I was thinking this document could be more valuable if we add some more text around the different steps sender and receiver go through.
By using "MAY" instead of "SHOULD" in the right places, I think this is possible without adding more requirements or bringing things into scope that are currently out of scope.
For instance, we can document /ocm-provider and describe how it is used in practice in some real-world interactions (even in an interoperable way across implementations!), but without making it a requirement. For instance when sharing only from and to the ScienceMesh, this endpoint is not needed.
Another example are the notificationType strings that are used interoperably across some (but not all) implementations in practice; we can at least describe the prevalent notificationType strings as things a server "MAY" support, so we don't end up with an eco-system where the same notification type is identified with more than one string.
I was thinking this document could be more valuable if we add some more text around the different steps sender and receiver go through.
By using "MAY" instead of "SHOULD" in the right places, I think this is possible without adding more requirements or bringing things into scope that are currently out of scope.
For instance, we can document
/ocm-provider
and describe how it is used in practice in some real-world interactions (even in an interoperable way across implementations!), but without making it a requirement. For instance when sharing only from and to the ScienceMesh, this endpoint is not needed.Another example are the
notificationType
strings that are used interoperably across some (but not all) implementations in practice; we can at least describe the prevalentnotificationType
strings as things a server "MAY" support, so we don't end up with an eco-system where the same notification type is identified with more than one string.