I did not say that. SQL embedded in C needs to communicate with the database, usually using ODBC, which Netezza supports. Your particular implementation of C probably has a library to handle the communications part and there is probably some configuration information somewhere so that the program knows what database to connect to.
As to your other question, Netezza does not support cursors, but stored procedures do have a LOOP construct that allows you to iterate through a result set in much the same manner as a cursor.
However, be warned! Single row operations (UPDATE, INSERT, DELETE) in Netezza are VERY expensive. Unless you are dealing with very few rows, such a process would perform very poorly (figure 5-6 operations per second for UPDATE or DELETE on a good day). You should consider changing the process to perform updates using set operations to process large numbers of transactions in a single SQL statement. Performance would increase by many orders of magnitude. If all you are doing is SELECT, just send the query and deal with the result set in the C program.
Also, it is possible (but not recommended) to run your own code on the Netezza host processor (it's just a LINUX machine after all). But it offers no real advantage as it only cuts down communication overhead, and creates a whole bunch of disadvantages/issues (support, system stability, etc...).
And you still need to go through ODBC to interact with the database, so the SQL still runs much the same as it would coming from an external machine.
