Compromise is in the air again, this time over the Object Management Group’s CORBA 2 object request broker specification and the decision on how to make different vendors’ object request brokers interoperate. After two and a half days of discussions at the last Object Group meeting in Berlin, the participants fell into two broad camps – some endorse a common protocol or Universal Standard Profile model, while others back a gateway approach. IBM Corp, Hewlett-Packard Co, Digital Equipment Corp and NEC Corp sit in the first category, under the Open Software Foundation umbrella. They submitted a Distributed Computing Environment-based proposal, whereby the Interface Definition Language requests mappings to DCE calls. Expersoft Corp is also included in the first category, but its approach is based on its own proprietary Exportable Object Reference communication protocols and a Universal Request Forwarding Option. This uses TCP/IP protocols and NDR encoding techniques. ExperSoft reckons its model could potentially encompass the other submissions, although the rest of the vendors don’t seem quite so sure. SunSoft, Iona Technology Ltd and another submitter, Bell Northern Research, on the other hand, plumped for the gateway or translator method, whereby multiple object brokers to co-exist in a single address space. And finally, ICL Plc submitted a reference implementation independent framework model, based on ODP, into which both approaches could potentially be incorporated. The firm has now written a document describing this general architecture, and has contacted all submitters with a view to forming a working group to discuss it and the relative merits of the different approaches. The pros and cons of the two technologies appear to be this. A gateway has the disadvantage of slowing things down because it forms a bottleneck when undertaking transactions between client and server machines. It has a complex topology, which means that only specific pairs of Object Request Brokers can interoperate at any given time. On the other hand, it can communicate with protocols other than standard computing ones such as TCP/IP. A common protocol, meanwhile, does not cause the above problems, but cannot bridge different protocol domains – for example, it cannot deal with specific telecommunications protocols, a disadvantage when many existing object users happen to be in just that market. There’s some consensus behind the ICL proposal but interested parties have until September 21 to revise their submissions.
Content from our partners
