed can't match/validate that actor id and silently drops the activity.
The same symptom was reported against Pixelfed with Hubzilla in 2024
(pixelfed/pixelfed#5387) with no response, so I'm hoping for a workaround
on this side of the fence:

Is there a site or channel setting to present the classic channel URL as
the actor id towards platforms that don't handle did:key gateway ids —
or would that require a code change?

Full payloads and unabridged logs available on request. Happy to test
patches against this instance pair.

show all 10 comments

Mike Macgirvin

Mon, 13 Jul 2026 21:59:09 +0000

G'day. I'm retired and stay out of public spaces these days. As far as I know, Pixelfed is the only application that has an issue with Accept/Follow from streams.

So the first question is why do you assume this is in fact a streams bug? If I encountered this issue, I'd be looking at pixelfed and finding out what went wrong; especially since other apps work just fine -- and they lied to us and sent back a 200 response (indicating everything is hunky-dory on their end - except it isn't).

But I also know dansup won't lift a finger to improve federation with our software. I think we've still got an open issue that they don't support url arrays - that's going on 7-8 years now. So I've looked at our implementation and found something that pixelfed "might not like"; though I have no way of proving it at the moment. I've just pushed this to the dev branch. I don't know if this will change anything, but I've cleaned it up and am crossing fingers - because I really don't need or want a pixelfed server.

It's highly doubtful the nomadic url's are involved. I note that a pixelfed account was able to follow you without issue, and this would've required fetching, storing, and using the nomadic url.

CMaB beta

Tue, 14 Jul 2026 08:16:17 +0000 from YunoHost Streams

Hello, thank you for responding.

First of all, I'm not assuming this is a Streams bug at all. On the contrary, I know that the implementation fault lies on the Pixelfed side. Like Mastodon, Pixelfed is not really interested in federation with anything other than their own platform. I'm not interested in running a Pixelfed server either!

What I'm trying to create is a group that shares content across different federated platforms. The idea is to host a kind of federated hub in the form of a group, from which I can publish content to the other members of the group – independent of the members' choice of platform. Likewise, any member of the group can post to the group and have their contribution shared automatically or semi-automatically with the other members – independent of their choice of platform.

I've been around Friendica and Hubzilla and have finally decided to opt for Streams. The ultimate goal is to make the Streams server capable of federating with most platforms that use ActivityPub, and maybe even the AT Protocol via bridging. BTW, the group is about classical music (and beyond). The members should be able to share text, photos, audio and video.

Some developers seem to care very little about federation outside their own platform, and I know I'm bound to run into issues with Pixelfed, Mastodon and probably other platforms. That implies having to resolve any federation issue from my side, even if the issue is caused by a sloppy implementation of the standards on the other platform. In the end, that means having to implement workarounds as long as platforms like Pixelfed or Mastodon continue to overlook the W3C standards.

The future members of this Streams group are not to blame for a faulty implementation of the protocols in their platform of choice. Most of them will simply have signed up for an account without any background knowledge of the implementation. My goal is to make this group federation work for as many users on as many different platforms as possible. That is why I naively asked for a channel setting that would solve the problem :)

I have a reproducible setup and will set up a test instance (streams <- -> Pixelfed 0.12.7) from the dev branch to verify whether your change fixes the handshake. Since you've retired from public spaces: where should test results and findings like these be reported – is there a forum or venue for Streams that is open to users and/or developers?

Thank you for your time and for your work on Streams.

All the best,
Poul Udbye Pock-Steen
Admin of streams.flamenco.info

CMaB beta

Wed, 15 Jul 2026 18:20:14 +0000 last edited: Wed, 15 Jul 2026 18:40:26 +0000 from YunoHost Streams

I set up my own Pixelfed instance (0.12.7, fresh install, admin access including the Horizon queue) and repeated the test against it, following the same streams channel as before. Same result as on gram.social:
Accept delivered, HTTP 200, and the follow request stays pending.

