UNIQUE INDEX unaware of transactions

It seems that our current way of enforcing uniqueness knows nothing 
about transactions ;(

when you 

create table t( i int4 primary key
);"""

and then run the following query 

begin; delete from t where i=1; insert into t(i) values(1);
end;

in a loop from two parallel processes in a loop then one of them will 
almost instantaneously err out with  

ERROR:  Cannot insert a duplicate key into unique index t_pkey

I guess this can be classified as a bug, but I'm not sure how easy it 
is to fix it.

-------------
Hannu


I tested it with the followiong python script

#!/usr/bin/python

sql_reinsert_item = """\
begin; delete from t where i=1; insert into t(i) values(1);
end;
"""

def main():   import _pg   con = _pg.connect('test')   for i in range(500):       print '%d. update' % (i+1)
con.query(sql_reinsert_item)

if __name__=='__main__':   main()