It's possbile that someone would want to archive a subset of columns of a local table in S3, so this would be a valid statement:

create table z02 (name varchar(20)) engine=s3 select user from mysql.user;

But currently CREATE TABLE is not supported in S3:

ERROR 1005 (HY000): Can't create table `s3_schema`.`z02` (errno: 131 "Command not supported by the engine")

It would allow to do a vertical partitioning excluding non important data or just split tables on separate S3 tables without the need to create them locally first.