Snelle doorstart - formule 1?

De voorbije dagen heeft iedereen - vooral dan de VLD - de mond vol van snelle doorstart. Wie die term bedacht heeft, weeet ik niet. Maar mij doet het eerder denken aan Formule-1 wedstrijden. Een snelle doorstart lijkt mij dan iemand die tegen te hoge snelheid door de pitlane raast. Daarvoor bestaan in Formule-1 ook straffen: stop & go penalty. Iemand moet dan verplicht terug de pits in en moet een tijd in straf blijven staan, vooraleer opnieuw te mogen vertrekken.

Ben benieuwd wie er bij de snelle doorstart in de nationale politiek de stop & go penalty krijgt: Yves Leterme? Stoppen, ja, maar go? Go home, zeker?

Professionele afscheidnemer

Over Yves Leterme gesproken... Vroeger - ongeveer 15 jaar geleden was er de "professionele afscheidnemer", Xavier Debaere, als vaste afsluiter van "Morgen Maandag". Onze premier lijkt eerder de "professionele ontslagnemer". Hoe vaak heeft de heer Leterme de voorbije 2 jaar niet ontslag genomen? Een overzichtje...

Allereerst was er in 2007 zijn ontslag als minister-president van Vlaanderen, om vanaf 15 juli 2007 zijn taak als formateur op te nemen. Toen de formatiegesprekken echter niet al te vlot verliepen, bood hij op 24 augustus 2007 zijn ontslag aan. Toen werd Jean-Luc Dehaene als verkenner op pad gestuurd. werden de koninklijke bemiddelaars aangesteld.

Na een nieuwe formatiepoging, die van start ging op 29 september 2007, en vele beleidsnota's kwam Yves tot de vaststelling dat de verschillen tussen Vlaanderen en Wallonië te groot waren en er geen toenadering mogelijk was, geen compromis aangaande een verregaande staatshervorming en legde op 1 december 2007 er opnieuw het bijltje bij neer.

Ongeveer een jaar geleden - ook tijdens Music for Life - moest Guy Verhofstadt de brokken lijmen, een voorlopige regering starten (gevormd op 21 december 2007) om de dringende zaken aan te pakken en vóór Pasen zou Yves dan toch premier mogen worden. Maar... in januari ging het met zijn gezondheid niet al te best en moest hij met spoed worden opgenomen in het ziekenhuis. Na ongeveer een maand, toen zijn situatie stabiel was, werd hij ook daar ontslagen...

De paasklokken brachten op 20 maart 2008 dan toch een nieuwe premier en er werd naarstig gewerkt aan een grote staatshervorming. Maar iedereen wist ondertussen al lang dat MR niet zou meestappen in dit verhaal, want dat zou betekenen dat zij federale bevoegdheden moesten overdragen aan de PS, die in het Waalse gewest de meerderheid heeft. Gevolg... half juli 2008 bood Yves opnieuw zijn ontslag aan. Hij nam zijn verantwoordelijkheid op, was de verklaring. Dat werd echter niet aanvaard door de Koning en enkele dagen later ging alles weer zijn gewone gang, want... Yves moest zijn verantwoordelijkheid opnemen (Huh??).

En dan was er de Fortis-saga (waar ook enkele ontslagen vielen aan de top van Fortis), het arrest van het Hof van Beroep, de aanwijzingen van politieke inmenging in het gerechtelijk apparaat en uiteindelijk het ontslag van de premier dat uiteindelijk op 22 december 2008 door de koning werd aanvaard...

De premier heeft dus zijn verantwoordelijkheid genomen in dit dossier en heeft op professionele wijze afscheid genomen van de regering (maar niet voordat eerst Jo Vandeurzen geslachtofferd werd). En al die tijd bleef DJ Reynders buiten schot. Hoewel het wel degelijk om een financieel dossier ging. Benieuwd wie hier nu een snelle doorstart gaat nemen... met of zonder stop & go penalty.

The annual Devoxx conference is already "recent history". I posted my last comments on conference talks. This is the right time to do a small recap of the conference.

The university days: not an overall success for me, at least not for all the talks I attended. Luckily I didn't have to choose between university and conference (I chose for the full conference). There were better conference talks of which a combination could have been a good replacement for university talks. Highlights for me are Ivar jacobsen's talk, the university talk on Websockets, the different conference talks on REST, including the talk on Spring 3.0 and the talk on workload estimation. A colleague of mine will probably add the "french fries" to this list :-).

