Conversation

Notices

1. Embed this notice
   silverpill (silverpill@mitra.social)'s status on Thursday, 01-May-2025 01:43:31 JST silverpill

   ActivityPub object identifiers are not IRIs. They are URIs:

   https://www.w3.org/TR/activitypub/#obj-id

   Unicode characters are not allowed there.

   #ActivityPub

   • Fish of Rage likes this.
   • Embed this notice
     silverpill (silverpill@mitra.social)'s status on Thursday, 01-May-2025 02:01:38 JST silverpill

     in reply to

     • opal

     @wowaname I saw it in Fedify docs a couple of days ago and today in Forgejo code.

   • Embed this notice
     opal (wowaname@freesoftwareextremist.com)'s status on Thursday, 01-May-2025 02:01:39 JST opal

     in reply to
     @silverpill for context who (or what implementation[s]) was suggesting they're IRIs?
   • Embed this notice
     Shawn MacIntyre (smacintyre@mastodon.social)'s status on Thursday, 01-May-2025 02:11:58 JST Shawn MacIntyre

     in reply to

     @silverpill That's a disappointing step backwards. English and the Latin Alphabet need to be de-centred in tech. Modern standards should be open to people who speak all languages.

   • Embed this notice
     silverpill (silverpill@mitra.social)'s status on Thursday, 01-May-2025 02:11:58 JST silverpill

     in reply to

     • Shawn MacIntyre

     @smacintyre I think URIs are preferable because they are simpler and easier to work with. End users don't normally see those identifiers

   • Embed this notice
     silverpill (silverpill@mitra.social)'s status on Thursday, 01-May-2025 02:19:20 JST silverpill

     in reply to

     • opal

     @wowaname Mitra doesn't allow IRIs.

     In practice, I've only seen an IRI once, and it was from WordPress. Maybe those other developers don't actually use IRIs, just the wrong term.

   • Embed this notice
     opal (wowaname@freesoftwareextremist.com)'s status on Thursday, 01-May-2025 02:19:21 JST opal

     in reply to

     • opal
     @silverpill that said, idk how many implementations do strict uri parsing or even use libraries that implicitly convert and could break an iri
   • Embed this notice
     opal (wowaname@freesoftwareextremist.com)'s status on Thursday, 01-May-2025 02:19:22 JST opal

     in reply to
     @silverpill good grief all we need is perpetual interop issues going forward
   • Embed this notice
     silverpill (silverpill@mitra.social)'s status on Thursday, 01-May-2025 05:28:32 JST silverpill

     in reply to

     • marius
     • opal
     • @reiver ⊼ (Charles) :batman:

     @mariusor @wowaname

     1. It is not clear what "shares the same URI / IRI conventions" means. This is the only place in the ActivityPub spec where a connection to IRIs is made, and all other statements refer to URIs. The section 3.1. (Object identifiers) says that identifiers must be URIs.
     2. Section 2.2 (IRIs and URLs) of ActivityStreams says "This specification uses IRIs", but it doesn't say how it uses them. That statement may refer to names of properties and types, for example, which can be IRIs. What follows after that is just nonsense: "URI may be used wherever an IRI is named". ActivityPub's "shares same conventions" might mean that relative URIs are not allowed, but that's anybody's guess.
     4. The vast majority of implementations use URIs for IDs and encode non-ASCII characters.

     cc @reiver

   • Embed this notice
     marius (mariusor@metalhead.club)'s status on Thursday, 01-May-2025 05:28:34 JST marius

     in reply to

     • opal

     @silverpill @wowaname what exactly is the problem with calling them IRIs?

     > ActivityPub shares the same URI / IRI conventions as in ActivityStreams.

     From: https://www.w3.org/TR/activitypub/#objects

     And following the link there, we get to:

     > Every URI [RFC3986] is also an IRI, so a URI may be used wherever an IRI is named. There are two special considerations: (1) when an IRI that is not also a URI is given for dereferencing, it MUST be mapped to a URI using the steps in Section 3.1 of [RFC3987] and (2) when an IRI is serving as an "id" value, it MUST NOT be so mapped.

     https://www.w3.org/TR/activitystreams-core/#urls

   • Embed this notice
     silverpill (silverpill@mitra.social)'s status on Thursday, 01-May-2025 20:48:16 JST silverpill

     in reply to

     • marius
     • opal

     @wowaname @mariusor

     This probably means they don't care what you use, URIs or IRIs.
     ActivityPub, on the other hand, has very clear requirements. I don't know how one could possibly interpret a vague reference to that AS section as a proof that ActivityPub identifiers must be IRIs.

   • Embed this notice
     opal (wowaname@freesoftwareextremist.com)'s status on Thursday, 01-May-2025 20:48:17 JST opal

     in reply to

     • marius
     • opal
     @mariusor @silverpill
     >a URI may be used wherever an IRI is named.
     so does this mean "IRIs are only allowed whenever activitypub spec mentions that an IRI is acceptable" or does this mean "activitypub also accepts IRIs where URIs are mentioned"? because otherwise i don't know why you would even quote the spec in the first place. IRIs are less-restrictive than URIs, which is what i initially meant to say
   • Embed this notice
     opal (wowaname@freesoftwareextremist.com)'s status on Thursday, 01-May-2025 20:48:18 JST opal

     in reply to

     • marius
     @mariusor @silverpill
     >I'm not sure why you're convinced that not all URIs are IRIs.
     i was reading it like "every time we say URI we also mean it can be an IRI". my apologies for being absolutely careless in my wording, because you're right, every URI is an IRI
   • Embed this notice
     marius (mariusor@metalhead.club)'s status on Thursday, 01-May-2025 20:48:20 JST marius

     in reply to

     • opal

     @wowaname I find that perfectly understandable: because all printable ASCII characters form a subset of Unicode characters, we can infer that the set of all possible URIs is included in the set of all possible IRIs.

     I'm not sure why you're convinced that not all URIs are IRIs.

     @silverpill

   • Embed this notice
     opal (wowaname@freesoftwareextremist.com)'s status on Thursday, 01-May-2025 20:48:21 JST opal

     in reply to

     • marius
     • opal
     @mariusor @silverpill
     >Every URI [RFC3986] is also an IRI
     like this type of language should NEVER have made it into a formal specification. *not* every URI is an IRI, just like not every rectangle is a square; you can't just suddenly bend reality because you're too lazy to fuckin say "IRI" everywhere you mean an IRI.
   • Embed this notice
     opal (wowaname@freesoftwareextremist.com)'s status on Thursday, 01-May-2025 20:48:22 JST opal

     in reply to

     • marius
     • opal
     @mariusor @silverpill spec writers gotta realise their goal is to write human-digestible information foremost, so whenever they footgun themselves like this with inconsistent wordy language instead of getting to the point, idfk why everyone acts surprised that every implementation goes ahead and does its own thing
   • Embed this notice
     opal (wowaname@freesoftwareextremist.com)'s status on Thursday, 01-May-2025 20:48:23 JST opal

     in reply to

     • marius
     @mariusor @silverpill so the spec is poorly-worded to boot? why are they calling them URIs whenever they mean the broader IRI category for shit like object ids? no wonder implementers get confused.