Open Hypermedia Systems Working Group

Beans, RMI, Voyager

Recommendations

In the evaluation of CORBA by Ken Anderson, three main requirements of the Open Hypermedia Protocol were identified, being the ability to:

• specify a set of related components.
• develop and run the components cheaply (e.g. for free) and on multiple platforms.
• exist in an Internet environment (e.g. with associated scale and security issues)

RMI is in its current incarnation not able to work with anything but other RMI programs. This does not exactly open up for a large set of related components, as RMI is not very widespread at moment.

The second requirement is RMI programs will work without recompilation on any platform supporting the Java Virtual Machine. As the JVM has been ported to nearly all computer platforms, RMI programs have a platform independence superior to that of CORBA or DCOM. As the Java Development Kit (of which RMI is a part) are available for free from JavaSoft, the price of this technology is hard to beat.

The third requirement is partly met by RMI. Ever since Java evolved from Oak into Java, the language and its class libraries have been designed to be the net programming architecture of choice. This is reflected in the Java security model, which is probably one of the best around and ensures via the use of security managers that the possiblities open to Java programs are clearly defined and strictly enforced. The 'sandbox' approach to security taken by JavaSoft seems to be safer than the after-the-fact-you-can-always-sue digital signature security of ActiveX. The other aspects of Java being well suited of the Internet (such as dynamic class loading and platform independence) are well known and well described elsewhere. RMI is however a lightweight ORB and is yet to prove itself as a technology that scales well.

The crux of the advantages and disadvantages with RMI is the Java-centric design. By limiting the architecture to Java, RMI have become nice, easy to use, and very well integrated with the rest of the language. RMI certainly seems to be very much easier to program than CORBA and DCOM.

However this comes at the cost of sacrificing all other programming languages and platforms. I am personally in favour of the Java platform, and it certainly has become a very fast growing architectures, but I think that in the context of OHSWG the sacrifices are too great. That is of course unless we decide to do everything in Java.

Please send feedback to Niels Olof Bouvin. Last modified: Mon Dec 15 07:03:45 1997