Ok, you now have the 3rd option - neither NULL nor SQL.NULL but just an empty field. This will come into a VarChar field as '', or an empty string {which is very different from null!}. What Minhajuddin wrote now applies, but only sort-of. Basically, unless you are loading to a VarChar2 Oracle these values will not be null so the nullability attributes on the columns really don't apply.
In the case of data types that do not allow '' as a value - i.e. date, time, timestamp, and any numeric type you would use the default value attribute to change this value to a legal value or you could make the column nullable and use the attributes to set a null value.
In the case of strings you could use what Minhajuddin suggested to do the same.
So your columns would come out of the sequential file stage as nullable and any illegal values in the file would be converted to nulls. Then you can use whatever method you like to handle those nulls in the subsequent stages.
But if I were to use Minhajuddin's suggestions, then I will have to set the Nullable property to 'Yes', right? This again is not my requirement (sorry!)
I want records having null values to be REJECTED by Datastage... these records should NOT come into the main process at all, it should be rejected while the input is read in the first place... but this doesn't seem to be happening...