Discussions

News: Don Ferguson interviewed on IBM and J2EE

  1. Don Ferguson interviewed on IBM and J2EE (14 messages)

    Don Ferguson is IBM's lead for the EJB and J2EE specification evolution; and was recently interviewed by JDJ. Don answers questions about Websphere 5, netbeans competition, SWT, Corba/J2EE integration, .NET, etc.

    Read Don Ferguson answers readers questions about IBM and Java.

    Threaded Messages (14)

  2. [extract]
    WebSphere provides a workflow engine that choreographs business processes in which the activities map to J2EE Java calls or invocations of Web services. It also supports a business transaction model with automated compensation-based recovery.[...] For Web services, there are two fundamental concepts: the "container" and the "bus". The container is the environment in which the service actually executes, and we have put a large amount of work into WebSphere to make it the Web services container. This is also the environment that provides support for current and evolving Web services standards like BPEL4WS, WS-Coordination/WS-Transactions, etc.
    [/extract]
    The vision of an internet scale, loosely-coupled, flow-coordinated bus is starting to become a reality. Here is a pointer to a great article from Adam Bosworth [1] on the same topic. IBM can talk the talk. But can they deliver?

    Edwin K.
    [1] http://www.cs.ust.hk/vldb2002/VLDB2002-proceedings/papers/S01P01.pdf
  3. I really appreciate IBM's support for and contributions to J2EE and the open source community, but that interview has to be taken with large grains of salt...

    <quote>
    There is no serious community around Forte or NetBeans. (...) We've seen that the massive participation in eclipse.org - two million downloads, 175 tool vendors committed to creating commercial Eclipse-based tools - demonstrates that there was a compelling need for it.
    </quote>

    They are probably right in that respect. Although NetBeans does have a community, it's rather centered around users of the IDE, less around tool providers. NetBeans resp. Forte/Sun ONE Studio Community Edition is a decent free IDE with some interesting special modules, but Eclipse definitely has its place and constantly gains mindshare.

    <quote>
    With WebSphere, IBM was the first major vendor to have a J2EE 1.3 certified product.
    </quote>

    In what world does IBM live in? Do they really count their unsupported WebSphere 5 "Preview"? And WebSphere 5 GA is just being released these days, isn't it? They are literally years behind with their production version...

    <quote>
    We have no plans to support JDOs. For J2EE, we recommend EJB 2.0 support for CMP entity beans, which we believe is simpler and better. Application development tool support for JDBC also provides a better model than JDOs.
    </quote>

    Perpetuating traditions is a fine thing, especially in that case, isn't it? Or should we call it ignorance? "Simpler and better", or arguably the worst choice for fine-grained persistent objects? Note that I don't want them to support JDO within WebSphere in the first place, just to refrain from such oversimplifying recommendations!

    Juergen
  4. <quote>
    We have no plans to support JDOs. For J2EE, we recommend EJB 2.0 support for CMP entity beans, which we believe is simpler and better. Application development tool support for JDBC also provides a better model than JDOs.
    </quote>

    What is the world they are talking about? I just happen to see Rod Johnson's new book and it clearly shares my idea thta entity beans have to die. It is so bloated that they are now "inventing" a language to support it - EQL ! I feel JDOs are the future. You cant do anything worthwhile with EJBs. I have faced several mission critical failures in IBM's own clients while working contracting for IBM due to over use of EJBs. Just my 2 cents worth thought...
  5. The world of J2EE according to IBM[ Go to top ]

    <Sanjaya Ganesh>
    You cant do anything worthwhile with EJBs. I have faced several mission critical failures in IBM's own clients while working contracting for IBM due to over use of EJBs. Just my 2 cents worth thought...
    </Sanjaya Ganesh>

    Do not generalize. EJB does not equal entity beans.
  6. Need detail on your comments[ Go to top ]

    Sanjaya, I do not follow the thinking behind your comments:

    "It is so bloated that they are now "inventing" a language to support it - EQL ! "
    How is the introduction of EJB QL evidence of, or a consequence of (alleged) EJB bloat?

    "I have faced several mission critical failures in IBM's own clients while working contracting for IBM due to over use of EJBs"
    You state yourself that there was EJB "overuse" (which I say is improper use). In that context, how do you draw any conclusion about the value of EJB? or how it compares to JDO?
    You might have legitimate concerns, but these comments are too vague to be of value
  7. <Ferguson>
    With WebSphere, IBM was the first major vendor to have a J2EE 1.3 certified product.
    </Ferguson>

    <Hoeller>
    In what world does IBM live in? Do they really count their unsupported WebSphere 5 "Preview"? And WebSphere 5 GA is just being released these days, isn't it? They are literally years behind with their production version...
    </Hoeller>

    WebSphere 5.0 Technology for Developers was not an unsupported preview; it was indeed a product, and it was the first product from a major vendor to be J2EE 1.3 certified. It was supported through a web forum, which had a fair amount of traffic. IBMers were assigned to monitor the forum and respond to queries.

    As Ferguson stated in one of his comments, there are two constituencies IBM is targeting: the early-adopter "developer" group who wants to play with the new APIs, and the "deployer" group who requires stability, scalability and the other "-ilities" needed for production use. WebSphere 5.0 TD was aimed at the first group; WebSphere 5.0 at the second.

    Randy Schnier
    IBM WebSphere Development
  8. IBM was late with J2EE 1.3, period![ Go to top ]

    Here's the reality as I understand it. In January 2002 IBM delivered WS Technology for Developers, which was not supported for production use. I think the IBMers are saying that it was "supported" in some general developer kind of way, but unless I'm mistaken, you couldn't put a Message Driven Bean into production, so let's get real and face the facts there.

