Skip to main content


Salmon fields in webfinger response


You might have seen that FediTest has begun publishing test results. As a first step they're only testing webfinger. For Friendica there are lots of failures due to Salmon fields in the webfinger response. According to the spec, link relations must be either an IANA type or a URI. Friendica returns "salmon" and "magic-public-key" link relations, which are neither.

From what I've read, the Salmon protocol is supposed to use URIs for these link relations. I'm guessing that the current approach is designed for compatibility with some other service that doesn't quite follow the spec either. Does anyone know if this is true, and what would break if URIs were used instead?

Should the technical violation of the webfinger spec be considered a bug?

!Friendica Developers

FediTest: FediTest | FediTest first draft WebFinger results (for FediDevs meetup today)

in reply to Matthew Exon

We would lose the connectivity with systems that use OStatus, which is used by old versions of GNU Social.
in reply to Michael Vogel

Right, so we're talking live, active servers running old versions of GNU social, such that it's not realistic to bring GNU social into line with the webfinger spec, and it's not reasonable to ask the servers to upgrade. Are there specific examples of such servers?

I could also ask FediTest to make exceptions for these specific cases, with that justification. Or we could just live with a bad score. I'm hoping that eventually FediTest will help tighten up the AP implementation. If there's too many known failures, it makes it less likely that FediTest will be useful.

in reply to Matthew Exon

We really have to check if there really still are servers out there who only speak OStatus. If that is not the case, I will gladly remove the OStatus part. This is some really old code and we can't test it anymore. So it makes sense to get rid of it when it doesn't hurt anyone.
in reply to Michael Vogel

That would be ideal! In that case: it's not a bug in Friendica, and nothing should change in FediTest, and there's some investigation to be done if there's time.
in reply to Michael Vogel

@Michael Vogel @Matthew Exon Professor Sekula's Chirp server!

But he has expressed that he won't be running it for much longer. Which reminds me that I want to scrape 800 pages of fedi history from there that can't be found elsewhere. Hope I get around to it while it's still up.

in reply to clacke: exhausted pixie dream boy πŸ‡ΈπŸ‡ͺπŸ‡­πŸ‡°πŸ’™πŸ’›

in reply to Michael Vogel

That all makes sense then. These are running services of an unmaintained project but with real live users who should not be just switched off arbitrarily. One option could be to downgrade OStatus to an addon. But presumably that would be quite a lot of refactoring.
in reply to Matthew Exon

It seems as if that server is running a version of GNU Social that does support ActivityPub, so the connection to that server shouldn't break.
⇧