On Saturday 18 October 2003 12:23, Valère Swinnen wrote: > > Let me know what you think. > > Don't hold back ... I didn't > > On the contrary I appreciate your honest opinion. > > Initially I was thinking about Java3D myself. Some time ago I played a > bit with it. > I certainly am not an expert on 3D stuff. To be honest I didn’t find it > particularly easy to work with but I think the reason for this was I > didn’t find a good “introduction to Java3D programming” on the net. > Secondly I had the impression that the development is not exactly a > priority for Sun so I was not sure if it was such a good idea to use it > for our project. I thought also that it might be hard to find > programmers with some experience in Java3D.
Combine that with: - a few weeks ago Sun announced to be supporting a OpenGL based Java3D replacement - Java3D implementations are known to be *very* buggy (which is probably the reason why there are so few people using it) > > I think that Java losing mindshare and going "out of fashion" is a more > > legitimate concern. > > What do you mean by this exactly? Are there any *signs* to support this? > I had the impression that Java is a settled *thing* in business > application development. > Are there any alternatives at the moment? Surely not C#? > > >>I have noticed a considerable improvement in speed compared to 1.4. > > > > It is good to hear that you think it is faster. Sun PR said that it was > > faster. The specific graphics operations I want are still too slow. (I > > said that the graphics operations are slow, not that Java is inherently > > slow.) > > I didn’t do any benchmarking on it but the software I am working on > showed a considerably improvement on start-up time which actually was a > pleasant surprise. > > > Now, for my random thoughts. In fact, my blunt advice: > > I will give you some more points to consider first. > > 1. The molecular modelling stuff will be done in stages. There are > different levels of modelling possible. Ab-initio is the best but also > the most CPU intensive. In fact you need a supercomputer or a cluster of > very big computers to handle at most 100 atoms or so but in the end you > are sure you know just about anything there is to know about your > molecule. There are other approaches to do modelling based on > semi-empirical and empirical algorithms, which are (a lot) less, > accurate but can handle from thousands until millions of atoms depending > on the algorithm. > We would like to incorporate some of the less CPU intensive algorithms > in our CAMD tool. VASP (or another) would be used separate from our tool > and is, as you said, for a multiprocessor environment. That is why I > mentioned that the CAMD tool needs to be able to show the result of the > simulation. That, but only that! Toda, I've been working on porting the VASP reader to CDK (cdk.sf.net) which is the chemical library the Jmol uses... Will upload this tomorrow. > In fact we would prefer that anybody with a reasonable recent computer > (2+ GHz) is able to use our CAMD tool. The same holds for the DC client. > > 2. [EMAIL PROTECTED] (the first Distributed Computing (DC) project) has been > developing a general purpose DC environment called BOINC. It is meant as > a platform for all DC projects. People will be able to switch between > projects without the need to load another DC screensaver. They will also > be able to participate at several DC projects at the same time. The > software is written in C++. > Although it is not used anywhere at the moment ([EMAIL PROTECTED] will start > with the distribution for their project soon) most people in our project > want to use it. I am almost the only one who is opposing it. Some are > doubting but tend to agree with the majority. > I am opposing it because I think most make the mistake in thinking that > “Joe average” is interested in participating in more than one DC > project. I do not think this is the case. No doubt there *are* people > that are interested in being able to switch but those are not the > *average* kind of people. > I think that *betting on one horse* is not the right way to proceed. > There are other projects interested in BOINC but they already have their > own DC environment so for them it is only an additional path to other > spare CPU cycles. (You can use both environments in parallel on the > server side). I haven't heard about a new DC project that *only* wants > to use BOINC (except ours that is). > > 3. From the start I was sure we had to add native libraries if we wanted > to use Java. (It is time consuming to develop molecular modelling > software even the ones using empirical algorithms). But I thought it > would not pose a problem using the JNI package. Do you have any > experience with this? Not directly... but have a look at JOElib (joelib.sf.net) which is a GPL-ed Java chemical library... they use JNI to link with the Ghemical libraries (in C++) to do basic quantum mechanics... > Actually I do not understand why most DC projects do *not* use Java. To > me it seems the most logical choice as Java (Sun) is mainly focused on > business application (database access and stuff) and I am sure Java is > optimised for this kind of applications. > > > I believe that C++ has an advantage over Java in the area of OpenGL > > support, but you are not currently thinking of using OpenGL. > > Actually I have been thinking about it. One of the reasons I thought > about Java3D in the first place. > Am I incorrect in thinking that Java3D uses OpenGL? Don't remember... > > I am not a fan of Java3D ... as a broad market product. Nevertheless it > > does work (I assume) and it does have its place. And I think its place > > is for applications like yours. Java3D will slowly gain broader support > > and acceptance over the coming years. > > Why didn’t you use it for Jmol? Was it a personal thing maybe? No, due to the many Java3D bugs in the implementation *and* that the plugin is not available for all browsers/platforms makes it very unportable... > Why do you think it *has* a future at all? Sun doesn’t seem to promote > it very much and they don’t seem to work on it (not much anyway). I have > been thinking about contacting the Java3D developers team about it but I > haven’t yet. Maybe I should! See above... look for the press announcement... I really don't think Java3D has any future... And seriously... I think Miguel's engine is more than what we could wish for from Java3D... Egon -- PhD Molecular Representation in Chemometrics Laboratory of Analytical Chemistry http://www-cac.sci.kun.nl/people/egonw.html ------------------------------------------------------- This SF.net email sponsored by: Enterprise Linux Forum Conference & Expo The Event For Linux Datacenter Solutions & Strategies in The Enterprise Linux in the Boardroom; in the Front Office; & in the Server Room http://www.enterpriselinuxforum.com _______________________________________________ Jmol-developers mailing list [EMAIL PROTECTED] https://lists.sourceforge.net/lists/listinfo/jmol-developers
