Re: Implementing sequences without sequences

From: Ilia Bashtanov <iliab_at_mmc.ru>
Date: 1997/04/10
Message-ID: <01bc45c6$200b3da0$25d9bec2_at_ib>#1/1


What about

create sequence seq_name start with 1 increment by 2

for the first database and

create sequence seq_name start with 2 increment by 2

for the second?

-- 

~~~~~~~~~~~~~~~~~~~~~~~~~~~
Ilia Bashtanov
iliab_at_mmc.ru
phone +7 (095) 974 30 30
fax   +7 (095) 974 30 20
~~~~~~~~~~~~~~~~~~~~~~~~~~~

Sadru Fidai <sfidai_at_castle.net> wrote in article
<3349A841.528B_at_castle.net>...

> Hi,
> We are making our application replication-aware. In this process, we
> have user control the sequence segments. E.G. a sequence in database A
> may go from 1-100, 201-300 while the same sequence in database be would
> go from 101-200, 301-400 and so on. If I use table based next number
> generator, I will have to commit everytime I get next number.
>
> I thought of an idea. Use regular sequence. Once that expires, use
> DBMS_JOB package to submit a job which will drop and create the
> sequence. The idea was that, DDL will happen in the separate session
> without affacting current transaction. Unfortunately, I just learnt
> today that using DBMS_JOB procedures 'submit' and 'run' perform implicit
> commit. Oracle support confirmed this. This throws is idea out of the
> door.
>
> I kicked around other ideas. Like using DBMS_PIPE package but that is
> not so apealing. This and other ideas need a continous process which we
> think is not a good idea.
>
> Our implementation is a client/server architecture and use
> Developer/2000 on the front and Oracle 7.3.X at the backend.
>
> Does anyone have any other suggesion?
>
> Sadru Fidai
>
Received on Thu Apr 10 1997 - 00:00:00 CEST

Original text of this message