Mastodon/Activity Pub question:
-
Mastodon/Activity Pub question:
As far as I can tell mastodon posts are expected to have a rich text / html-subset representation i.e. https://docs.joinmastodon.org/spec/activitypub/#sanitization)
I'm writing a tiny activity pub service, mastodon is currently stripping out all html tags when displaying posts from my dev account (see e.g. https://mastodon.social/@sarah@tap.resistant.tech/114643976114333084), plain text posts work fine (see e.g. https://mastodon.social/@sarah@tap.resistant.tech/114643533179472725)
I'm attempting to work out why that is the case, but this is proving to be a hard thing to search for.
-
Mastodon/Activity Pub question:
As far as I can tell mastodon posts are expected to have a rich text / html-subset representation i.e. https://docs.joinmastodon.org/spec/activitypub/#sanitization)
I'm writing a tiny activity pub service, mastodon is currently stripping out all html tags when displaying posts from my dev account (see e.g. https://mastodon.social/@sarah@tap.resistant.tech/114643976114333084), plain text posts work fine (see e.g. https://mastodon.social/@sarah@tap.resistant.tech/114643533179472725)
I'm attempting to work out why that is the case, but this is proving to be a hard thing to search for.
My current working theories in no particular order:
- I've missed something about how create activities/notes are constructed which conveys some kind of explicit/implicit content type?
- does mastodon do some form of tiered server trusting when it comes to deciding what to format?
- a text encoding issue somewhere?
-
Mastodon/Activity Pub question:
As far as I can tell mastodon posts are expected to have a rich text / html-subset representation i.e. https://docs.joinmastodon.org/spec/activitypub/#sanitization)
I'm writing a tiny activity pub service, mastodon is currently stripping out all html tags when displaying posts from my dev account (see e.g. https://mastodon.social/@sarah@tap.resistant.tech/114643976114333084), plain text posts work fine (see e.g. https://mastodon.social/@sarah@tap.resistant.tech/114643533179472725)
I'm attempting to work out why that is the case, but this is proving to be a hard thing to search for.
@sarahjamielewis The JSON of 1749328245 looks normal/correct at a glance, but I can't pull it into my own Mastodon server because your WebFinger at https://tap.resistant.tech/.well-known/webfinger?resource=sarah%40tap.resistant.tech doesn't currently seem to resolve, so my Mastodon server fails to put together the actor's profile. Is that known and/or intentional? Maybe it was working previously and m.s cached it?
-
@sarahjamielewis The JSON of 1749328245 looks normal/correct at a glance, but I can't pull it into my own Mastodon server because your WebFinger at https://tap.resistant.tech/.well-known/webfinger?resource=sarah%40tap.resistant.tech doesn't currently seem to resolve, so my Mastodon server fails to put together the actor's profile. Is that known and/or intentional? Maybe it was working previously and m.s cached it?
@julian ah looks like the webfinger lookup was a little too restrictive (previously it only resolved acct:sarah@tap.resistant.tech), now it resolves other versions including the one you linked
-
@julian ah looks like the webfinger lookup was a little too restrictive (previously it only resolved acct:sarah@tap.resistant.tech), now it resolves other versions including the one you linked
@sarahjamielewis Ah, no, that's my mistake. Going by the standard, the "acct:" schema isn't optional, I just have a bad habit of forgetting about it.
But there's still something going on with fetching your AP URIs from Mastodon. I can pull up your profile if I search for the double-@ handle, but not if I search for the URI, even though you appear to serve correct AP. Might be the same reason why I can't fetch your test post. I tried from my Mastodon server and from https://activitypub.academy.
-
@sarahjamielewis Ah, no, that's my mistake. Going by the standard, the "acct:" schema isn't optional, I just have a bad habit of forgetting about it.
But there's still something going on with fetching your AP URIs from Mastodon. I can pull up your profile if I search for the double-@ handle, but not if I search for the URI, even though you appear to serve correct AP. Might be the same reason why I can't fetch your test post. I tried from my Mastodon server and from https://activitypub.academy.
@sarahjamielewis Maaaaaybe it has something to do with the content type? It looks like you serve the HTML if explicitly asked for text/html, and the JSON otherwise. I usually do it the other way around.
I think Mastodon refuses to load plain JSON documents, for security reasons (long story), this might be linked to that. Only sending the AP JSON for the content types listed in section 3.2 of the spec could be worth trying.
I'm poking at this kinda blindly though. Tough situation to debug.
-
@sarahjamielewis Maaaaaybe it has something to do with the content type? It looks like you serve the HTML if explicitly asked for text/html, and the JSON otherwise. I usually do it the other way around.
I think Mastodon refuses to load plain JSON documents, for security reasons (long story), this might be linked to that. Only sending the AP JSON for the content types listed in section 3.2 of the spec could be worth trying.
