• In netCDF-4 jargon, there are atomic types, and user defined types.
• Atomic types are those which can't be further subdivided.
• All six classic model types (BYTE, CHAR, SHORT, INT, FLOAT, DOUBLE) are atomic, and fully supported in netCDF-4.
• The following new atomic types have been added in netCDF-4: UBYTE, USHORT, UINT, INT64, UINT64, STRING.
• Use these just like the classic model types, for example, use this function to write an attribute of UINT64:

  EXTERNL int
  nc_put_att_ulonglong(int ncid, int varid, const char *name, nc_type xtype,
  		     size_t len, const unsigned long long *op);
  

• Take care when using the unsigned types, as they are not supported in Java or Fortran. Who can read an unsigned long long? Only C/C++.
• The string type will efficiently store arrays of variable length strings.

2007 Unidata NetCDF Workshop for Developers and Data Providers > Using NetCDF-4 Features, Part 1