One of your Hibernate entities maps to a table called order and has a column called user, both reserved words in the database. What options does Hibernate give you to make the generated SQL valid, and what is the risk of switching on the setting hibernate.globally_quoted_identifiers?

## Why reserved words break SQL engines reserve words such as order, user, group, table and select. If Hibernate emits `select o.id from order o`, the parser sees the ORDER keyword and rejects the statement. The fix is always the same in principle - wrap the identifier in the database's quoting characters (double quotes in standard SQL and PostgreSQL, backticks in MySQL, square brackets in SQL Server) - but Hibernate must know which identifiers need it, and must translate to the right character per dialect. ## Per-identifier quoting The portable way is to mark the name in the mapping. Hibernate accepts backticks around a name in any name attribute: @Table(name = "`order`"), @Column(name = "`user`"), also inside @JoinColumn and @JoinTable. JPA's own convention is a name that starts and ends with an escaped double quote. Either way Hibernate stores the Identifier with quoted = true and the dialect renders it with the right characters. This is the surgical fix and it keeps the rest of the schema unquoted and case-folded as usual. A subtlety: once an identifier is quoted, it is exact. On PostgreSQL an unquoted identifier folds to lower case, so order quoted as "order" is fine, but "Order" would be a different column from order. Keep quoted names lower case unless the existing schema says otherwise. ## Automatic keyword quoting Hibernate offers hibernate.auto_quote_keyword, which asks it to quote names it recognises as reserved for the configured dialect while building the mapping model. It saves you from auditing every entity, and it is particularly useful for schema tooling. The caveat is coverage: the keyword set comes from the dialect and JDBC metadata, and databases add reserved words between versions, so a name can slip through. Treat it as a safety net, not a policy. ## Global quoting and its cost hibernate.globally_quoted_identifiers = true makes Hibernate quote every table, column, sequence, schema and catalog name it emits. It certainly solves reserved words. The price: - Case sensitivity everywhere. Every identifier now means exactly what it is spelled as. If your mapping produces createdAt and the table has created_at or createdat, statements fail where they previously worked through the engine's case folding. Retrofitting this onto a live schema is a breaking change. - Coupling to the exact spelling. Views, stored procedures, reporting queries and migration scripts written by other teams must quote identically. - Ergonomics. Hand-written SQL against the schema becomes tedious, and mistakes are silent until runtime. - A companion setting, hibernate.globally_quoted_identifiers_skip_column_definitions, exists because quoting inside generated column definitions broke some databases - a hint that the global switch has rough edges. ## How this interacts with naming strategies Quoting travels with the Identifier through the physical naming pass. A custom PhysicalNamingStrategy that rebuilds identifiers with Identifier.toIdentifier(text) and forgets the quoted flag will silently strip quoting and reintroduce the reserved-word failure; always pass name.isQuoted() through. Conversely, a strategy can be the place to implement the policy - for example, quoting any name that appears in a curated keyword list. ## What I would do Prefer renaming: a table called orders instead of order removes the problem for everyone, including hand-written SQL. Where the name is fixed by a legacy schema, quote that identifier explicitly. Reserve the global switch for schemas that Hibernate alone generates and owns.