The Middleware Company has released a video interview with Salil Deshpande on the Application Server Platform Baseline Specification.

In this interview, Salil Deshpande talks about The Middleware Company's Application Server Platform Baseline Specification, the members of the expert group for the Spec, and looks at the concept of 'case studies' as opposed to benchmarks. He discusses the October 2002 .NET vs J2EE case study, the aftermath and controversy surrounding it, and performance issues in .NET and J2EE.

Results:
TMC indicated that results from several case studies based on this framework (e.g., a Performance case study, a Productivity case study, and an Interoperability case study) will be released over the next month or two.

TMC also stated that a technical interview will be going live next week which will talk about implementation details of the specification, what technologies were used, and why. It will also compare the TMC MPetstore to JPetstore.

View the interview:
https://www.theserverside.com/talks/library.tss#deshpande

View the Case Study home page:
http://www.middleware-company.com/casestudy

View coverage and complaints from the original petstore:
http://www.middleware-company.com/casestudy/coverage.shtml

Some questions answered are:

- What is The Middleware Company Application Server Platform Baseline Specification and why is the name so long?
- Who are the members of the expert group?
- Why did you limit participation to these experts? Why not just open it to the community at large?
- Why PetStore? Why not make it something entirely different?
- So why create a new specification when there are already SPEC jAppServer and TPCW?
- Let's talk about the October 2002 case study. You had some rather vocal critics; Were they invited to join the expert group?
- So with all these waves you created in the J2EE community, what were some of the results and aftermaths of the case study?
- So come on, you made some mistakes. Can you review what they were?
- You've touched upon some of the mistakes from the past; what's different this time around?
- So why do you think there was only a 10% increase in performance?
- This question is from the META Group: Were you surprised at the results of the October 2002 study?
- Another follow up question from the META Group: What does your experience leave you thinking about .NET and J2EE?
- What do you say to detractors of this case study and people who try to discredit you or the study?
- All right, so you mentioned other people having agendas. What's your agenda?
- What were some of the hot issues discussed by the Expert Group?

good honest discussion, pleasantly surprised.

As one of the experts invited to participate, I have to say that TMC did this in an entirely professional manner. I didn't agree with everything that ended up in the spec, but these were purely technical issues, and TMC have published all my comments and those of the other experts. I saw no sign of any hidden agendas, so I think TMC deserves the trust of the J2EE community. (Please note that I wasn't paid to participate and wasn't asked to post here, so I don't have any agendas here either.)

Rod Johnson, author of Expert One-on-One J2EE Design and Development

Of course you didn't see any signs of hidden agendas. They were HIDDEN :D What would a conspiracy theory have if it didn't have a complete lack of evidence?

Steve

Oh God! Here come the conspiracy theorists. There goes the neighborhood (again).

/Jim

Could you please make the code downloadable? It's linked to, but clicking the link leads to a "Forbidden" page.

Sorry. This is fixed now.

Salil Deshpande
The Middleware Company

I guess I don't understand why all this drama came about. Any moderatlely skilled developer knows that there are 3 kinds of lies:

lies
damn lies
statistics

Benchmarks, at best, tell you how a certain piece of code/application perform under a given set of conditions. It's tough enough to compare the performance of 2 java applications, let alone 2 applications written in different languages that have vastly different runtimes.

I go back to the Voting Booth guidelines on Slashdot:
"This whole thing is wildly inaccurate. Rounding errors, ballot stuffers, dynamic IPs, firewalls. If you're using these numbers to do anything important, you're insane. "

First of all, there might be more than 10 Java Petstore implementations in Java.
Can someone compare?

Petstore Xdoclet (Xpetstore EJB)
Petstore Xdoclet (Xpetstore Servlets)
Petstore JDO (Libelis)
PetStore Sun (EJBs)
PetStore Ibatis

...

Its worth to me compare Java architectures than .net x J2ee

While TMC has certainly made an obiovus effort to do a study in a correct professional manner this time and attmepted to remove issues concerning the Oct 2002 study they stil do not get it..

The temptatiosn are still there and without being open and frothright about those temptations TMC will always be looked upon through the oct2002 study rose colored glasses..

But that is a very minor point..centering on EJB as the major repsentative of J2ee vs NET is I think a vast mistake..

The focus should be on all of j22ee together vs NEt not just ejb..

Not to say that TMC may not recognize this,, in fact I believe TMC is currently working on 2 other case studies to prepare for this shift in j2ee vs NET debate..


It sjust that after al lthe efrforts on TMC's part its goign to take a more open forthright approach..

An exmaple might be open it up to allowing the Mono Project coders to particpate..they do not have an economic incentive to offer TMC and thus the temptation will be vastly lessor in degree..

Then of course is the debate of how wrong the orginal 2 versions of PetStore were and how some opensource groups designed a new PetStore implementation to repair the issue that were wrong..



A question I might have of TMC on this effort is does TMC see any value in using an independent set of expert j2ee and NEt audtiors to audit PetStore implementations they use in order to have a realistic factual/case example basis on which TMC and the developer public can factually rely on the results of the case studies?

> repsentative of J2ee vs NET is I think a vast mistake..

they are not, one of their codebases does not use EJB.

Hi Fred,

Many of these issues have been discussed in a prior thread:
https://www.theserverside.com/home/thread.jsp?thread_id=19407

The docs and codebases have been available for several months now. All are invited to modify and make submissions. Please see submission guidelines on the case study website. Or if you work for an organization that has resources, you can take everything we have and do your own studies and I'm sure TSS would be interested in publishing your results.

Salil Deshpande
The Middleware Company

A question I might have of TMC on this effort is does TMC see any value in using an independent set of expert j2ee and NEt audtiors to audit PetStore implementations they use in order to have a realistic factual/case example basis on which TMC and the developer public can factually rely on the results of the case studies?

Have at! (You can download all the implementations, or submit your own.)

There are only two things that I would have liked to have seen done differently in this "revisited" PetStore:

1) The proposed "rules" should have been made public from day one, and the process of finalizing them should have been public. TMC selected several dozen people (I was one of them) to look at the "rules" to comment and make suggestions on how to improve them, and tried to take that feedback into account. I don't think their approach was "wrong", but IMHO any time you can put the public spotlight onto something that will eventually become controversial, you reduce the chance of someone calling it unfair and conspiritorial.

2) The "rules" do define many architecture and design decisions. I lobbied for an official "no-holds-barred" category which would not be so encumbered, to encourage implementations that don't necessarily follow the "boring" ways of designing applications. ;-)

That said, I think that TMC has been pretty open with it, and tried to solicit outside feedback and make it a "public" process. If the implementations do shine, it is the community that will benefit, from finally having a PetStore implementation worth looking at ;-)

Peace,

Cameron Purdy
Tangosol, Inc.
Coherence: Easily share live data across a cluster!

So did I. I'd like to see a Perl or PHP Pet Store :-) There are a whole lot of technologies out there for developing web applications, and I think it's a mistake to focus on the J2EE vs .NET rivalry. Of course there are plenty of "enterprise" class applications that languages such as Perl aren't a rational option for, but these are a lot more complex than Pet Stores, and maybe we need that kind of benchmark.