--------------15FB748359E2B6001CFBAE39
Content-Type: text/plain; charset=us-ascii
Content-Transfer-Encoding: 7bit
Roger Session's newsletter has previous issues that document his
'conversion' between camps. This current story, though, illuminates some
of the more, uh, personal battles between standards players.
Substantively, I think he's right how far Microsoft has come, but the
crux of the argument is cross-platform or not? Because MS has never
convincingly delivered on cross-platform promises. (I've heard enough
screams of swapping agony from IE4Solaris... and even that's only one
UNIX).
Rohit
http://www.objectwatch.com/issue10.htm
--------------15FB748359E2B6001CFBAE39
Content-Type: text/plain; charset=us-ascii; name="issue10.htm"
Content-Transfer-Encoding: 7bit
Content-Disposition: inline; filename="issue10.htm"
[Image]
ObjectWatch Home - What's New - Newsletters - Book Store
[Image]
ObjectWatch Newsletter Number 10
Focus on Distributed Object Technology
There are now over 3,000 subscribers to this newsletter! This newsletter may
be freely redistributed as long as it is redistributed in its entirety and
is not modified. For your own free email subscription, send mail to
sub@objectwatch.com saying
subscribe <your-name>,<your-eMail>
To unsubscribe, send mail to sub@objectwatch.com saying
unsubscribe <your-name>,<your e-mail>
----------------------------------------------------------------------------
CONTENTS
Main Article - Roger vs. The 850 Ton Gorilla
Jeri Edwards Responds
The Netscape Fiasco -or- I Hate to Say I Told You So, But...
Invitation to the Microsoft Corporate Solutions Summit
Book Review - Essential COM by Don Box
Latest on Roger's Public COM/DCOM/MTS class
----------------------------------------------------------------------------
MAIN ARTICLE - ROGER VS. THE 850 TON GORILLA
Apparently, something I did upset it. I didn't mean to. I just thought the
gorilla had gotten too big, was moving too slow, and was no longer
competitive in the gorilla business. I had met another gorilla, one that was
nimble, moving quickly, and seemed to have the necessary smarts to find good
bananas.
So I switched gorillas. No big deal, I thought. I never meant the first
gorilla to take this personally. After all, who wants to be the target of an
850 ton gorilla's grudge?
But the 850 ton gorilla didn't see it that way. It doesn't like people to
leave its tree. As long as you continue stroking it, feeding it, and
grooming it, you are acceptable. But should your glance happen to wander to
another gorilla, it snarls. Should you stop your grooming, it threatens.
Should you actually have the audacity to climb out of its tree, the gorilla
goes ape (so to speak).
The good news is that 850 ton gorillas don't move very fast. The bad news is
they have a lot of momentum.
I first made the acquaintance of this particular gorilla seven years ago,
back when the gorilla was a mere 200 tons. IBM introduced us. I had come
down to Austin to work on an IBM skunk works project implementing object
technology. The project was called SOM, for System Object Model. I was to be
responsible for the persistence portion of the technology.
IBM soon became interested in the CORBA standard for distributed objects.
Mike Conners, the technical lead of SOM, decided to refocus the SOM project
as IBM's implementation of CORBA. This was the beginning of my six year
affair with the (then) 200 ton gorilla known as OMG (Object Management
Group).
The OMG is a software consortium. At the time IBM joined it had about 200
member companies at, I figure, one ton per company. While it is true that
many member companies are so small they barely tip the scale, they are more
than compensated for by companies like IBM, Sun, Intel, and Oracle. There is
no question about it. OMG was one big gorilla then. And it has quadrupled
since. It takes a lot of bananas to feed this gorilla.
While I worked at IBM, I was widely regarded as the primary spokesperson for
IBM's CORBA implementation, SOM. Many people described me as Mr. SOM. I
spoke at over 30 industry conferences on CORBA and SOM. I wrote more than a
dozen articles about CORBA and SOM. I was a lead architect of one of the
CORBA services and attended many OMG meetings throughout the world. I was
well known in the OMG and the IBM community.