I'm poking at this kinda blindly though. Tough situation to debug.
@sarahjamielewis For what it's worth, @fedify fetches and parses your actor correctly (tested with https://github.com/fedify-dev/fedify/tree/main/examples/actor-lookup-cli), so it's gotta be something where Mastodon is more strict.
-
My current working theories in no particular order:
- I've missed something about how create activities/notes are constructed which conveys some kind of explicit/implicit content type?
- does mastodon do some form of tiered server trusting when it comes to deciding what to format?
- a text encoding issue somewhere?
What we have learned so far:
- taking the json my service produces and mirroring on another AP service results in mastodon rendering the HTML on the mirrored post - so it unlikely to be an issue with the Note object at least...
- just in case, I've tweaked a couple of things in webfinger, nodeinfo, and http header signing to better align the behaviour that mastodon is has, to no effect.
-
What we have learned so far:
- taking the json my service produces and mirroring on another AP service results in mastodon rendering the HTML on the mirrored post - so it unlikely to be an issue with the Note object at least...
- just in case, I've tweaked a couple of things in webfinger, nodeinfo, and http header signing to better align the behaviour that mastodon is has, to no effect.
I have officially ran out of ideas...probably time to take a break....
Request: If anyone who knows anything about mastodon internals has any idea why plain text messages would show fine, but basic html formatted messages would show empty (see above) - I would greatly appreciate any insight.
Edit: Thanks all, This has been Solved! (https://mastodon.social/@sarah@tap.resistant.tech/114644828209750037)
-
I have officially ran out of ideas...probably time to take a break....
Request: If anyone who knows anything about mastodon internals has any idea why plain text messages would show fine, but basic html formatted messages would show empty (see above) - I would greatly appreciate any insight.
Edit: Thanks all, This has been Solved! (https://mastodon.social/@sarah@tap.resistant.tech/114644828209750037)
@sarahjamielewis can you share a sample json you're sending?
-
I have officially ran out of ideas...probably time to take a break....
Request: If anyone who knows anything about mastodon internals has any idea why plain text messages would show fine, but basic html formatted messages would show empty (see above) - I would greatly appreciate any insight.
Edit: Thanks all, This has been Solved! (https://mastodon.social/@sarah@tap.resistant.tech/114644828209750037)
@sarahjamielewis Hey @evan any ideas?
-
@sarahjamielewis can you share a sample json you're sending?
The note: https://tap.resistant.tech/sarah/posts/1749337528.json (doesn't include the wrapping Create activity, but sending plain text messages work, so I assume the Create activity is OK...I can pull one from the debug logs)
The empty render: https://mastodon.social/@sarah@tap.resistant.tech/114644584327450993
-
@sarahjamielewis Hey @evan any ideas?
@soatok @sarahjamielewis they've got a list of tags they allow. Could you be using something at the top level that they sanitize out?
https://docs.joinmastodon.org/spec/activitypub/#sanitization
-
@soatok @sarahjamielewis they've got a list of tags they allow. Could you be using something at the top level that they sanitize out?
https://docs.joinmastodon.org/spec/activitypub/#sanitization
@soatok @sarahjamielewis also, very excited to see your service!
-
The note: https://tap.resistant.tech/sarah/posts/1749337528.json (doesn't include the wrapping Create activity, but sending plain text messages work, so I assume the Create activity is OK...I can pull one from the debug logs)
The empty render: https://mastodon.social/@sarah@tap.resistant.tech/114644584327450993
@sarahjamielewis Nothing is obviously wrong with it. This is just a guess, but it might be the capitalization on the unicode escapes? Mastodon capitalizes them. Here's a mastodon note that I keep around for regression testing
https://github.com/Letterbook/Letterbook/blob/main/Tests/Letterbook.IntegrationTests/Data/mastodon_post.json -
@sarahjamielewis Nothing is obviously wrong with it. This is just a guess, but it might be the capitalization on the unicode escapes? Mastodon capitalizes them. Here's a mastodon note that I keep around for regression testing
https://github.com/Letterbook/Letterbook/blob/main/Tests/Letterbook.IntegrationTests/Data/mastodon_post.json@sarahjamielewis I put this together to help with interop testing. It's really useful to have access to mastodon's logs while you're trying to do this stuff
https://github.com/Letterbook/Sandcastles -
@sarahjamielewis Nothing is obviously wrong with it. This is just a guess, but it might be the capitalization on the unicode escapes? Mastodon capitalizes them. Here's a mastodon note that I keep around for regression testing
https://github.com/Letterbook/Letterbook/blob/main/Tests/Letterbook.IntegrationTests/Data/mastodon_post.jsonjenniferplusplus@hachyderm.io I think it might be the Unicode escapes too. I don't think they're necessary.