The daily Parlays magazine offered extra background information for a number of sessions; in case you were still in doubt what to choose, these articles could help make up your mind. It was nice to see my blog quoted in 2 Parlays issues (on Thursday about WebSockets and on Friday about this year's buzzword). To me this means that the Devoxx-people think I have something interesting to tell... :-)

Overall the conference was a big success. Apart from the known "issues" (catering, wireless connection during the first days and hardly usable browser clients in the cyber café), this was, again, a great conference, an excellent opportunity to keep up with the evolution. This was my 6th time; I'll be looking forward to the 7th.

The effort of 8 months preparation for this conference is definitely worthwile. I certainly hope Stephan drove home on Friday afternoon with a big smile on his face.

The last talk of this year's Devoxx conference that I attended was at bit in the extent of the previous talk: it was about XML persistence. The speaker seems to be a regular "conference closer", because I remember his talk of the last day of last year's Javapolis conference on ESB's and messaging, with a main focus on the finance and banking business.

This talk was related to his talk of last year: how to persist XML messages. I was very interested in this talk, because I have been involved in projects where we had to persist XML messages of dealing room activities. According to his "that was acquired by Oracle" (referring to both TopLink and Tangosol) and "see your license fee doubled", he didn't seem to be very fond of the O-company. He adviced not to opt for vendor specific solutions regarding to XML persistence, but stick to open standards as much as possible.

You don't want to dissect an XML document inito its composing entities and create tables for them. This would lead to a far too scattered relational database model, with joins that fill an entire page and take ages to execute. Relational databases just aren't meant for this purpose. They haven't even evolved with today's needs. To illustrate his statement, he showed a hierarchic representation of a simple swap (dealing room product) modeled in FPML (Financial Products Markup language).  The best way to represent an XML document, is an object. The best way to query XML-documents, is XPath.

His conclusion was a bit surprising, but, given the current technological situation, not such a bad idea: keep everything in memory as much as possible and only persist for archiving purposes. Why? Memory is cheap. A server nowadays comes with at least 32 GB of memory. You can have a cluster of servers, say 20 servers (= 640 GB of RAM, minus the memory required for running applications), keeping all data in memory. Apparently there is an investment bank that keeps all XML messages of dealing room activities of 1 day in memory and only persists the data at the end of the day.

I guess that these servers should somehow be setup redundantly, so that the crash or power failure of a server doesn't lead to a disaster (compare it to RAID-configured disks).

The last day was one without interruptions: all talks were held the one after the other. For the second talk I didn't even have to change rooms. This talk was about XSLT and more specifically XSLT 2.0 and Java.

Per default the Java runtime comes with a Xalan XSLT-processor. However, this processor only supports XSLT 1.0. If you want to make full usage of the new features of XSLT 2.0, you have to point the processor factory to a XSLT 2.0 aware processor, e.g. Saxon.

This presentation gave a nice overview of the XSLT 2.0 capabilities (like e.g. language specific collation), how to invoke them from within Java and how to extend XSLT with your own functions. The presentation also dealt with nice eye-candy features, like generating an SVG pie chart based numbers in an XML document and generating a PDF document using XSL-FO.

In a world driven by annotations, RESTful web services and stuff, it is good to see that XML isn't dead yet but that the focus is more on "data", rather than on configuration or RPC protocol.

This talk was about the way Scrum in combination with XP was introduced in the Belgian Postal services. It was an interesting story, touching upon all the growing pains of the rollout of an agile approach.

But the "enterprise" in this case is still located in 1 single building. This makes it still possible to colocate all team members, including the business. But things are all different once your company is divided over different buidlings, even over different city's and worse, different countries. The colocation becomes really hard... to say the least. This postal guy probably has no clue, but Erich Gamma, lead developer of Eclipse, knows all about it!

This was a recap of the presentation of last year, given by the same speaker. I've looked up my blog post of last year again and apparently there was not much difference in the talk of this year, compared with the talk of last year. The idea of profiles was already mentioned last year, but the only profile that is concrete today, is the web profile (this was already mentioned last year, but the building block of this profile are more or less fixed now). Things like EJB 3.1 light (only session beans) and the single class/no interface approach of EJB's were new.

The difference with last year? Last year Q3 2008 was meant as the milestone for completing the entire JEE6 spec; most sub-specs are in their final review stage, but the beta reference implementation was meant for Q4. That will be hard to meet...

After all the "theoretical" presentations about REST (now that I finally got the picture clear after the talk of Stefan Tilkov), it was time to see how this was all implemented in Java. Because to be honest, the first day's university talk "REST (in peace) with Java)" was NOT what I expected of a presentation on REST and Java; it even made it all more fuzzy to me. The JAX-RS talk however made it all bright and shiny again: it's all in the annotations :-)

All previous talks that mentioned REST talked about resources and URI's and also HTTP operations and statuses, but how on earth could you define an URI? This talk showed how it could be done using Jersey.

