Hi Mircea,

>when you design a system for backup/recovery you want to be sure that
>if the testing is ok for 2 or 3 systems then everything will work fine
>with all of them.

Good point. It�s add another perspective. In a server <--> server 
replication (hub to hub), at this point I would even think about using 
different dbms.From ours, we did not mean we would like Object replication 
with backup/recovery features "no matter" your rdbms. As you say, we do 
not want to add heterogenity to the application with major unwanted 
consequences.
We just think, based in some own replication experience, in a client <--> 
server replication (hub & spoke architecture), the client (the 
offsite/offline app client) tipically just need to read master tables (RO 
access) and just need to add/update records impacting in a DB subset. For 
this situation,we envisioning the clients using small footprint db�s (like 
HSQL). 

But no matter the different requirements we might identify in the 2 
scenarios (client <--> server or server <--> server replication), objects 
replication is still grabbing our attention.

Thanks for the input and best regards,
Gustavo.



[EMAIL PROTECTED] 
29/11/2003 05:59 a.m.
Please respond to
"OJB Users List" <[EMAIL PROTECTED]>


To
"OJB Users List" <[EMAIL PROTECTED]>
cc

Subject
Re: Object replication vs RDBMS replication (little off topic)






Hi Gustavo,
> Let&#194;&#180;s dream for a minute: OJB will let us replicate objects 
across
> different dbms, without even having the same schema (due to the
> repository.xml mapping magic!!!!)
IMHO, this will be a dream for some time. Our purpose is, in the end,
to have the same set on objects on different machines. The assumption
of different providers for databases is not a realistic one because
when you design a system for backup/recovery you want to be sure that
if the testing is ok for 2 or 3 systems then everything will work fine
with all of them. Adding such heterogenity to your application can
have major unwanted consequences.

Of course, we can implement a 'SmartPersistenceBroker' that send data
to its subscribers and execute localy some broker.op(), but keeping
those subscribers syncronized is not easy. You may take a look over
the replication packages available for PostgreSQL or BerkeleyDB.

>Finally and again: Thanks Thomas and the whole OJB team!.
Thanks Thomas for this framework, I look forward to see it at version 1.1 
:)

Best regards,
--
Mircea Lazar

---------------------------------------------------------------------
To unsubscribe, e-mail: [EMAIL PROTECTED]
For additional commands, e-mail: [EMAIL PROTECTED]


Reply via email to