comments

proto/catalog.proto

@@ -52,6 +52,11 @@ message StreamSourceInfo {
   // deprecated
   plan_common.RowFormatType row_format = 1;
   string row_schema_location = 2;
+  // This means *use **confluent** schema registry* and is `false` for **aws glue** schema registry.
+  // Eventually we will deprecate it and rely on `enum SchemaLocation` derived from `format_encode_options` below.
+  // * schema.location     false
+  // * schema.registry     true
+  // * aws.glue.schema_arn false
   bool use_schema_registry = 3;
   string proto_message_name = 4;
   int32 csv_delimiter = 5;

src/connector/src/parser/avro/parser.rs

@@ -115,6 +115,11 @@ impl AvroAccessBuilder {
                 }
             }
             WriterSchemaCache::Glue(resolver) => {
+                // <https://github.com/awslabs/aws-glue-schema-registry/blob/v1.1.20/common/src/main/java/com/amazonaws/services/schemaregistry/utils/AWSSchemaRegistryConstants.java#L59-L61>
+                // byte 0:      header version = 3
+                // byte 1:      compression: 0 = no compression; 5 = zlib (unsupported)
+                // byte 2..=17: 16-byte UUID as schema version id
+                // byte 18..:   raw avro payload
                 if payload.len() < 18 {
                     bail!("payload shorter than 18-byte glue header");
                 }