New languages in

By eweek  |  Posted 2004-07-19 Email Print this article Print

Suns future?"> I understand there could possibly be some new languages coming out of Sun. Is there anything you can say about that? Sun is a company with no secrets (laughter). We have all been really happy with people doing other languages on top of the Java VM [Virtual Machine]. The Java language itself is pretty good for a lot of things, but the real magic in the Java phenomenon really doesnt have anything to do with the language; its really about the VM and the properties of the VM. And there are a lot of people who have built languages that target the Java VM. We generally havent talked about them a whole lot, for no particularly good or bad reason. And there have been people inside Sun who have done other languages on top of the Java VM, like Ive done a couple.
Weve only got one group thats seriously doing some language stuff, and thats a part of project that targets HPCS, or High Performance Computing Systems. And we have this contract with DARPA [Defense Advanced Research Projects Agency] to do R&D into building high-end numerical computing hardware and software. So weve got some folks working on [it], and its not clear how exactly it will go, but theyre looking at programming language support for doing scientific computing. And thats a research project. Where it will go nobody knows. Its not a product. Theres no FCS [first customer ship] or alpha date. Its just a few guys with propeller heads trying to figure out which way is up.
Click here to read eWEEK Technology Editor Peter Coffees analysis of Javas new direction. Ive been spending a lot of time with folks who do scientific computing. I had been moderately heavily involved with that group until about 8 months ago. And then I got a real job, which was a mistake (laughter). But the world of people doing scientific computing is pretty tough right now, if only because theres really no economic model to support people doing things like software tools for them because there are so few of them. And yet theyre a population that does things that are really, really important, like predicting the weather.
Well, as a follow-on Id like to know what your reaction was when you heard Microsoft was looking at doing something with high-performance computing? Kind of like anything else from Microsoft. Its impossible to detect from their press releases what theyre doing. They issue press releases about pretty much absolutely everything. Its kind of hard to believe that theyre really serious about it. Click here to read more about Microsofts supercomputing plans. And one of the things that I think is one of the unsung Sun stories is that, depending on how you calculate it and who you ask, somewhere between 15 and 30 percent of Suns revenue is high-performance computing. Most of what we talk about is enterprise stuff, but we sell a lot of stuff to people who do enterprise computing. And one of the problems that people in the software world have had is that you have to have some kind of economic model to support it. If youre an independent software company and youre trying to do software for people who use scientific computing, theres only like 50 or 100 of them [organizations] in the world that do the really high-end stuff. So you take your up-front engineering and divide it by the size of the market and youve got a really big number for your price tag. And it just doesnt compute. There are almost no small companies out there, other than two guys-and-a-dog kind of outfits, that are doing anything for scientific computing. And yet at a company like Sun, where we get a lot of hardware revenue, the hardware doesnt sell unless theres software to enable it. And weve kind of been coasting in the HPCS world because most people in the HPCS world have been saying Fortran, Fortran, Fortran, so weve got crews of people that give them Fortran, Fortran, Fortran. But even in the die-hard Fortran crowd thats getting a little hard to live with. We had an interesting review with a bunch of diehard Fortran folks at Los Alamos [National Laboratory] and they had done some interesting side-by-side comparisons between Fortran and Java, and it was fascinating. Then DARPA has been kind of panicky about how this has been going because theyre really dependent on the existence of numerical computing tools. So they gave us, IBM and Cray big buckets of money that have been firewalled off to do HPCS stuff. IBM is just building an array of basically taking a bunch of their Linux boards, putting them in a big box and calling it a supercomputer … which doesnt really work. Were actually building machines, so … Next page: The dangers of aspect-oriented programming.


Submit a Comment

Loading Comments...

Manage your Newsletters: Login   Register My Newsletters

Rocket Fuel