> With postgres once an error occurs in a transaction
block you need
> to rollback. None of the transaction will commit.
>
> This behaviour makes sense as it assumes that the
transaction block
> is atomic and it should all succeed or all fail.


This is VERY counter-intuitive. I can have really
important data
for say 5 tables which has committed properly but at
the 6th
insert into a non-important auxillary table, I may
encounter a
transient exception. I still want to be able to commit
my data.

There are many similar scenarios such as the above,
right ?

As a programmer, shouldn't it be upto me to decide
when to
commit and when to rollback ? Is this even within
spec ? And
at the very least, commit() should then not fail
SILENTLY ! (and
this should be documented).

:-]

Best regards,

--j



__________________________________
Do you Yahoo!?
Yahoo! Mail - Helps protect you from nasty viruses.
http://promotions.yahoo.com/new_mail

---------------------------(end of broadcast)---------------------------
TIP 6: Have you searched our list archives?

http://archives.postgresql.org