cbsmith at envise.com
Sat Dec 26 09:50:19 PST 1998
Gerhard Paulus wrote:
> if you are interested in RMI you might have a look at Weblogic's RMI
> implementation :
> They claim to have a drop-in RMI replacement which is more efficient and
> faster then Sun's reference implementation.
The catch is that it's really not an independant RMI implementtion. It
really is a binding of RMI to their "Tengah" communications sever, which
is completely proprietary. :-(
> I looked at RMI in JDK 1.1.3 or so because I thought it might be useful
> to design a distributed OODBMS. At that time I had the impression that
> Sun's implementation is really limited if you want to allow a large number
> of exported objects. Any client seems to open a socket connection for each
> and every exported object s/he is working with. For a small number of
> objects that's OK but what if an application needs more then the available
> number of ports and for some reason you cannot split the application between
> servers ?
Sun does provide some multi-plexing in their protocol, although I
believe it is largely to get around firewall issues. In general, RMI was
never really meant to be a huge, robust, enterprise level, distributed
objects system. For that you use CORBA. RMI's real value is that it
provides a very simple manner to provide distributed objects to small
Of course all that being said, once there's a GPL'd implementation out
there, there's nothing to stop somebody from writing an improved wire
protocol with better performance. ;-)
> Anyway, for the OODBMS (Jacob) I decided to implement a messenging
> system based on raw sockets.
I've heard of Jacob before.... Do you have a URL to it? I'm always
curious about Java-based OODBMS's.
More information about the kaffe