[ https://issues.apache.org/jira/browse/CASSANDRA-7492?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14051383#comment-14051383 ]
Jorge Bay commented on CASSANDRA-7492: -------------------------------------- If it involves deserialization of stored value vs and serialization to the output value in a different format, I agree it looks like an unnecessary overhead... For the C# driver it makes sense to only support Udts under protocol 3, so you can "wont fix" this. > Udt inner collections under protocol 2 > -------------------------------------- > > Key: CASSANDRA-7492 > URL: https://issues.apache.org/jira/browse/CASSANDRA-7492 > Project: Cassandra > Issue Type: Bug > Components: Core > Environment: Cassandra 2.1-rc2 > Reporter: Jorge Bay > Priority: Minor > > Consider the following schema: > {code:sql} > CREATE TYPE phone (alias text, number text, country_code int); > CREATE TYPE contact (first_name text, last_name text, birth_date timestamp, > phones set<phone>, emails set<text>); > CREATE TABLE users_contacts (id int PRIMARY KEY, contact_value contact); > {code} > Under protocol v2, Cassandra serializes the phone *set* within contact udt, > it uses protocol v3 collections format (4 byte lengths). > As an UDT is a composition of other types, it makes sense to maintain the > other types formatting (in this case 2 byte length). > Possibly related to CASSANDRA-7472 -- This message was sent by Atlassian JIRA (v6.2#6252)