This time I could see what happens on the Pixelfed side:

nginx: two POSTs to /users/badmin/inbox at 16:19:50 UTC, both 200.
Horizon: no failed, silenced or pending jobs. Completed jobs: two InboxValidator runs (16:19:50), then ActivityHandler, FeedFollowPipeline and FollowPipeline (the latter tagged App\Follower:1)
So one of the two activities went through: the reciprocal Follow. Confirmed in the UI: "Follows You" badge, follower count 1, notification "@test followed you".
The Accept produced nothing. No log entry either – laravel.log doesn't even exist on this instance.

One thing I noticed in the two payloads (from the Horizon job details):

"type":"Accept",
"id":".../apgateway/did:key:z6Mkp…/follow/31/d28d9a8a25c076a874499c2e5d60d823"

"type":"Follow",
"id":".../apgateway/did:key:z6Mkp…/follow/31/d28d9a8a25c076a874499c2e5d60d823"

The Accept and the reciprocal Follow carry the same activity id. My gram.social logs from July 10 show the same thing (both activities had .../follow/28/24068ef2… that day). Whether that's what makes Pixelfed drop the Accept I can't say, but it's the only difference I can point to between the activity that worked and the one that didn't.

The instance is at https://test.flamenco.info and I'll keep it running with a test account on it, @pup@test.flamenco.info
I can run follow requests from it against any channel you name and post what the queue and the payloads show.

CMaB beta

Thu, 16 Jul 2026 08:16:22 +0000 from YunoHost Streams

Great news! My production streams is on 26.5.3, so the quickest verification from my side would be if you follow @pup@test.flamenco.info (or @badmin@test.flamenco.info) from a channel on your dev instance. I'll post what the queue and the payloads show.
Meanwhile I'll look at getting a dev-branch streams running to stay ahead of things.

Mike Macgirvin

Thu, 16 Jul 2026 19:25:03 +0000

If that works I'll do another quick release. IIRC, the Pleroma issue only affects people that connect to the same account twice; (the first time works fine but not the second) - and they reject all of our other incoming messages and three otherwise competent devs have looked at it but can't tell me anything; though it's fixed in Akkoma (a pleroma fork) -- so long story short that will have to be a lower priority right now.

CMaB beta

Thu, 16 Jul 2026 21:48:03 +0000 from YunoHost Streams

Requested a follow to @yorg@macgirvin.com tonight; it went through.
In the queue: one InboxValidator run, then ActivityHandler and
FollowPipeline creating the follower relation, no failed jobs.
The Accept carries its own id:

.../apgateway/did:key:z6Mki7gw…/follow/144/ec61ec88f2de56442709b7179b475da3

with my original Follow (its own id) embedded as the object.

"payload": 
"{"@context":"https://www.w3.org/ns/activitystreams","id":"https://macgirvin.com/.well-known/apgateway/did:key:z6Mki7gw1FZgXTkWQkJnPEvVbtNgrjA3dMkTFSyNJu64LP6s/follow/144/ec61ec88f2de56442709b7179b475da3","type":"Accept","actor":"https://macgirvin.com/.well-known/apgateway/did:key:z6Mki7gw1FZgXTkWQkJnPEvVbtNgrjA3dMkTFSyNJu64LP6s/actor","object":{"@context":"https://www.w3.org/ns/activitystreams","id":"https://test.flamenco.info/users/badmin#follow/983448361476132867","type":"Follow","actor":"https://test.flamenco.info/users/badmin","object":"https://macgirvin.com/.well-known/apgateway/did:key:z6Mki7gw1FZgXTkWQkJnPEvVbtNgrjA3dMkTFSyNJu64LP6s/actor"},"to":["https://test.flamenco.info/badmin"],"cc":[],"proof":{"type":"DataIntegrityProof","cryptosuite":"eddsa-jcs-2022","created":"2026-07-16T21:25:18Z","verificationMethod":"did:key:z6Mki7gw1FZgXTkWQkJnPEvVbtNgrjA3dMkTFSyNJu64LP6s","proofPurpose":"assertionMethod","proofValue":"z3YMLvQ2wwkzG6rjWUQ92VPR4exBu4fyCvdLCP5kpuLMc92TH3fC9vvQ362MmW5Ck2LUqxQJ7c4xU8CN57un3j53z"}}",
"queue": 
"follow"
}