For me this was like the icing on the cake, after all the conceptual talks on REST. Not that I know how to do it myself, but now at least I know where to start looking for more information once I may need to do an implementation of a RESTful web service. This talk, preceeded by Stefan Tilkov's talk, would have been a great replacement for the REST in peace talk of the first day.

This talk focussed on the challenge of testing AJAX and/or web service based applications and the tools that can be used for it. The speaker started building his own testing tools after his request for ordering the Mercury toolset got denied. This resulted in the PushToTest toolset, which integrates with other tools, like Selenium, soapUI and the likes.

However, he didn't do a commercial talk, presenting his toolset. He focussed on the distinct test cases and how a unit test can be promoted to/reused as a functional test or even a stress and load test or a service availability test (for monitoring purposes), how htmlUnit can emulate the role of a user operated browser an can be launched in parallel (multiple threads) for stress testing purposes.

This was an interesting presentation. To me it was a teaser to take a closer look at the tools that were mentioned in this presentation, including the TestMaker bundle of PushToTest.

Links:

• PushToTest: http://www.pushtotest.com
• soapUI: http://www.soapui.org
• htmlUnit: http://htmlunit.sourceforge.net
• Selenium: http://selenium.seleniumhq.org

My first session of the day was the one of Ivar Jacobson on agile development: "be smart". I knew Ivar Jacobson as one of the three amigos: Jacobson, Grady Booch and James Rumbaugh are the founding fathers of the Unified Modeling Language. But I never attended a talk of him. This certainly was a great experience. This man - according to Wikipedia nearly 70 years old - can be concidered as an encyclopedia of information technology (he must have seen it all: from punch card systems onwards). So he knows very well what the pitfalls of a classical waterfall approach are, because he has experienced it all. And still - according to his presentation - about 95% of all projects around the world apply this classical waterfall approach: shoveling paper all the time (yes, our company still does so, too).

Content-wise this story has been told so many times already, but what is so great about Jacobson, is that he can tell his story in a fascinating way. He is really a gifted speaker, knowing how to hold the attention of his audience. Just as an illustration: the room was really crowdy and as far as I know, no-one left mid-session.

What to remember

He had this statement about testers, that testers were (and in many places still are) concidered as the ones that have to do the cleansweeping. The developers, the 'thinkers', like he called them, write code, with tiny little bugs in it. And the tester have to track them down. He compared this with a relationship, where the man - the thinker - has finished his meal and leaves the table, but then there is the wife how signals the man that he has to put his plate in the dish washer himself. This is the same with software development: the developers should take care of their waste first. We are all responsible for the quality of what we produce and cleanup our mess.  in other words: collective ownership.

No beat boxer during today's keynotes, not even a family picture. Today's keynote session started with a keynote of Josh Bloch on Effective Java, which was mainly based on the new book he wrote on the topic. Even though I hadn't been drinking Leffe, Orval, Duvel, Westmalle or what other Belgian beers he mentioned, this was way too much code than my brain could deal with at that time of the day. Even though I usually start working around 9 o'clock, I'm not your average Java developer (I type a lot of words every day, but in a word processor instead of an IDE).  Sorry, Josh, not that your talk didn't interest me, but for me this talk was the excellent opportunity to recap on my blog. Knowing the difficulties I had the previous days, there were still a number of talks I had to discuss. So during this first keynote, I wrote 3 new blog posts. Me very happy .

The next keynote was about Java 7 and modularizations. By then I had written the blog posts I wanted to write and besides that, my battery power had reduced so much, that I had to stop blogging or my laptop would have forced me to. Anyway, the second keynote was about Jiggsaw, the project dealing with modularizing the JDK. This reminded me of a talk on the last day of last year's Javapolis, on OSGi, by a guy from IBM. He was very annoyed that all their effort in trying to standardize their modularization solution bumped against a brick wall. I remember him saying: "In the Java Community Process, all participants are equal, but unfortunately some are more equal than others". Anyway, it is a good thing that they will at least try to integrate with OSGi, which is already very well adopted on the server side, in e.g. Websphere AppServer - off course - and also in the SpringServer. But, as Mark showed in his HelloWorld example, the modularization shouldn't be uniquely focusing on servers; even simple stand-alone applications should benefit from it (and especially JavaFX will benefit from it a lot). And besides, it is not only a matter of solving the JAR-hell; they will also deal with platform specific packaging (like generating DEB or RPM packages).

Finally he talked about what language features will be in JSE7 and what not. It wasn't new, there had been a serious - and sometimes emotional - debate going on about this topic, but the number 1 feature to be removed from the list was closures. Right before he moved on from Google to Mickeysoft, Neill Gafter had finalized his prototype. Unfortunately for him, his little baby won't see the light, at least not the "sunlight"...