After five years at IBM, I left to do CORBA consulting. I wrote a book on
the CORBA persistence service. I developed IBM's CORBA training program. I
attended an OOPSLA conference on behalf of the OMG (with all expenses paid
for by the OMG). I developed and taught IBM's portion of the OMG CORBA
Academy.
For six years, nobody in the OMG questioned my CORBA expertise. I was
feeding the gorilla, and the gorilla was happy.
But then a funny thing happen. One of the editors at John Wiley & Sons asked
me to write a book about Microsoft's COM and DCOM. This technology directly
competes with CORBA. At first I said I wasn't interested. I already had a
gorilla. But then I decided to have a look.
The more I looked at Microsoft's technology, the more convinced I was that
Microsoft had a vision behind its technology. COM and DCOM, I thought, were
reasonable alternatives to CORBA. MTS was addressing some fundamental issues
that CORBA had never even looked at. By the time my book was finished ("COM
and DCOM; Microsoft's Vision for Distributed Objects"), I had decided it was
time to change gorillas.
Since then I have publicly discussed the main deficiencies of the CORBA
strategy. This recently culminated in a Webweek article titled "A Critic of
CORBA [that's me, folks!] Knows Whereof He Speaks" by David F. Carr in the
December 12 issue
(http://www.webweek.com/1997/12/08/software/19971208-critic.html). The
article gave a rather balanced view of my main objections to CORBA. I
thought Carr did a good job. But the 850 ton gorilla was not amused.
In the January 20 issue of Webweek, the gorilla spoke up. In an article
titled "CORBA Community Responds to Sessions' Criticism", also by David F.
Carr (http://www.webweek.com/current/software/19980120-corbra.html) we heard
how the giant consortium addressed my technical concerns.
What were my concerns? Chiefly that Microsoft's Transaction Server, Message
Queue Server (Falcon), and Wolfpack provide important functionality that is
completely missing from CORBA. Included among these are component tier
scalability, automatic transaction management, automatic security
management, failover, and robust asynchronous communications.
How did the OMG spokespeople respond to these issues? By explaining why
these are not important areas? By listing all of the specifications that
address these areas that I had somehow missed? By showing how CORBA vendors
implement these in a portable, coherent manner?
Not exactly. According to the article, "The response from the Object
Management Group is that Sessions is exaggerating his claim as a CORBA
expert. 'As far as I know, he was involved in exactly one CORBA service
specification, one that was so bad no one ever implemented it, and it's now
being replaced,' said Richard Soley, CEO of the industry consortium."
In other words, I , who barely understood CORBA, managed to convince the
OMG, the world's largest technical consortium, to accept a bogus
specification for one of its most important object services, the one
defining Object Persistence. The best minds in the software industry were
hoodwinked by little know-nothing me. Not bad, for a simple country boy. You
would think the OMG would be too embarrassed to admit this.
Jeri Edwards continues where Richard Soley left off. According to the
Webweek article, "I don't want to slander Roger," she said, "But I do think
in this area it's worth noting that he has a book out, and nothing sells
better than controversy in terms of books."
In other words, because I wrote a book on COM/DCOM, all I care about is
generating controversy, selling my books, and making huge piles of money.
Obviously, book authors can't be trusted. Jeri, on the other hand, will give
us her unbiased opinion on all this.
Who is Jeri Edwards, you might ask. Now Jeri Edwards happens to be a Vice
President of BEA, a corporation that has bet its future on CORBA. You can be
sure that Jeri has huge stock options that will be worthless unless CORBA
makes it big time. Do you think she might have her own little gorilla to
protect? And perhaps her own not-so-little stash of bananas to look after?
Oh yes, and if being the author of one book on COM/DCOM makes me less than
trustworthy, Jeri might have mentioned that she is the co-author of four
books on CORBA.
