> I'm sorry I didn't communicate this clearly. When I say "network computer", > I mean a particular scheme to make a very inexpensive Java-executing engine > with centralized administration for use in primary school. I'd be very cautious about promoting "Java". It's true, we're close to getting a free Java implementation that "sort of" works. The compatibility won't be very good though - most people will not be satisfied with it. Also, the proliferation of Java APIs means that the free software community is only likely to clone the core bits and a few subsets of the entire API. Overall, there aren't many people working on the free software Java stuff, so I don't expect the situation to change anytime soon. I'd leave it up to Corel to promote Java on Linux (on their StrongARM NC). They've got enough resources (and less constraining principles) to play the proprietary software licensing game. We don't. Long term, I personally feel that some of the 'true' free software language efforts will bear fruit, and Java isn't really going to matter all that much. Java is only one language out of many, and the Java JVM and the just-in-time compiling concept aren't the optimal way of obtaining fast machine code. (it's still useful for "applets" though, not much else) The portability of the bytecode is nice, but there is going to be less real need for it in the future, since most computer makers are moving to Intel's beat. ie. StrongArm -> owned by Intel; Alpha -> made by Intel, still owned by DEC, deprecated for future use; PA-risc -> succeeded by Merced. Really, only the PowerPC, Sparc, and MIPS chips seem like they're left - and they occupy such tiny segments of the market that they are almost irrelevant. They're important, since they provide competition, but I don't think they're going to nudge Intel off the store shelves anytime soon. The world is becoming a less heterogeneous place - so the overhead of JIT compilers seems unnecessary. Native code schemes that support multiple architectures seem like the way to go. The Debian packaging system is one such scheme, so is ActiveX (yuck). Tcl/Tk/Python/Perl is cross platform, and quite successful. And I'm really liking Guile-Gtk too (but it's too young to be cross-platform). Cygnus's GNU/Win32 stuff is cool, but immature too. On the other hand, I have built something very similar to a Network computer using Debian and Netscape 3.01 in a fvwm2 session with customized X resources so the only program available was Netscape - and the user could only use Netscape for web browsing and email. I set up the window manager so Netscape always appeared full-screen, and it was impossible for the user to resize it or move it around (thus lessening the confusion). I also used tkgoodstuff's windowlist feature to provide a nice little toolbar for switching windows, and an exit button. Overall, it was quite useable, and require very little training and zero maintenance. The user was very happy with it. I've still got the setup here somewhere. Cheers, - Jim
Attachment:
pgpu3cyG5zB41.pgp
Description: PGP signature