The device fragmentation technical session turned out to be... well... interesting. The Sun and Orange representatives had a presentation prepared, but the presentation was quickly derailed by frequent developer comments. As there was a Q&A time reserved in the end, it would have been proper to let the presenters to go though the presentation without interruptions (even though I know this is an important topic for the developers). At least that way the audience would have been able to hear the whole presentation, just not two thirds of it. But what do I know...

Basically what the presentation was about how the developers should not try to create a different version of the app for each target handset. The development should not start with one device in mind only, but the developer should design the application in a way that it fits as many device models as possible. Nothing new here.

They also presented a number of detailed case studies of various application areas, like localization, handling key different key codes, handling different screen sizes, and HTTP 302 handling.

Localization can be done with JAD files, JAR resources and localization classes with own resource bundle implementation (suggested approach). Graphics can be fitted on different screens with tiling or using clipping methods from Graphics package. Key handling can be done with getGameAction() or with mapping keys to actions in JAD files. The most interesting part was HTTP 302 handling, but unfortunately I missed the URL they referred for more information as the time ran out. (Basically there are three different behaviors among the Java ME implementations. Either the implementation handles the redirection properly, the implementation returns 302 with address set to null and after the redirected address is received 200 is returned,  or then the redirect address is inside the response content due to bad parsing of HTTP headers.) 

A good point from one developer was that it could still cause more work, if one tries to have smaller number of builds than one for each device, as in that case one needs to keep track which version works with which devices and still the apps need to be tested with all the possible devices.

Or as one developer put it: "We developers bring the innovation. The title should be 'how the carriers work with the developers..."

Four carrier representatives (Orange, AT&T, Sprint, and China Mobile) sat in the panel in front of an audience of maybe 60-70 people strong (at 9:00 pm - nice feat...). Here are some discussion highlights.

Naturally the issue of carrier specific signing came up (although it was not the first question like I expected, but only second...). It was surprising to hear Sprint's candid response. Basically the Sprint rep stated that they spend 10 billion a year for network upgrades and device subsidies. To bring in money they need to sell value-add services. They need to make profit.

AT&T's response was along the security and privacy protection. Orange said that they go by the standards (they have no changes in the security domains).

The discussion then moved to how a hobbyist can get certified. Buying the Verisign certificate for hundreds of bucks per year is too much for a hobbyist. Orange said that they endorse Java Verified and they also have a no-cost Orange Endorsement. However they need to have the developers traceable - that is usually much easier for companies. AT&T needs a SLA (Service Level Agreement) to make sure the customers get the quality of service they are paying for - if you cannot scale to support many many customers, getting that SLA might be impossible. Sprint has a developer certificate (Orange btw promised to look in to that).

In response to what should an aspiring game developer to do to raise interest among the operators. The responses varied again. Orange rep said that we have already enough games, give use something else. Something innovative and new.

AT&T and Sprint mentioned that gaming is important, but that brand names sell. Good point. If you do not have name yourself or do not have a game for certain recognized brand, you might have a hard time selling your game to the carriers.