I don't really mean to put down Richard Soley, Jeri Edwards, or any other
individual associated with the OMG. I have always considered Jeri to be a
great writer and recommend the books she has co-authored with Bob Orfali and
Dan Harkey as the best sources of information available on CORBA. I have
worked closely with these and many other talented OMG associates over the
years, and have learned a great deal from all of them.
But we must ask why these people spend their time attacking me personally
rather than addressing the fundamental technical issues I keep raising.
These issues include:
- CORBA is an unscalable architecture because of its reliance on component
state. The central role the Name Service plays in the CORBA model is
inconsistent with a Transaction Processing Monitor model for component
scaling.
- CORBA component code is not portable. It never has been and it is getting
worse, not better. BEA's Iceberg, IBM's Component Broker, and IONA's
OrbixTPM are all moving in their own proprietary directions.
- CORBA is difficult to program because of its heavy reliance on programmer
APIs rather than runtime environment.
- Customers using CORBA are using functionality defined five years ago. Not
only has the Persistence Service gone nowhere (which Soley says is my
fault), but few of the other specifications have gone anywhere either. Most
customers today are using only the most basic IIOP ORB specification (for
ORB inter-operability). I have yet to meet a single customer using the
Relationship Service, the Event Service, the Query Service, the Licensing
Service, the Properties Service, the Time Service, the Trader Service, or
the Collection Service. Even services as basic as the Naming Service are
rarely used. How does Soley explain the underwhelming popularity of these
services, none of which were tainted by my unholy presence?
- The OMG is a political organization, not a technical organization. Its
cumbersome processes give it little maneuvering room in this rapidly
evolving technology. Although the OMG rules require that specifications be
based on existing technology, in fact this rarely ever happens. The
consensus process favors political compromises, not technical advancement.
- Microsoft has already made up for a three year OMG head start and has now
leapfrogged CORBA technically. There is little if any chance the OMG will
ever overtake Microsoft again.
These are interesting questions, and you would think the OMG would want to
discuss them. Why, instead of addressing these important issues, does the
CEO of OMG tell people that I don't understand CORBA? If I don't understand
CORBA, now is your chance to show the world! Show us how stupid I am!
Please!
Here is my offer to Richard Soley. Rather than debate personalities, I
challenge you to debate technology. I would be happy to meet you in a public
technical debate at a conference of your choosing. I will explain the
problems I see with OMG and CORBA and you can respond. You can explain the
problems you see with COM, DCOM, and MTS, and I will respond. Then we will
let the technical public decide which approach makes sense and which
doesn't. And, by the way, who understands what and who doesn't. Are you
willing?
So these are the exciting questions for this month. Will the 850 member OMG
agree to a technology debate with little old Roger Sessions, a poor country
boy who doesn't understand these oh so complicated issues? Will the OMG
crush Roger Sessions and leave him nothing but an empty mental shell? Will
Roger Sessions ever be able to look a banana in the face again? Stay tuned
for the thrilling conclusion.
- Roger Sessions
Austin, Texas
1/30/98
----------------------------------------------------------------------------
JERI EDWARDS RESPONDS
In the interest of fairness, I sent Jeri Edwards an advance copy of my
article and offered her an opportunity to reply. She sent the following
letter, which I am happy to reprint in its entirety and without editing.
- Roger Sessions
.....
Roger,
As you know, I do not intend to start a mud-slinging contest with anyone,
much less with you. So, thank you for offering me this opportunity to
respond to your concerns directly, without the press in the middle.
As you accurately point out, I'm Strategy VP at BEA and coauthor of four
books that support CORBA. But you confuse cause with effectBEA's strategy
and our book's soap boxes reflect my views, not vice versa. While there are
certainly differences between CORBA and DCOM (see discussion in Orfalis,
"Client/Server Programming with CORBA and Java") the key issues are
strategic, not technical:
* CORBA lets 'the better implementation win.' Unlike DCOM, CORBA defines a
set of interfaces and protocols, not an implementation. True, vendors must
work together to test that various implementations interoperate. But this
approach ultimately results in a dynamic market that produces the best
products and the lowest prices. Where you see this as the inability of the
CORBA community to agree, I see this as the basis for a vibrant,
multi-product marketplace based on open standards. It lets customers
mix-and-match best-of-breed solutions.
