Ovum’s recent report Ovum Evaluates Object Request Brokers dismisses the Object Management Groups Common Object Request Broker Architecture (CORBA) as a flawed architecture which could become a doomed standard. These dramatic conclusions have been widely quoted, but the analysis on which they are based is itself deeply flawed. A close reading of the 370-page report reveals research that is seriously out of date. The report lists the first four Corba services, plus eight more still being defined, and refers to them by a name (COSS) that has not been used for over 3 years. In fact, 15 services have now been published and more are on the way. Ovum states that OMG has more than 500 members, which is what OMG was saying two years ago. There are now over 780 members. The statement that Enterprise Objects – from NeXT, is a planned product helps with the carbon-dating process, as this product first shipped in October 1994. Although there are well over a dozen Corba products on the market, Ovum only evaluates four ORBs (Sun’s Neo/Joe, BEA’s ObjectBroker, Iona’s Orbix, and Visigenic’s VisiBroker) and briefly reviews a fifth (IBM’s DSOM and Component Broker Connector), plus Microsoft’s Distributed Common Object Model (DCOM).
Serious factual errors
Admittedly, those chosen are among the market leaders, but what about TCSI, Hewlett Packard, Fujitsu, ObjectShare, DNS, Chorus and Siemens-Nixdorf? And why did Ovum choose to ignore two home-grown British ORBs, ICL’s DAIS and DOME from Real Objects? ORB vendors point out that Ovums line on the internet Inter-ORB Protocol (IIOP) is uncannily close to taunts that Microsoft used to employ two years ago. Ovum analyst Laurent Lachal has declared that IIOP is a technology that doesn’t work, not at all. Its a reference specification, and everyone is implementing the naming services on which IIOP relies differently in each ORB. Only Sun and IBM are committed to making it work. These three sentences contain four serious factual errors. (1) IIOP simply maps Corba messages onto TCP/IP – it has nothing to do with any Corba services. (2) IIOP works perfectly – like HTTP and SMTP, it is simply not the sort of thing that can go wrong. (3) The whole point of having a vendor-neutral specification is that everyone can implement it differently as long as the results are the same. Lachal’s complaint is like criticizing a Hitachi TV set because it uses a different power supply from a Sony TV. (4) Not only IBM and Sun, but all vendors of Corba 2 products are committed to IIOP. They have no choice – it is a mandatory part of the specification. Another Ovum criticism that has not been true for a year or more is that Corba products do not interoperate. Actually any two properly implemented Corba 2.0 products interoperate with each other.
By Tom Welsh
The Distributed Systems Technology Center (DSTC) in Australia has been running an open demonstration called Corbanet for over a year, which can be accessed on the web. It features at least seven different commercially available ORBs, all of which interoperate with all the others. The report also alleges that the increase in the number of platforms supported has not been realized. Most vendors support only the Unix and Windows platforms, and even then only the better known Unix platforms are supported. Actually Corba is supported on at least 30 platforms, including MVS, OpenVMS, Tandem, AS/400, OS/2, Macintosh, a dozen or more Unix variants, and all Microsofts platforms. In fact Corba supports more Microsoft platforms than DCOM does. There are Corba products for DOS, Windows 3.1, Windows for Workgroups, Windows 95 and Windows NT, whereas Microsofts DCOM supports the last two only (plans to port DCOM to Macintosh were recently abandoned). Some ORBs, such as Orbix, support as many as 20 platforms. Others, like Suns and Hewlett Packards, run only on their own hardware. But this does not matter, as all Corba 2.0 products interoperate with each other. Another unjustified criticism is that the Corba architecture is not suited to high performance, reliability or availability. This would come as a surprise to Chorus Systemes and Real Objects, which offer optimized real-time ORBs, as well as the UK Home Office Immigration Service, whose Suspect Index system – based on ICL’s DAIS ORB – ran continuously for its first six months with no downtime whatsoever. Ovum’s statement that Corba is hardly being used seriously seems to rest on a myopic view of the market. Boeing relies on Corba to link software from Baan, SDRC/Metaphase, CIMLINC and Trilogy in what has been called the world’s largest business process redesign project, to be rolled out to 45,000 users. Hongkong Telecom IMS is using Java and Corba (with IIOP) as the middleware backbone for its $1bn Interactive Multimedia Services video-on-demand system aimed at tens of thousands of set-top boxes. Motorola built the vital control software for the Iridium satellite communications network around Corba, and there are now 22 Iridium satellites in orbit, each carrying and relying on an embedded Corba ORB.
Rampant author bias
ABN Amro, Bank of America, Britannia Building Society, British Aerospace, British Airways, CNN, Deutsche Presse-Agentur, McKesson, Telstra, Vodafone, Wells Fargo Bank and Zeneca are building or have built production systems with Corba. OMG lists no fewer than 261 organizations that are creating, adopting or using Corba-based systems. The report complains that there is little tool support for Corba. In fact PowerBuilder, Dynasty, FortT, IBM’s VisualAge, Neuron Data Elements, Hitachis Object IQ, Templates Snap, Antares ObjectStar, SuperNova, Rational Rose, Ptech, Select Software, Unify Vision, Cayenne’s ObjectTeam, Aonix StP, and the leading Eiffel and Smalltalk tools either support Corba or will do so in the near future. Jens Christensen, chief technology officer at Visigenic, sums up the unspoken feelings of most Corba vendors and users. The Ovum piece is technology fiction that is based upon a vast amount of outdated, simply false information Ovum customers are paying for sloppy, outdated research that lacks any real foundation for its conclusions other than rampant author bias.
Content from our partners
