Total Trash! At the very least, the guy could write a few paragraphs outlining why it is dead, instead of saying "its too expensive" so buy our crap. I expect more from JDJ than this, maybe I shouldn''t.

The best Sun can do to follow BEA Systems, Gartner Group and IBM vision on Application Platform Suites and SOA.
But it is nstill ot enough to catch the front runners.

Several respondents picked up on this immediately: Another thinly disguised (if it were disguised at all) sales pitch, with a direct attack against IBM.

What "Jonathon" fails to recognize is that it is crazy (cost prohibitive) to rip and replace legacy systems. Oracle pitches a similar message (one source - Oracle), I think. A company who does that risks its own demise: while they spend money rearchitecting existing processes and functionality, their competition will push them into oblivion.

Web Services is one of the many keys to the future, along with information integration (both EAI and EII). Middleware helps make that happen.

jackprogrammer: If you can write a program to generate marketing hype, I am not sure if you are a genius or sick...

A catchy title yes but what you are offering is just your own middleware with Sun branding. In short a sales pitch for your own product. Lame. Joining the party rather late don''t you think? A dollar short and a day late. Must have read those briefs by Gartner that this area of the market is hot!

This is just another sales pitch for the newest fashionable abstraction. And those web services, what do they run behind the scenes, if not middleware ?

This is really useful stuff for all of us Java programmers. Please give me more! and more! Long Live the Schwartz! I think I could write a program to generate this sort of garbage.

Hello John,

Yes, the lack of Canonicals costs me far more in up-front integration effort than differences in format and protocol. Couldn''t agree more. It''s amazing that format and protocol get so much of the SOA discussion bandwidth.

But if the challenges in establishing Canonicals were mostly technical, I wouldn''t still today be having semantics discussions with EDI trading partners. How long will it take for semantic (not technical) conformance to emerge? Recent experience with UCCnet is making me think this (the semantics, not the mechanism/technology) is much harder than most people thought.

Regards.

Mr. Rosenthal -
Yep, I work a few levels under Tony Scott. So regarding decoupling the semantics from the app that creates it... Take an example of the UDEF ID''s being assigned to each data element in a schema, or each data element in an RDF taxonomy. This exists as an online resource, and there would be resolver services on the wire to perform lookups and transforms in real time. By the way, an example UDEF ID appears as d.t.2_8, which is literally translated as purchase.order.document_identifier. So the ID is an attribute of the actual data element.

Then consider an approach like the new Content Assembly Mechanism (CAM) TC in OASIS to actually compose the content.

Using common, or Canonical, based formats like OAGIS, we can achieve real human-free integration, where the formats are well defined, and all data elements are labeled with UDEF IDs.

See more on the CAM, OAGIS and NIST proposal for a Proof of Concept at the udef.builders discussion group. [visit link]

john

Hello Yadi,

If shared services through standard protocol and formatting make creating point-to-point connections between applications easier, than I submit they are evil.

If in a few years, a shared service is replaced with another which has somewhat different content and semantics, then each and every application that was coded to directly use that service may need to adapt. In an environment with a large number of point-to-point connections between shared services, the arithmetic of change becomes absolutely frightening. Inter-application "spaghetti" is not a result of multiple protocols and formats or even semantics, it''s a result of multiple point-to-point application connections.

If I had to make a choice between point-to-point shared services versus hub-and-spoke messaging through proprietary interfaces, I''d choose hub-and-spoke every time.

Fortunately there is no need for such a choice. Shared services will just make the job of connecting services to my middleware layer easier.

Regards.

John, Huy and all other dear readers and software engineers:

I strongly believe the whole problem with the article is in the different interpretations of the word middleware. Mr. Schwartz is referring to middleware as opposite of shared service. I understood what he means by middleware is really repeated, un-sharable, scattered services that are floating in an enterprise. If this is what he is referring to then he really has to fix the terminology. And if he is referring to such services, I agree with that. The age of un-shareable services has come to an end.

But what is next. SOA techniques have not really matured to a level that we can say SOA implementation solves all the problems. For example we can talk about reuse and SHARED SERVICES. A service cannot be called a shared service until others outside a project use such service by adoption and integration with their system. So even term shared service is a very delicate term to use. This is an extensive discussion by itself. The other issue with shared service is versioning. How do we version such service? How do we deprecate old services? There are many answers but what is the unified, standard answer to such issue.

SOA as you said has to go through many trials and implementations until we can come up with a set of standard practices and patterns, which can be used in most enterprise implementations.

