Oh, and I used MS SQL Server to test that code. Looks like there's also a portability issue :)

I figured you knew about naming columns when inserting, I just wanted it in the record for any SQL newbies out there.

I guess to be completely clear and sure of your results, you'd have to do the following to be equivalent to the left outer join version (in MS SQL Server at least):

SELECT pt.surname, pt.town 
FROM (
  select people.surname, town.town
  from people, town
  where people.person_id = town.person_id (+)
) pt
WHERE pt.town <> 'Washington' AND pt.town IS NOT NULL
[download]

I just realized that although you've requested a left outer join, what you get is the same as an inner join because all of the columns in your example match up. You need to add a person who isn't in the town table at all:

insert into people (
  person_id, surname
) values (
  5, 'Drifter'
)
[download]

What really surprised me was that you got Jones in the results - it seems like MS SQL Server did this:

SELECT people.surname, t1.town 
FROM (
  select *
  from town
  where town <> 'Washington' AND town IS NOT NULL
) t1,
  people
WHERE people.person_id = t1.person_id (+)
[download]

Essentially, it pared down the town table before joining - probably good for performance, although it gives strange results. Interesting - you learn something new every day here at sqlMonks. What? This is perlMonks?

---

Update
That part about it not being an inner join instead of a left outer join? Forget about it - it is a left outer join because of MS SQL Server's apparent order of operations (see my last SQL snippet above).