DB2 SQL join with USING need to support nullable column

I have a problem with a JOIN when columns are nullable.

summary:

I ask for an correction of the USING feature : in the sql engine do a IS NOT DISTINCT FROM when the column is nullable, so USING will do its job.

details:

for a better explanation, you can reproduce the problem with this sequence:

create table qtemp.autinf as ( select * from QSYS2.AUTHORIZATION_LIST_INFO ) with data
;

in this table, there are always either oname is null or path_name is null


when I try to join the table on itself with a USING


select
AUTH_LIST, ONAME, OSCHEMA, PATH_NAME, SYS_OTYPE
from qtemp.autinf B join qtemp.autinf A using(
AUTH_LIST, ONAME, OSCHEMA, PATH_NAME, SYS_OTYPE )
;

the resulset is empty.
translation in user language: the "keys" in file B are not in file A. inconsistant.


The reason is : NULL values never "equal" each other
the right way is to use the IS DISTINCT FROM clause


select
b.AUTH_LIST, b.ONAME, b.OSCHEMA, b.PATH_NAME, b.SYS_OTYPE
from qtemp.autinf B join qtemp.autinf A
on ( b.AUTH_LIST is not distinct from a.AUTH_LIST
and b.oname is not distinct from a.oname
and b.oschema is not distinct from a.oschema
and b.path_name is not distinct from a.path_name
and B.SYS_OTYPE = a.SYS_OTYPE)
;

a few ugly and difficult to explain to user. For my users, the concept of null is just foggy.



I would like to be able to use
USING (AUTH_LIST, ONAME, OSCHEMA, PATH_NAME, SYS_OTYPE )

I ask for an correction of the USING feature : in the sql engine do a IS NOT DISTINCT FROM when the column is nullable, so USING will do its job;
and probably more performant because the verification of
b.AUTH_LIST = a.AUTH_LIST
is restricted to the case where both values ar not null ... that is to say, null values are verified before the equal.