* CORBA is cross-platform. This is an absolute requirement for enterprises
and the Web. I hear this from customers every day. Their installed base and
need for scalable servers means they will continue to use a variety of
systemsnot just NTfor the foreseeable future. Software/AGs DCOM port is
not the answer. It carries a prohibitive price tag and doesnt support MTS;
and MTS, as you frequently point out, is what makes DCOM a server platform.
In contrast, many vendors provide cross-platform CORBA ORBs and Object
Transaction Monitors (OTMs), including BEA.
* Transactions are very well integrated into CORBA. As you point out,
transactions, scale, and distribution are key to mission-critical computing.
Thats why companies like BEA and IBM are implementing CORBA-based OTMs
(Object Transaction Monitors). Both companies have early versions of their
products on the market. You accurately note that MTS is already generally
available, while these products are still in beta. But MTS as it stands is
just a toy. Its an OTM that doesnt scale, doesn't perform, doesn't support
distributed applications, and is not cross-platform. MTS can only be used
for small applications today. In contrast, BEA Iceberg is based on an engine
tested in Tuxedo, the leading distributed TP Monitor. Therefore, it
automatically garners years of experience running millions of transactions a
day. Finally, even if Microsoft had an enterprise-class product today, they
wouldnt know how to sell or support it in enterprise shops. Supporting
these customers requires a completely different sales and business model
than the one they've perfected for the commodity market.
* CORBA is working on a MOM-based protocol over IIOP. I agree with you that
MOM is important. That is why BEA bought MessageQ from Digital. We are also
contributors to the CORBA messaging specification and are integrating our
advanced MOM capabilities (based on MessageQ and Tuxedo \Q) into Icebergour
OTM/ORB. IBM also incorporates asynchronous communications into
ComponentBroker (its OTM) using its MQSeries product. Both MessageQ and
MQSeries are leading MOM products, seasoned in many enterprise applications.
By comparison, Microsofts technologyFalconis embryonic.
* CORBA and Java are coming together nicely. The proof of this is the
Enterprise Java Beans (EJB) specification and RMI over IIOP. Note that these
JavaSoft specifications were the result of collaboration by many key
players, including Oracle, IBM, BEA, Netscape, Sun (of course), Sybase,
Gemstone, and most of the tools vendors. Java is not a sidetrack from CORBA,
but a key element of our direction.
In summary, few enterprises have committed to DCOM as their only networking
model; it just doesn't meet their requirements. However, vendors and
customers alike acknowledge the importance of Microsoft and their COM and
Active/X 'standards' on the desktop and in workgroups. So, software vendors
like BEA support it in their solutions (see our Desktop Connection and
ActiveX Builder products).
If Microsoft really wanted to win the enterprise market they'd adopt CORBA,
thereby coopting it. Their DCOM-only vision is not a strength: It will make
them a niche player in the enterprise. But, lets not tell them that. :-)
Best regards,
Jeri Edwards
----------------------------------------------------------------------------
THE NETSCAPE FIASCO
- OR -
I HATE TO SAY I TOLD YOU SO, BUT...
In newsletter # 8 (http://www.objectwatch.com/issue9.htm), I discussed the
plans of Netscape to develop a version of its web browser written entirely
in Java.
I quoted Marc Andreessen, Netscape's Executive VP of Products, who had said,
"Java will eventually prove to be as fast, if not faster, than today's
operating system-specific programming environments." I said that this
statement was unbelievably naive, and predicted that this effort would be a
failure.
