RMI Lives! But Where?

In "CORBA, Java, and the Object Web" (October 1997), the authors' assertion that "JavaSoft will abandon the proprietary ORB on which RMI is currently built" is refuted by the folks at JavaSoft. If by this the authors mean that the ORB possibilities will no longer be "limited" to a proprietary Java Remote Method Protocol (JRMP), then I stand corrected. But I believe most folks would take it to mean that JavaSoft intends to join the herd and dump JRMP, which most of us have been led to believe is not true. Those who are developing under RMI and need pass-by-value capability and an "all-Java" implementation of distributed objects believe that there is a c omfortable market position for RMI.

Hal Arnold
harnold@telegroup.com

JavaSoft will continue to support RMI over JRMP until the Internet Interoperable ORB Protocol (IIOP) is fully RMI-ready. This is consistent with what's stated in our article. JavaSoft will first deploy a subset of RMI on IIOP as it is today. JavaSoft is also working with Netscape/Visigenic on the OMG object pass-by-value and RMI-to-IIOP specifications. These specifications might provide the full RMI semantics over IIOP. (Netscape/Visigenic's Caffeine is an example of this architecture.) Hopefully, the specifications will have been presented -- in final form -- to the OMG by the time this appears in print.

When IIOP is fully RMI-ready, JavaSoft might continue to support JRMP for backward compatibility. But JavaSoft's partners intend to support RMI over IIOP only in their virtual machines. Of course, Microsoft doesn't support either. Enterpise JavaBean s support only the RMI/IIOP subset. They need IIOP to pass Java Transaction Service (JTS) transaction contexts.

JRMP is dead. Microsoft doesn't support it, and neither will BEA, IBM, Netscape, Oracle, and so forth. However, it shouldn't matter to you if IIOP is the only ORB that JavaSoft (and friends) support as long as it provides full RMI semantics. RMI will have a long life, but on top of IIOP. -- Robert Orfali