Thanks @pfefferle! All I did was implement the protocol as many others have done. Half the credit would go to you as well for the other end of the integration!
It is really cool to be able to follow a blog from a forum. Are comments from the fediverse integrated?
@pfefferle I think there's a bit of a push to provide a full fediverse client (that includes two-way interaction between users). That's what the NodeBB integration is trying to do — to be that all-in-one fediverse client.
But the reality is there's no reason you have to do that. Simply federating content out and maybe ingesting responses back as comments would be a feature complete integration as far as I'm concerned! Anything else is gravy.
@pfefferle categories in NodeBB are exposed via ActivityPub just like users, so they can be followed by remote users (just like how local users can "watch" a category, I suppose).
We follow FEP-1b12 at the moment, which means all posts are Announced by the category actor to any followers.
@julian I do that for the blog-wide user. It is from the type group and boosts all posts from every author. But it seems to be not compatible with NodeBB for now?!?
Hey @pfefferle, I looked into it, and NodeBB can successfully resolve your instance actor via webfinger, but cannot make an S2S GET call to https://notiz.blog/@notiz.blog
I receive the following error:
{
"error":"This method requires an authenticated user"
}
@julian The HTML version (HTTP Request without Accept header) does redirect to the main page, but if you add the Accept header to get the JSON it will work even without signature. (I have not yet enabled AUTHORIZED_FETCH).
I ran some tests and it seems to work up to the Follow Accept response where I receive a 500 from the NodeBB inbox.
The issue was that apparently for the blog actor WP sent actor URI and not username@host - which AFAIK is perfectly valid, NodeBB just assumed the latter was always the case 🙂
@pfefferle it was actually because of @ in actor IDs - the codepath for choosing a webfinger lookup over just accepting the URL as the ID only checked for its presence, and the code for webfinger lookups was only meant to run for the acct: protocol.
So yeah, not WP fault at all, just a combination of a bug and missing feature 🙂 Now both should be fixed.
Add comment