CMaB beta

Fri, 10 Jul 2026 12:23:08 +0000 from YunoHost Streams

Et testopslag fra cmab
#testopsag-cmab

CMaB beta

Thu, 09 Jul 2026 21:10:42 +0000 from YunoHost Streams

Ballhaus wrote the following post Thu, 09 Jul 2026 21:10:39 +0000

Vi prøver med en ny meddelse

@Ballhaus

CMaB beta

Thu, 09 Jul 2026 21:11:45 +0000 from YunoHost Streams

Så burde man også kunne sende et foto.

CMaB beta

Thu, 09 Jul 2026 20:56:13 +0000 from YunoHost Streams

Ballhaus wrote the following post Thu, 09 Jul 2026 20:56:12 +0000

Jeg tror meddelelsen nåede frem

@Ballhaus

CMaB beta

Thu, 09 Jul 2026 20:41:22 +0000 last edited: Thu, 09 Jul 2026 20:41:37 +0000 from YunoHost Streams

Just a test message to share

CMaB beta

Thu, 09 Jul 2026 19:08:55 +0000 last edited: Thu, 09 Jul 2026 19:12:56 +0000 from YunoHost Streams

It doesn't get more beautiful than this today!

IRMELIN ROSE
The English composer Frederick Delius (1862 -1934) held profound connections with Danish intellectual circles of his time, with their literature playing a crucial role in shaping his compositions. Early in his career, Delius ”collaborated” with Danish literary giants like H.C. Andersen and Holger Drachmann, turning their works into musical pieces.
His close friendship with poet Helge Rode further deepened his engagement with Danish literature. The bond shared between them led to the creation of Delius's symphonic poem "Lebenstanz", directly influenced by Rode's play "Dansen gaar."
However, the most profound impact on Delius's work came from poet and novelist Jens Peter Jacobsen. Despite never meeting him, Delius was deeply influenced by Jacobsen's writings. Jacobsen's novel "Niels Lyhne" provided the foundation for his opera, "Fennimore and Gerda." and six of the songs from "Seven Danish Songs" was based on Jacobsen's poems, among them the poem 'Irmelin Rose':

There lived a king in days departed
Many treasures rare he owned
But far beyond the best of these
He thought his beauteous Irmelin
Irmelin rose
Irmelin sun
Irmelin loveliest of all!

Every knightly helmet shining
Mirrored back her image bright
Every poet sang of passion
Breathed her name by day and night:
Irmelin rose
Irmelin sun
Irmelin loveliest of all!

Knights by hundreds, noble wooers
Thronged the ancient castle hall
Wooed the maid with fond devotion
Wooed with sweet and flowered word:
Irmelin rose
Irmelin sun
Irmelin loveliest of all!

But the princess would not listen
Cold her heart was, cold as steel
She lightly mocked their marks of anguish
And scorned their pangs of hopeless love
Irmelin rose
Irmelin sun
Irmelin loveliest of all!

Several composers have composed music inspired by the poem, being Carl Nielsen's magnificent interpretation perhaps the most recognized, at least in Denmark. However, I have a special affinity for Delius's version, as I generally appreciate Delius's music.
You can listen to the song in a performance by the Aarhus Symphony Orchestra.

Seven Danish Songs: Irmelin Rose

Provided to YouTube by The Orchard EnterprisesSeven Danish Songs: Irmelin Rose · Aarhus Symphony Orchestra · Henriette Bonde-Hansen · Frederick Delius · Bo H...

CMaB beta

Wed, 08 Jul 2026 12:44:29 +0000 from YunoHost Streams

Test 10.
Dette er blot plain tekst, for at se om den rammer Bluesky

CMaB beta

Wed, 08 Jul 2026 12:28:24 +0000 from YunoHost Streams

Bridgy Fed for Bluesky wrote the following post Wed, 08 Jul 2026 12:28:24 +0000

Welcome to Bridgy Fed! Your account will soon be bridged to Bluesky at cmab-beta.streams.flamenco.info.ap.brid.gy. See the docs and your user page for more information. To disable this and delete your bridged profile, block this account.

