Re: [Highly OT] Webservices are NOT distributed objects

Gerald Brose <[EMAIL PROTECTED]> wrote in message news:<[EMAIL PROTECTED]>...
> Mark Woyna wrote:
> > Gerald Brose <[EMAIL PROTECTED]> wrote:
> >
> > Does SOAP actually put the meta-data in the message, or a reference
> > to the meta-data, i.e. targetNameSpace? I'm no SOAP expert, but I
> > believe only a reference to the schema is included, not the 
> > entire schema file.
> 
> Correct.
> 
> > If so, this is not really much different than DII, since the client
> > must perform a remote call to retrieve the schema before parsing the
> > document for the first time.
> 
> I was referring only to the tagged document structure that
> lets you parse anything even if you don't have a schema.

But the schema defines the data types. I would imagine that many
types of applications couldn't do much with a simple string without
knowing how to convert it to a "proper" parameter.

> 
> > Now, I'm not saying that DII is efficient, but it could be implemented
> > in basically the same fashion.
> 
> What is missing is the ability to skip parts of a GIOP message
> without retrieving the full type information, and to do that
> you would need to modifiy the message layout.

But one only has to retrieve the full type information once. If you
have to make any call, you might as well retrieve the layout for the
entire message.

> 
> > True, but I'm in the camp that believes that Web Services provide
> > nothing that CORBA doesn't provide, or couldn't have provided with
> > a little bit of effort. Instead we're basically back to where we were
> > 10 years ago, trying to standardize the basic protocols and common
> > services, and ensure interoperability. On top of that, as Michi
> > pointed
> > out in a previous thread, we're stuck with a text-based protocol that
> > is inefficient for many types of applications. Lastly, just to add
> > insult to injury, I'm hearing that some people are shocked that XML
> > on the wire is slow, and there are pending proposals to introduce *yet
> > another* binary protocol!
> 
> Let's rephrase the point I am making again: to me, Web Services
> are technically a subset of CORBA, but with practical advantages
> for certain application types that do not demand high performance
> but benefit substantially from "loose coupling". You are
> challenging my claim thatWeb Services will be simpler to build
> for that application domain.
> 
> I have no statistics to prove my claim, only architectural
> arguments that somehow surround the point of loose coupling,
> especially not breaking interfaces when only minor changes
> are required. I also said that the XML world generally provides
> richer tools, which I believe is true for both modeeling and
> development.

I won't dispute that the tools available for XML are richer, only
that the same, or similar tools, could have been developed for
CORBA within the same timeframe. I don't believe the existence
of these rich tools preceded the adoption of XML as the language of
choice. 

> 
> > Imagine if Microsoft had dumped DCOM, and embraced CORBA, thus
> > ensuring cross platform compatibility. Imagine if Sun had 
> > selected CORBA and IIOP as the basis for J2EE, thus ensuring
> > cross-language interoperability, rather than Java-only RMI?
> 
> Sun's choice (for whatever actual reasons) was a clear indication
> that CORBA would not make it on the "mass market". If that had
> been an objective, it would have been wise to restrict the
> technical complexities that went into CORBA over the years
> and do more marketing. This did not happen. Obviously, making
> CORBA a headline topic was not a common goal.

I believe Sun's choice was driven by their desire to promote Java
everywhere, not the fact that CORBA couldn't make it on the mass
market. Sun has never been a big proponent of CORBA, or at least
never developed an orb that was worth using. Keep in mind that at
the time (~1997-8) there was no guarantee that Java was going to have
a server-side market, either.

The irony of this situation is that Microsoft was the one who more
or less pushed the idea of cross language interoperability! They pointed
to the fact that Sun wanted the whole world to use a single language,
but that they would allow you to mix languages. (Gee, what a interesting
idea). Of course, MS wanted you limit you to a single OS platform.

The mistake Microsoft made was that their .Net platform wasn't ready
for prime time. This gave IBM, BEA, Iona, and a handful of other J2EE
vendors a chance to establish J2EE as a viable platform for doing Web
Services. They realized that the Java everywhere message wasn't selling,
or at least that it would take a while before every legacy application
was rewritten in Java. Eventually, Sun reluctantly accepted Web Services
as a reality.

> 
> > Imagine if the OMG would have fixed the firewall problem 8 
> > years ago, when Netscape first put an ORB in a web browser,
> > "inventing" the Object Web [Orfali, Harkey, and Edwards], thus
> > allowing web-based RPC-style applications? Would we be have a 
> > need for "Web Services"?
> 
> The problem with the OMG is that it cannot do things by itself,
> you always need coalitions of vendors that see ROI for certain
> things, otherwise nothing gets done. So, in effect you can
> say: "if the larger ORB vendors had..."

But these same vendors, IBM, Iona, Borland, are firmly behind Web Services!
If they had shown the same commitment to CORBA as they're showing
with Web Services, the problems would have been fixed by now.

> 
> >>I guess most people are simply disgusted by the hype surrounding
> >>the XML world, which indeed is annoying. I am not trying to
> >>add to that, but I see practical value in XML Web Services.
> > 
> > I totally agree that there's a lot of hype. But more importantly,
> > there's
> > a massive amount of wasted effort in reinventing the wheel. We've
> > been down this road before. CORBA was, and still is, the standard
> > for cross-platform interoperability. Remember that the first "I" in
> > IIOP stands for "Internet". :-)
> 
> Sure, but we just mourned over the OMG's missed chances. As noted
> in a different thread, I truly believe that, e.g., security
> issues have a much better chance of seeing technical solutions
> in the XML world than with CORBA. Starting with a clean sheet
> can be very liberating sometimes, and although it may sometimes
> look this way, there are actually many technical experts working
> in the Web Services world that did learn their lessons from
> CORBA.

True, but the OMG has corrected several specs by basically starting from
scratch, so it's possible to start with a clean sheet within a
standards organization.

> 
> > Yes, the CORBA specs could use a diet, but the Web Services
> > specifications
> > are getting just as bloated. :-) Not simply because the writers don't
> > know what they're doing, but simply because that's the nature of
> > real-world distributed applications.
> 
> Again, most of the people I have met that work on Web Services
> are not dumb, and you do see grey-beards there. I agree that
> distributed apps are inherently complex, but it's a matter of
> defining application domains and requirements to be able to
> make an informed choice of which tool does the job best.
> 
> > On top of that, I'm *still* not convinced that a basic web service
> > application is any easier to build or understand than a basic 
> > CORBA application, if one uses the appropriate tools. :-)
> 
> Installing a current AppServer product (with CORBA under the
> hood, but without them saying so for marketing reasosn) can
> be very painful, and getting through the supposedly simple
> GUIs to generate thousands of classes and deployment archives
> is more complex than what I have seen with Microsoft .NET
> tools - just as a teaser ;-). (I know there are good IDEs
> with built-in IDL compilers.)

But this has nothing to do with CORBA. These same app servers are
being marketed as Web Services platforms. I guess you'll be saying that
Web Services are difficult in a few years because these same app servers
are still hard to install and configure. ;-)

> 
> >>The point about the expenditure may be true, but it is hard
> >>to tell how much of that is marketing and how much is
> >>engineering. 
> > 
> > Don't you mean "re-engineering"? There's not much new here. ;-)
> 
> Granted.
> 
> > P.S. By the way, I really like your CORBA/Java Book. I use it 
> > as the main text for my Distributed Objects course
> > (http://www.cs.iit.edu/~cs447/readingList.html).
> 
> Thanks. It could do with some updating, I guess. Let's see if
> I can find the time and if there is sufficient interest in
> a new edition.
> 
> Cheers, Gerald.

Mark