February 1997
Distributed object computing (DOC) has received a substantial amount
of attention over the past several years. DOC uses OO techniques to
distribute reusable services and applications over multiple, often
heterogeneous, computing and networking elements. The most widely
publicized DOC tools are based on OMG's CORBA, Microsoft's Distributed
COM (DCOM), and JavaSoft's
Java
Remote Method Invocation (RMI).
With potentially lucrative technologies like DOC, there are always
opportunists who ``ride the wave'' to further their careers,
regardless of whether they truly understand or support the technology.
Over the past decade, I've witnessed this phenomenon with AI expert
systems, CASE tools, OO methods, software process maturity models,
C++, frameworks, components, patterns, and Java. It's remarkable to
watch the same opportunists re-position themselves time and again,
with every change in technological fashion and fortune.
Recently I've spoken with several notable technologists who feel that
the hype associated with DOC technologies is yet another example of
the triumph of marketing over technology. While I share their distain
for the shameless hucksterism surrounding DOC, I don't think it's
quite as easy to dismiss distributed object computing as it was to
dismiss earlier generations of over-hyped technologies like CASE
tools. Here's why:
- DOC products have been used successfully in practice to solve
real problems (particular for business applications and network
element management). In
contrast, CASE tools generally weren't useful enough to justify
their enormous price tags. CASE tools solved a relatively small
number of real problems (e.g., tracking dependencies between
documentation and code). But due to the high costs, many sellers
and buyers wanted to believe that they solved practically all
their problems and worked flawlessly.
- Some very technically talented engineers are building the
next generation DOC products at companies like IONA, JavaSoft,
Microsoft, and Visigenics. They are shipping increasingly good
products. In a year or so, I suspect that the existing performance
bottlenecks in DOC products will largely be alleviated.
- There are some very bright researchers extending and
standardizing complex aspects of DOC (such as
fault tolerance, replication, caching, and real-time quality of
service). For instance, take a look at the latest white
paper from the Real-time CORBA SIG,
which is authored by many respected members of the Real-time
research community. Within the next 2 years, I predict there will be a
number of credible real-time
DOC products available on a range of OS platforms.
- DOC standards, particularly CORBA, are improving with respect to
key features like portability, asynchronous messaging, multimedia
streaming, and end-to-end quality of service. At this stage,
it's more productive to work on improving CORBA, rather than
trying to start a new standard from scratch.
However, this is no call for complacency. I'm continually amazed at
companies and managers that are so desperate for distributed object
salvation that they completely ignore anything negative about existing
CORBA, DCOM, or Java RMI products. In this context, the hucksters
will keep right on peddling their (vapor)wares, unless we keep them
honest.
This issue of C++ Report contains three articles that focus on CORBA.
Jennifer Hamilton describes the use of Distributed SOM's
``DirectToSOM'' C++ compiler; Steve Vinoski and I present an overview
of the CORBA Events Service; and Rocky Stewart, James Storey, and
Dehua Huang illustrate how they've enhanced the CORBA Events Service
and applied it to a commercial CORBA-based telecommunications
management system. We'll be covering other C++ DOC technologies, in
particular DCOM, in subsequent issues of the C++ Report.
Back to
C++ Report Editorials home page.