    Somewhere in that general timeframe BEA and others put out equally non-supported versions. Up to this point, it was all a bunch of hand waving.

    The real deal was when BEA launch WebLogic Server 7.0 in April, with PRODUCTION SUPPORT FOR J2EE 1.3. WebSphere 5.0, the production version, came out when...December 2002. That's 8 months behind WebLogic.

    IBM people, correct me, or drop this claim to "first with J2EE 1.3" when we know its a crock. Alpha-ware doen't help me solve my problems. Production code does...

    Looking forward to any clarificaitons.

    Harvey
  9. IBM was late with J2EE 1.3, period![ Go to top ]

    I have to agree with Harvey. I could write production J2EE 1.3 EJB's with no problem on Weblogic 6.1 long before WLS 7.0 came out, and they ran perfectly fine on 7.0, no changes.

    Oh, I like your eclipse product though.
    Steve
  10. IBM was late with J2EE 1.3, period![ Go to top ]

    |
    |I could write production J2EE 1.3 EJB's with no problem on
    |Weblogic 6.1 long before WLS 7.0 came out, and they ran
    |perfectly fine on 7.0, no changes.
    |

    In fact, Weblogic had a "developer release" of some features of EJB2.0 (namely MDB's) which ran on Weblogic 5.1 (!) before 6.0 was released (in January 2001)!

    I just dont get IBM sometimes. They have all the cash resources and the engineering talent to be first to market with standards support - instead, they choose a proprietary implementation (aka WAS4E Message Consumption from MQSeries only) when MDB's had been a stable part of the spec for ages.... why?

    As for claims that Websphere 5.0 TD was a "product" - I disagree. You couldnt buy it and it didnt come with a support contract. It was definitely a J2EE1.3 "implementation" but a "product" is stretching it. What is more, it cannot be said that WAS5.0 was just a well tested version of WAS TD. I can tell you that as of July 2002, there were many core features still under development.

    -Nick
  11. Depends on your point of view.[ Go to top ]

    Harvey, your statements about timeline are reasonably accurate, except that WebSphere 5.0 TD was actually released in December 2001. However, nothing you said contradicts the fact that WebSphere 5.0 TD was a product, it was supported by IBM, and it was J2EE 1.3 certified, the first product from a major vendor to do that. It was definitely not alpha-ware; as I stated before it was targeted at the J2EE developer set. Being targeted at a specific type of customer (an important set of customers), does not mean it's not a product...not all products are targeted at the "deployer" customer set I mentioned in my last post.

    I don't recall exactly what BEA was making available at that time relative to non-supported versions, but whatever it was, it had not been demonstrated to be J2EE 1.3 certified. As I'm sure you're aware, that set of tests is not easy to pass and to be able to do it is a demonstration of a vendor's ability to produce a correct, valid J2EE product.

    I'm gathering from your remarks that when BEA came out with WL 7.0, they said it was OK for customers to run in production. That's fine, customers can of course make their own decisions on the level of code they are comfortable with putting in production. If some customer went ahead and ran WebSphere 5.0 TD in a production setting, that would be their decision too. I once worked with a customer who was considering using the WebSphere Test Environment in VisualAge for Java as their production J2EE server, but fortunately we were able to talk them out of it :)

    As always, customers make the final call. By this measure, WS 5.0 TD achieved what it was designed to do. A J2EE 1.3 certified developer-targeted product evidently wasn't the most important thing you were looking for, which is fine. There are a sizable number of customers though who were looking for just that.

    Randy
  12. Depends on your point of view.[ Go to top ]

    I don't think IBM support shares your view to use your best judgement and put something like Technology for Developers into produciton. I could swear I read something at the time that suggested that it wouldn't be supported, which is warning enough for me. That's my only point. If I can't put it into production for 12 months after release, it's not a release of the full products. I'm not buying the spin. We had MDBs in production well before IBM or Oralce had even released something, so our app server choice couln't include them if we wanted to do anything asynchronous.

    Next you are going to tell me that WebSphere is easy to configure...woof.

    Harvey
  13. In "Bitter Alternative"
    the author implies that SQL should be used for complex apps and O/R for simple ones.
    O/R does not gain anything, since in the end someone has to know SQL to save to DB, and not OQL, EQL, etc.

    This is incorect IMO, a thread on servlets vs EJB explains O/R vs JDBC RowSet.

    SQL should allways be used.

    .V
  14. I'm not in the thread hijacking business, but I had to reply.

    <hijack>
    I just had a look at the "Bitter Alternative" chapter that you reference, and it in no way implies what you state. Are we reading the same thing?

    Who would use a O/R Mapper for a supposedly simple solution? It would be massive overkill. Are you telling me TopLink is better put to use in a small database driven website than plain hand-written SQL? The cost of learning to properly use the tool alone would probably dwarf the complexity of the project.

    OTOH, why write all that SQL if you have a sufficiently complex application, when it can be easily generated from meta-data? Or do you think that you can generate data acccess code for 150 tables faster than a code generator based on the database schema? Sure the generated code may only get you 80% of the way there, but I deal with the other 20%.

    I think you are way off on this...
    </hijack>

    You can now return to your regularly scheduled thread.
  15. OOOPs Chris,
    I posted in the wrong thread. Please reply in thread on Bitter.
    Here is a quote from book:
    From the Alternatives chapter in book:
    "When dealing with complex query processing on relational data sets, SQL can be an indispensable component of an architecture. However, using SQL for simple queries can be problematic. In general, Java application developers are more comfortable thinking in object terms than in relational terms, and will sometimes create quite inefficient queries or queries that do not do exactly what the developer intended."