Currently data is lost when reverse engineering a data model and the generated sql requires extensive modification due to the lack of support for character sets and collation.
For example
CREATE DOMAIN domain_name
AS datatype [CHARACTER SET charset]
DEFAULT {literal|NULL| USER|CURRENT_CONNECTION|CURRENT_ROLE|CURRENT_TIME|CURRENT_TIMESTAMP|CURRENT_TRANSACTION|CURRENT_USER|GDSCODE|'NOW'|ROW_COUNT|SQLCODE|SQLSTATE}
[NULL|NOT NULL]
[CHECK (dom_search_condition)]
COLLATE collation;
If I wanted to create a binary UUID in a UTF8 database I would need to be able to define the character set at the very least.
CREATE DOMAIN UUID
AS CHAR(16) CHARACTER SET OCTETS
NOT NULL
COLLATE OCTETS;
Although the examples I provided are from firebird, most databases support defining both character sets and sort collations.