Carlos H. Reimer wrote:
> Hi,
>
> I would like to use Slony-I to replicate our database but some tables don?t
> have primary keys defined. One alternative is to let Slony-I define it for
> me but in this case some applications could be affected, right?
>
>   
Yes, a couple of problems can arise:

1.  There is a new failure mode that no longer existed; applications
could insert values to the table and encounter failures due to
non-uniqueness of the new column.

That shouldn't happen if you're not messing with the sequence that is
used, but it's a potential problem.

2.  SELECT * from some_table;

...now has an additional column with a pretty funky name.  Potentially,
the new column might not forcibly be "at the end", where apps might be
prone to ignore it...
> Depending how the SQL statements are used we could have more or less work,
> or am I wrong?
I'm not sure what you mean by "more or less work."

- The added column means that the table gets a little bigger.

- The added index consumes some resources, both disk space and general
resource consumption.

- If the table is large, it will take some time to alter the table to
add the new column, populate it, and then index it.  That could be an
undesirable outage...