client_uri should be a property on SIOP Request, not .registration

[jmandel]

client_uri should be not in registration but an optional parameter directly in the request?

client_uri should move up to be a sibling

[madaster97]

Having trouble accessing the Openid Spec for Dynamic client registration, but the OAuth RFC I think it's based on has "client_uri" in the metadata.

[p2-apple]

Having trouble accessing the Openid Spec for Dynamic client registration, but the OAuth RFC I think it's based on has "client_uri" in the metadata.

Yes, just like e.g. scope and responseType, so I think this move makes sense.

[jmandel]

Uh, now that I got around to implementing this, I'm not so sure it's correct :-)

I'll follow up.

[madaster97]

@p2-apple scope and responseType are not part of the Metadata definition. That's why they aren't in the registration object.

Those two fields are specific to a given request, and not generically to a given client. Note, response_types is part of the metadata definition, and on any given request the responseType chosen should be present in that array.

[jmandel]

Agreed @madaster97. I'm double checking with the person who initially raised this issue to make sure I'm not missing anything, but my reading matches yours.

[p2-apple]

I was comparing it with the OAuth2 metadata keys, but right, client_uri has a different meaning there. From an OID perspective it actually seems to make sense to keep it in registration, thanks for the close reading @madaster97 !

[jmandel]

Thanks -- I've reverted the fix and will close this out.