James Gosling: on the Java Road

« Sleep is unlikely | Main | MyGame.java.sun.com... »

20070508 Tuesday May 08, 2007


Moore's law turns scale into evolution

One of the questions that has popped up about JavaFX Mobile is "does this replace midlets & xlets"? The answer is "no". It actually supports both midlets and xlets. Midlets were initially designed for the very constrained cellphones of 5 years ago. It has evolved since then and is the volume Java platform today. The next step up are xlets, which are starting to appear on CDC based phones. But the trajectory is clear: these APIs are approaching the full capability of a desktop: JavaSE. As the hardware constraints disappear, the API capabilities step up.

This is one of the great places to apply Moore's law: at any point in time you can look across the axis of scale and turn it on its side: today's high-end is tomorrow's volume. Permalink Comments [5]

Comments:

How long do you think it will be before average (or slightly better than average) phones are running JavaSE? Perhaps with the Java kernel project it might be sooner than generally expected.

Posted by Ricky Clarkson on May 09, 2007 at 09:57 AM PDT #

The real question is : will JavaFX mobile (read SavaJe OS) be GPLed or not ?

Because, to me, the same way that SavaJe failed to confront to MS on the highend mobile market, the same way Sun will undoublty fail.

MS got strong partnership with operators and manufacturers (I mean big bucks in the background) . Whatever price Sun will propose the JavaFX mobile license, MS will propose a better deal to them. Only one reason could make those "old-time" partners quit MS OS : open source ! If opensourced, SavaJe (oups, I meant JavaFX mobile ... did I said the new name is confusing, why didn't you call this "JavaOS mobile" ?) there would be a strong momentum brought to this OS including from operators looking for cost-cutting opportunities and customization opportunities. No doubt that a community would help Sun to port the OS to multiple platform

Opportunities for Sun is to sell customization (binary license) and service, plus high level software.

SavaJe OS was full JavaSE at some point of time and I got no doubt that JavaFX mobile is also close to this. James, going further to your point, if JAvaFX mobile is opensourced and become very popular, there is no doubt that JavaSE API wil win over more restricted APIs (read CLDC, MIDP, CDC, etc).

The choice for Sun is clear : 100% of a small market or some percent of a huge market ;-)

But is Sun ready to bring Java back to the client side and be an first-class actor there ?

Posted by bjb on May 09, 2007 at 12:14 PM PDT #

"f JAvaFX mobile is opensourced and become very popular, there is no doubt that JavaSE API wil win over more restricted APIs (read CLDC, MIDP, CDC, etc)." You MUST be kidding....I don't see Linux making any headway on the desktop no matter how "open source" it is. It takes more than just "open sourcing" to guarantee the success of a product. Sun is going about it the right way...by building on top of the success of MIDP, which has the momentum and past success that guarantees that the various stakeholders will pay attention.

Posted by asj on May 09, 2007 at 08:43 PM PDT #

"MS got strong partnership with operators and manufacturers" Actually, Microsoft doesn't hold a very strong position in the mobile phone market seen from a global perspective (in US is may be a bit different) and are easily dwarfed by companies like Nokia, Motorola and SonyEricsson etc. However, I agree with would that it’s very important than Sun commits the full JavaFX mobile stack to the open source community, which I understood from the key note that they intend to do (it would have had a bigger impact if Rich Green on stage had announced that the source code for both Java FX Script and JavaFX Mobile was available from today.. in a Steve Jobs kind of way ;-). Equally important is that Sun remains extremely committed to the future development of the platform. I have the impression that SavaJe’s work wasn’t really production ready and that there remains a bit of work before it is both complete and stable enough for manufacture’s to consider in a large scale. Not that JavaFX is a bad platform, but there is so many mobile platforms out there that a new one must be brought to them on a silver plate in order for it to be interesting. JavaFX has potential but it remains to see if that’s enough. I’m totally agrees with you on the branding issue. Sun have done it again! overloading a brand to the extent that it is going to confuse people for many years to come (I have already been explaining the difference between JavaFX Script and JavaFX Mobile several time, even to attendees here at JavaOne!). James, I agrees with you on the CLDC/MIDP thing. The whole CLDC/MIDP + optional packages now constitutes such a huge platform both in terms functional completeness and market acceptance that it is not likely to go away for foreseeable future. The richness of the CDC-stack/JavaFX Mobile is compelling for many people but remains suitable only for high-end devices like smart phones and even if Moore’s law also has its effect on mobile technology the low-end devices (where it real volume is) remains relatively constraint in order to keep the unit cost down and reduce power consumption. Another aspect is that the demand for resources always seems to exceed the availability, so that even if we gets more CPU power, memory and better display on the low-end devices the resources consumption of both the CLDC/MIDP stack as well as the CDC/FP/PBP (etc.) stack increases equally (the resource demands/availability remains to some extent constant over time) so instead of toss way the CLDC/MIDP platform it would be better to continue its development. I have actually for some time been advocating for something like the JavaFX Mobile platform just based on the CLDC/* stack instead of the CDC/* stack. However, that will among other things require a further development of the CLDC specification (haven't more or less been touched in 6 years), which Sun have been very reluctant to do. People have been advocating that CDC could be used instead but I believe that it contains so many elements that compared to CLDC are either redundant (e.g. the “old” file API vs. the GCF-based API) or “deprecated” (reintroducing e.g. finalizers would probably not be a good idea) that extending CLDC with the needed parts from CDC probably would be better than letting CDD supersede CLDC.

Posted by Jesper Zuschlag on May 10, 2007 at 06:29 AM PDT #

@asj: If Linux was not opensource, it would have died long time back, such as OS/2 or BeOS have died ! And it would never have been successfull on the server side. About MIDP, really, if you have tested SavaJe OS only once, you know that coding on plain old Java SE is way simple and more portable than having to cope with all the Java ME flavors (various magnetude of support of various tons of APIs, depending on manufacturer "good-will" and Java ME implementation support). MIDP is very strong at this time on the public market. But this is not draining a very important mass of IT services. Midlets are usually done by (relative small) product teams (inside or outside big telecom operators). I don't see MIDP or CDC beeing able to cope with a huge corporte application that is required by mobile users.

@Jesper : I agree, that Richard Green needs to "jobify" it's announcement, less marketing hype more "whaaooo effect" ;-) Ok, I'll try to clarify. MS position on corporate phone market (read "high end market") is quite high IMHO. Here in europe, most mobile corportate applications are implement know on a CE stack (phone or not). And I don't see how a MIDP or even JavaFX stack can compete with this. AFAIK, SavaJe OS was ready for primetime. I can not be sure for all the telephony part, but for the Java SE part, I have tested this years back when you could download it and set this up on an iPaq. At those time, my MS zealots-coworkers next to a demo asked me: "ey this is very fast CE app, how did you do that on this machine ?". I just answered : "I trashed the CE and put Java inside !". At that point, they were just stucked on the wall. That said, obviously SavaJe was not as bullet-proffed on all the topics (quid of long term memory management), but to me this was the core of a key element for the success of Java SE at mobile clent side. I must agree with you about the naming (why didn't they simply make it "JavaOS mobile" adding a sticker "JavaFX compliant" ?), again inconsistent marketing people stroke here. Anyway, let's wait and see if Sun go for OSS or not ...

Posted by bjb on May 13, 2007 at 02:47 AM PDT #

Post a Comment:

Comments are closed for this entry.