In that newsletter, I also made the following statement:
"The industry will be watching this port very carefully. If Netscape
succeeds in creating a competitive product and delivering it on time, this
will be a major vindication for the 100% pure Java approach. If Netscape is
unable to deliver on this promise, it will be the kiss of death for the 100%
pure Java bandwagon. So the success or failure of this venture will have
ripples that will reach far beyond Netscape itself."
According to the San Jose News, Netscape is now giving up on Java. According
to the article, "People working on Java projects at Mountain View-based
Netscape are among the hardest hit in a current round of layoffs, taking
place for the past several days. Netscape has said it would lay off about
400 of its 3,200 people after reporting it would lose up to $89 million in
its most recent quarter."
Based on this news, I think we can pretty much kiss the 100% pure Java
browser goodbye. Unfortunately, it is the programmers who are now paying the
price for Marc Andreessen's misguided vision. And, as I said in that
article, if this effort fails, it will be the start of the unraveling of
Sun's vision as well, that of Java being the operating system, the whole
operating system, and nothing but the operating system. Hopefully now we can
settle down to thinking of Java as the great programming language that I
have always believed was to be its final destiny.
The San Jose news article can be found at
http://www.sjmercury.com/business/center/netscapejava012498.htm
----------------------------------------------------------------------------
THE MICROSOFT CORPORATE SOLUTIONS SUMMIT
I would like to offer a special invitation to the readers of the ObjectWatch
Newsletter to join me at the Microsoft Corporate Solutions Summit on
February 19 at the Rosemont Convention Center just outside of Chicago.
This is a free conference, and I will be giving the keynote address! So
come, and don't be late!
To register, call 1-800-550-4300 or check out the conference web site at
http://www.microsoft.com/usa/midwest. I'll see you in the windy city!
Here is an abstract of my planned keynote talk:
Microsoft has let loose the opening volley in what promises to be a rough
and tough battle for the hearts and minds of corporate programmers. At the
center of this struggle is control of the platform for N-Tier applications.
The next generation of N-Tier applications will be based on components.
Object-oriented programming is pass. Components are the objects of the
future. Microsoft is working hard to make Windows NT the platform of choice
for component oriented applications and in particular, those applications
that run the world of commerce.
COM and DCOM have for years been Microsoft's lingua franca of components.
Microsoft is now investing heavily in a set of advanced technologies that
can collectively be described as component oriented middleware. Microsoft is
offering a tempting deal to today's software developers: you develop your
business logic as COM components, and we will provide you with all the run
time support you need for scalability, transactions, security, robustness,
and more.
This talk introduces components and the idea of component oriented
middleware. We will show how COM, DCOM, MTS, Falcon, Wolfpack, MSMQ, and
other Microsoft technologies all fit together. And don't throw away your
Java, C++, and Visual Basic. They have their important roles to play as
well. Components are the future of software, and the future is now.
----------------------------------------------------------------------------
BOOK REVIEW - ESSENTIAL COM
Essential COM by Don Box, published by Addison-Wesley (1998). ISBN
0-201-63445-5
I can't remember the last time a book's publication was awaited with such
eager anticipation. But Don Box's book is finally out, and I'm happy to say,
it has been worth waiting for. This is a book by a great writer about a
topic he knows and loves. This is The Essential COM by The Essential Don Box
It's important to understand for whom this book is written. This is a book
for advanced C++ programmers with some experience programming COM who now
want to become advanced COM programmers. If you are not a hard core C++
addict, you will have trouble with this book. And this is not the book to
get your first introduction to COM either. But for those of you in the
target audience, this will become your Bible.
In some ways, this book complements my book (COM and DCOM; Microsoft's
Vision for Distributed Objects). My book is very high level vision, looking
at the whole area of components and component oriented middleware. Don's
book is getting down to the dirty details of getting your design to work in
COM. If you are just starting out with COM, I suggest you start with my
book, and then move on to Don's.
Don starts out by discussing some of the problems with C++. He then shows
C++ programming techniques you would normally use to solve these problems.
But Don has tricked you! He has led you unknowingly into the world of COM.
