That is probably the simplest. Should be just a tiny parser tweak to get hat ALTER accepted I think.
Postgres is similar. For SERIAL columns, they create the table as is, create and attach sequence, restore value with a routine, and then do the table inserts. They don't do IDENTITY so the BY DEFAULT vs ALWAYS isn't an issue. If you insert values manually into them, it will get a duplicate key when the sequence comes to it.
That is probably the simplest. Should be just a tiny parser tweak to get hat ALTER accepted I think.
Postgres is similar. For SERIAL columns, they create the table as is, create and attach sequence, restore value with a routine, and then do the table inserts. They don't do IDENTITY so the BY DEFAULT vs ALWAYS isn't an issue. If you insert values manually into them, it will get a duplicate key when the sequence comes to it.