By the way, in my earlier posting I said that Mr. Schwartz is right in his assessment. I should have said also that it is true if he thinks of services (e-mail, portal, application services, and etc) being middleware not the real meaning of middleware(the glue).

Now if I am correct in my assessment of the article here are some good definition for Mr. Schwartz to consider. Let look at some of the definition I found on the web:
1. In [visit link] it says: Layer of software that is between client and server processes that deliver the extra functionality. This basically refers to RPC, SOAP, RDA and etc..
2. In [visit link] : Middleware: Software that connects two otherwise separate applications OR separate products that serve as the glue between two applications. It is, therefore, distinct from import and export features that may be built into one of the applications. Middleware is sometimes called plumbing because it connects two sides of an application and passes data between them. (For example, there are a number of middleware products that link a database system to a Web server. This allows users to request data from the database using forms displayed on a Web browser, and it enables the Web server to return dynamic Web pages based on the user''s requests and profile.)
3. In [visit link] : Software that mediates between an application program and a network. It manages the interaction between disparate applications across the heterogeneous computing platforms. The Object Request Broker (ORB), software that manages communication between objects, is an example of a middleware program.

Yadi Hooshmand
Sadra Inc.
www.sadrasoft.com

Hi John

I assume the following

> GM use Web service and XML to exchange information (or online transactional invocation) over some standard protocol such as HTTP. This requires high bandwidth
> Significant upgrade of hardware and software infrastructure to accommodate a new SOA

I would carefully consider the following

> ROI of services within internal as most organization have similar development standard and platform (except case of so much difference è I would question about the standards governance practice)
> Security of key services as far as I know this is still key issue with WS
> Provide a level of playing field to partners with Web Services is a fair call. However, be more careful with high volume transactions as high bandwidth and slow consumer will cause issues with expired transaction at the provider end. I face this once by mistake with our design / architecture.

Well I believe Middleware still around for a while and certainly Jonathan Schwartz are too optimistic of his SUN strategy with Enterprise Java offerings.

Messaging still dominate the bridge and message-compliant product vendor will watch SOA closely ever. We need some good and large SOA implementation for all of us to learn and increase its profile so key players are getting serious about this.

It is good to hear you are doing this. Please email, as I would love to discuss and to know how you are going with this with its achievement and any challenge you may have faced.

-h

Technical Architect
American Express
Operations JAPA

(e) huy.nguyen@consultant.com

Good morning John,

Is it any wonder EDI won''t die? Even with phonebook-sized X.12 manuals and carefully constructed implementation guides, we still work through semantic issues with EDI.

The XML based messaging standards have a long way to go before reaching even that level of conformity. It will take multiple organizations with influence like GM to push this forward. To me, this looks like a problem that is years away from a solution.

In the meantime... as you pointed out previously, how do you deal with differences in semantics, never mind format and protocol? And as I mentioned earlier, I would add the need to de-couple content from the specific application that produces it, allowing for a change in application architecture over time. Don''t hard-wire your enterprise business content semantics to an arbitrary set of application silos that existed at one point in time. Middleware is the solution of course.

By the way, it''s always entertaining to listen to Tony Scott. Do you work in his organization?

Regards.

Huy - Regarding the SOA from large corporations, I am currently working as a Chief Architect at General Motors, where the focus of the group I work in is not only delivering ebXML to a large group of trading partners (30K +) but also an internal enterprise wide SOA.

And I agree with your comments: the protocol I believe should be one or several of the OASIS and W3 based methods, specifically Schema and CAM to name a few. I believe that several will work in concert to provide these kinds of services.

- john john.hardin@gm.com

Agree in some aspects

The answer to that need to include protocol to bridge various systems which often have a completely different infrastructure such as O/S, protocol, languages, threads, etc.

IMO, Java Desk Top and Enterprised is not the answer. One of a credible solutions to that is Service Oriented Architecture which as I mentioned still need to convince from big players.

-h

Middleware will be dead, as soon as there is an answer to the semantic equivalency problem! One of the purposes of middleware is to bridge between formats that have the same data elements, but that are named differently. For example in OAGIS is the same as in xCBL. If a company needs to transform from OAGIS to xCBL, there is a human mapping effort, and map code in middleware to execute.

The Universal Data Element Framework is gaining steam in standards bodies to provide an open standards based mechanism to resolve the semantic equivalency between disparate formats.

Check out [visit link] and [visit link] for info including an upcoming NIST and OAGIS proposal for a Proof of Concept.

john hardin
chief architect, general motors