[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
Re: [dist-obj] CORBA implementation ideas
- To: <email@example.com>
- Subject: Re: [dist-obj] CORBA implementation ideas
- From: "Mark R. Thomsen" <firstname.lastname@example.org>
- Date: Mon, 14 Dec 1998 22:48:57 -0800
- Delivered-To: mailing list email@example.com
- Delivered-To: moderator for firstname.lastname@example.org
- Mailing-List: contact email@example.com; run by ezmlm
Heck, why not add a couple of experience points. I should state that after
watching others in pain, and not being a fan of RPC generally (for reasons
of strong type and temporal coupling), I am not a fan of CORBA. A bias.
Informed, but a bias.
1) Several years ago the USG/DoD/agencies were defining the new generation
of systems. They liked OO. They build distributed systems. They had used
the other RPCs that precursed CORBA. OMG and CORBA were big at the time. A
few experiences with DCE were OK.
So naturally they dived into IDL and CORBA to define the new systems. This
interface. These objects. They especially wanted to use network-based
discovery to overcome stovepipe systems, to provide types of flexibility not
found before in their systems.
Since then a number of development efforts have gone on with CORBA. Large
efforts. And problems have developed. Among them
a) The expertise requirements go up - not down - to work CORBA. The DoD
explicitly wanted to lower the expertise required to make this area happen
for their systems.
b) A lot of decisions in design and implementation above and beyond CORBA
must be carefully coordinated.
c) The projects are in more trouble than usual, including metrics that cover
injecting new technology.
d) IIOP performance, to put it delicately, sucks.
2) Our company works with Active Software. Their founder is Jim Green,
previously the head of the area that did DOE, CORBA, and such at Sun.
Active appeared out of attempts to use CORBA to build integrated systems in
Sun's manufacturing area. What they discovered is a lot of additional
software had to be written ... CORBA is largely fancy RPC, and a low level
Names. Management of higher level concepts. Decisions, such as distributed
memory management. All require substantial effort above and beyond mere
In their latest release they have eliminated the ORB entirely. I have
talked at length with Jim Green and other technical staff at Active. Their
words on CORBA are not positive, some of them being among the most
experienced with distributed messaging and function techniques I have met.
My Conclusion. I would rarely tell someone to use CORBA. Certainly I
hesitate in the real world. Perhaps it might be a good path to learning in
an academic environment. But I would first ask the participants to read
Hoare's CSP book, to get a grasp on issues central to distributed systems,
and to give some perspective on CORBA vs. alternatives.
Mark R. Thomsen
Alodar Systems, Inc.
To unsubscribe and other requests, mailto:firstname.lastname@example.org
Archives, FAQ, etc. http://www.distributedcoalition.org/mailing_lists/