new versions of schema are created when schema hasn't changed when using ccompat endpoint

[zzztimbo]

Each time I run this:

 ./bin/kafka-avro-console-producer --property schema.registry.url=http://localhost:8080/api/ccompat --topic SNACKS_AVRO --property value.schema='
{
  "type": "record",
  "name": "myrecord",
  "fields": [
      {"name": "name",  "type": "string" }
    , {"name": "calories", "type": "float" }
    , {"name": "color", "type": "string" }
  ]
}' < ~/Workspace/kafka-serialization/snacks_2.txt

I get another version in the apicurio registry. I don't think this is the expected behavior.

[alesj]

Auto-register is by default (afaik) in Confluent’s serde. It now just depends how this is handled server-side - would need to check in more details If you use Confluent’s serde perhaps turn this auto off. Or use Apicurio’s serde which has more config options.

…

On Thu, 28 May 2020 at 22:22, Tim Chan ***@***.***> wrote: Each time I run this: ./bin/kafka-avro-console-producer --property schema.registry.url=http://localhost:8080/api/ccompat --topic SNACKS_AVRO --property value.schema=' { "type": "record", "name": "myrecord", "fields": [ {"name": "name", "type": "string" } , {"name": "calories", "type": "float" } , {"name": "color", "type": "string" } ] }' < ~/Workspace/kafka-serialization/snacks_2.txt I get another version in the apicurio registry. I don't think this is the expected behavior. — You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub <#551>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AACRA6H4YU2IV7BZAHPDD2DRT3BZZANCNFSM4NNMYGPQ> .

[EricWittmann]

Yeah I do think that the Confluent registry does a content check when registering a schema. If the schema already exists in the registry (based on the actual bytes of the schema) then they won't re-register a new one - you'll get back the same GlobalId. I'm not sure whether a new version will get created - I'm guessing not. Although if you register the same schema to a different subject, you'll get a new subject+version but keep the same GlobalId. That we will not do. It's a storage optimization that isn't really necessary IMO.

That said, we can probably tweak the register method to better mimic the confluent behavior. I've marked this as a bug.