mysql collation_connection gets wrong value

The create_database method of databases.rake creates a mysql database with default charset utf8 and default collation of utf8_unicode_ci. These values can be overridden by providing a :charset resp. :collation key/values in database.yml.

When connecting to the database though by default it will use these values:

 {"Variable_name"=>"character_set_client", "Value"=>"latin1"},
 {"Variable_name"=>"character_set_connection", "Value"=>"latin1"},
 {"Variable_name"=>"character_set_database", "Value"=>"utf8"},
 {"Variable_name"=>"character_set_results", "Value"=>"latin1"},
 {"Variable_name"=>"character_set_server", "Value"=>"utf8"},
 {"Variable_name"=>"character_set_system", "Value"=>"utf8"},
 {"Variable_name"=>"collation_connection", "Value"=>"latin1_swedish_ci"}
 {"Variable_name"=>"collation_database", "Value"=>"utf8_unicode_ci"},
 {"Variable_name"=>"collation_server", "Value"=>"utf8_unicode_ci"}

The only way to get some control over this by explicitly defining encoding:utf8 in database.yml. You'd then get:

 {"Variable_name"=>"character_set_client", "Value"=>"utf8"},
 {"Variable_name"=>"character_set_connection", "Value"=>"utf8"},
 {"Variable_name"=>"character_set_database", "Value"=>"utf8"},
 {"Variable_name"=>"character_set_results", "Value"=>"utf8"},
 {"Variable_name"=>"character_set_server", "Value"=>"utf8"},
 {"Variable_name"=>"character_set_system", "Value"=>"utf8"},
 {"Variable_name"=>"collation_connection", "Value"=>"utf8_general_ci"}
 {"Variable_name"=>"collation_database", "Value"=>"utf8_unicode_ci"}, 
 {"Variable_name"=>"collation_server", "Value"=>"utf8_unicode_ci"}

I.e., the collation_connection value is still using a wrong value in this case (it's value is always the default collation for the character set of the connection, in this case of the utf8 character set).

The only solution I see is by patching the +configure_connection+ method of the mysql adapter.

There are two options:

1. Instead of doing SET NAMES 'utf8' it should be possible to do SET NAMES 'utf8' COLLATE 'utf8_unicode_ci'. For this I propose the +configure_connection+ method should (re)use the :charset and :collation values from database.yml (in favor of :encoding), just like the databases rake tasks do.

2. Instead of doing SET NAMES 'utf8' it should do SET CHARACTER SET 'utf8'. In that case the collation_connection value will be set to the value of collation_database.

Which option is preferable?