@Bridgy Fed for Bluesky

CMaB beta

Mon, 06 Jul 2026 23:55:42 +0000 from YunoHost Streams

CMaB beta wrote the following post Sun, 14 Jun 2026 20:25:35 +0000

9. test: Sharing a Youtube link within a text file
https://www.youtube.com/watch?v=B4d3WtixKBg

CMaB beta

Sun, 05 Jul 2026 19:00:10 +0000 last edited: Mon, 06 Jul 2026 21:18:28 +0000 from YunoHost Streams

Dette er er tekst med et foto

CMaB beta

Sun, 14 Jun 2026 20:25:35 +0000 last edited: Sun, 14 Jun 2026 20:29:21 +0000 from YunoHost Streams

test: Sharing a Youtube link within a text file

https://www.youtube.com/watch?v=B4d3WtixKBg

Poul Pock-Steen

Sun, 14 Jun 2026 20:27:07 +0000 from hubzilla

@CMaB beta The post also reached Hubzilla

CMaB beta

Sun, 14 Jun 2026 20:22:53 +0000 from YunoHost Streams

test:

Sharing a video link to Peertube witin a text file
https://ballhaus.media/w/eTm6VPSeweejZextHMR3FY
Dalum Kirke revisited

A low-fi video with hi-fi content filmed and recorded on an iPad. The video documents a concert at Dalum Church, Odense, Denmark titled ‘Dalum Kirke revisited’. The concert was a site-responsive co...

CMaB beta

Sun, 14 Jun 2026 20:00:14 +0000 from YunoHost Streams

Seventh test:
Fifth test:
Sharing a video link to Peertube witin a text file
https://ballhaus.media/w/eTm6VPSeweejZextHMR3FY
Dalum Kirke revisited

A low-fi video with hi-fi content filmed and recorded on an iPad. The video documents a concert at Dalum Church, Odense, Denmark titled ‘Dalum Kirke revisited’. The concert was a site-responsive co...

PARLA

Sun, 14 Jun 2026 20:06:32 +0000 from friendica

The post reaches Friendica

CMaB beta

Sun, 14 Jun 2026 17:31:28 +0000 from YunoHost Streams

Sixth test:
Sharing a Youtube link within a text file
https://www.youtube.com/watch?v=B4d3WtixKBg

CMaB beta

Sun, 14 Jun 2026 17:25:52 +0000 from YunoHost Streams

Fifth test:
Sharing a video link to Peertube witin a text file
https://ballhaus.media/w/eTm6VPSeweejZextHMR3FY
Dalum Kirke revisited

A low-fi video with hi-fi content filmed and recorded on an iPad. The video documents a concert at Dalum Church, Odense, Denmark titled ‘Dalum Kirke revisited’. The concert was a site-responsive co...

CMaB beta

Sun, 14 Jun 2026 16:49:37 +0000 last edited: Sun, 14 Jun 2026 16:58:50 +0000 from YunoHost Streams

Fourth test:
A video link from Youtube within a text file

CMaB beta

Sun, 14 Jun 2026 16:37:30 +0000 from YunoHost Streams

Poul Udbye Pock-Steen wrote the following post Sun, 14 Jun 2026 16:37:30 +0000

@CMaB beta
Dette er en video sendt fra Mastodon
https://ballhaus.media/w/eTm6VPSeweejZextHMR3FY

@Poul Udbye Pock-Steen

CMaB beta

Sun, 14 Jun 2026 16:29:39 +0000 last edited: Sun, 14 Jun 2026 16:33:19 +0000 from YunoHost Streams

Third test: A video link to Peertube wihtin a text file
Dalum Kirke revisited

A low-fi video with hi-fi content filmed and recorded on an iPad. The video documents a concert at Dalum Church, Odense, Denmark titled ‘Dalum Kirke revisited’. The concert was a site-responsive co...

CMaB beta

cmab-beta@streams.flamenco.info