It turns out that the techniques you would use to solve C++'s problems are
almost exactly the techniques COM uses. And from there, you are on a high
caffeine look at every bit in every COM flag.
A few words of warning. First, this really is targeted at C++ programmers.
Java and Visual Basic programmers will find this book confusing or downright
scary, as much of the bit/byte discussion doesn't apply to them. Second,
there is virtually no information about MTS and how this will be changing
applications design. This is not a major criticism, since the material in
this book will still be valid. You can think of MTS as being a layer on top
of the material Don is presenting.
If your idea of breakfast food is Jolt Cola, then this is a book by one of
your own. For C++/COM programmers, this will be THE book of 1998.
You can purchase Don's book from the ObjectWatch on-line bookstore,
http://www.objectwatch.com/bookstor.htm
----------------------------------------------------------------------------
LATEST ON ROGER'S PUBLIC COM/DCOM/MTS CLASS
Many of you have asked when the public COM/DCOM/MTS class will be offered
that I keep promising. My first public class has been slightly delayed due
to a heavy corporate class schedule and preparations for the Microsoft
Corporate Solutions Summit.
I have almost nailed down the date as being Thursday, March 5 in Austin,
Texas. As soon as I am sure, I will send out mail to this list. Plan on
signing up early, because I am strictly limiting registration to the first
20 participants. If you are not on this mailing list, now is the time to
sign up.
Here are some of the topics we will cover in this one day class:
- Component oriented programming
- Packaging your business logic as components
- Distributed components over a network
- Creating a four-tier component based architecture
- The difference between fat client and thin client programming
- Using MTS as an efficient component runtime environment
- The importance of stateless components
- Designing for scalability
- Making use of object pools and just in time activation
- The critical security algorithms you must understand
- Using role-based security for your components
- The importance of asynchronous communications
- Using Falcon's message queues
- Transparent transactions
- The real reason two-phase commit is critical to your design
- Using relational databases within the Microsoft architecture
- How clusters will change your notion of hardware
- How Microsoft's technology compares to CORBA
- The weaknesses in the Microsoft architecture
These are critical areas YOU MUST UNDERSTAND to design large web based
applications in the Windows NT environment. If you are not familiar with
these areas, DO NOT PASS GO, DO NOT COLLECT $200, DO NOT GO ANY FURTHER
until you have taken this class.
I hope to see y'all (or at least 20 of y'all) in Austin Texas this March.
----------------------------------------------------------------------------
Roger Sessions is the author of COM and DCOM; Microsoft's Vision for
Distributed Objects, just published by John Wiley & Sons. This book is a
gentle introduction to the three-tier scalable distributed object
architecture of Microsoft. This book can be purchased at your local computer
book store, or online at http://www.objectwatch.com/bookstor.htm.
Roger Sessions is the President of ObjectWatch, Inc. He is the author of
four books and dozens of articles and a frequent conference speaker. Visit
our web page at http://www.objectwatch.com or contact Roger Sessions at
roger@objectwatch.com for details on how ObjectWatch can help your
corporation succeed with objects.
----------------------------------------------------------------------------
Recent issues of this newsletter are available at http://www.objectwatch.com
# 9: Scalability in Three-Tier Component Systems
# 8: Reflections from the Colorado Software Summit 1997
# 7: Java - The Microsoft Perspective
# 6: Letter From the Microsoft Professional Developer's Conference
# 5: The World's Largest Software Company
----------------------------------------------------------------------------
Legal Notices:
The ObjectWatch Newsletter does not accept advertising or rent out its
subscription list.
This newsletter is copyright by ObjectWatch, Inc., Austin, Texas. It may be
freely redistributed provided that it is redistributed in its entirety, and
that absolutely no changes are made in any way.
ObjectWatch is a registered trademark of ObjectWatch, Inc., Austin, Texas.
All other trademarks are owned by their respective companies.
----------------------------------------------------------------------------
--------------15FB748359E2B6001CFBAE39--