What is Your opinion about using datatype TIMESTAMP as the Primary Key of database tables?

It works well for me:

a) I don't have to use a sequence (one object less).

b) I don't have to set up another (in my case required) column "DATE_CREATED".

c) Space savings 11 bytes against 16 - our standard for sequence generated primary keys or even 20 - our standard changes recording sequence

But there are doubts also:

a) The distribution of column values is not uniform (as that produced by the sequence) any more, so the key should not hash well. I have no access to tkprof so is somehow useless for me to set up benchmarks.

b) The use of recursive insert (packaged) procedure (all tests went well - simultaneous insertion from three sessions using 1..100000 loops produced 300000 rows -> no missing rows). Here is the template:

procedure do_insert(p_a varchar2,p_n number,p_d date,the_count in out nocopy number) is
begin
the_count := the_count + 1;
insert into t (the_key,a,n,d)
values (timestamp(9),p_a,p_n,p_d);
commit;
exception
when dup_val_on_index then
if count < 1000 then
do_insert(column_values_in,the_count);
else
raise_application_error(-20000,'Retry count exceeded');
end if;
end;

Can You comment on what was said above. Do You see any other drawbacks? My proposal was rejected with the argumentation that TOAD and some in house made tools cannot handle timestamps.

Thanks and wish You long life (despite my almost daily visits to he site it is the first time in four years to catch the opportunity to submit a question)

ops$tkyte%ORA10GR2> select vsize(1234567890123456789) from dual;

VSIZE(1234567890123456789)
--------------------------
                        11