We've gotten feedback that db.instance.id and db.name are confusing for a couple of reasons:

• those are existing terms that mean different things in different database systems (e.g. db.name should be broken down into individual layers #714)
• some databases have both concepts, and some have only one of them

To address this, one proposal we discussed in the database client semconv group is to combine these two into a single (more generically named) attribute, e.g. db.scope, which wouldn't conflict with existing database terminology.

DB-specific conventions could still define attributes which correspond to their preferred terminology, e.g. db.oracle.instance and db.oracle.schema.

We would need to define for each database what to populate into this new attribute, e.g. for Oracle it could be {db.oracle.instance}.{db.oracle.schema}.

Related to

• Should db.instance.id replace elastic and mssql specific attributes? #725
• db.mssql.instance_name vs db.instance.id and server.port #727
• Is db.elasticsearch.node.name the same as db.instance.id #728