When Sun finally DOES open source Java, it'll likely be to late. [LWN.net]

When Sun finally DOES open source Java, it'll likely be to late.

Posted May 14, 2004 12:34 UTC (Fri) by Duncan (guest, #6647)
In reply to: JBoss CEO Opposes Open-Source Java (CRN) by dang
Parent article: JBoss CEO Opposes Open-Source Java (CRN)

While I agree with you in the BK instance (tho I believe it will
eventually either go libre or sink into oblivion, Larry saw Linus' need
and directly met it where nobody else, libre or not, was doing so, and the
fact that he went proprietary to begin with was a choice I won't begrudge
him making, as it was his to make, since it was him that stepped up to
provide the solution), Java is an entirely different ballgame. I'm with
IBM there -- the time has clearly come to FLOSS it.

> [V]ery real concerns about keeping Java portable.

As mentioned recently in discussion following a C#/Mono story, Open Source
doesn't need to worry about "portable", because open source
is /by/ /definition/ portable. Even where it isn't directly portable, and
good code tends to be at least /almost/ directly portable, all it needs to
be ported is a bit of work removing say 32-bit specific assumptions when
porting to 64-bit (an area I know a bit about as I'm writing this on my
dual Opteron Linux hobbyist-desktop-workstation), and porting of any
dependencies. Ironically, Microsoft's entry, C#, as implemented in
open-source Mono, is at this point easier to use than Java for
open-sourcers. As the world turns toward open source, Java is in danger
of becoming irrelevent, because it refuses to do so as well, and the
biggest strength it claims, portability, open source already has.

> [W]here [is the win] for going Open Source[?]

Both demonstrating the above point and supplying the answer here, is my
very real experience. Closed source must be ported by the suppliers. If
they refuse to do so, users are out of luck. Open source can be ported
(or supported in other ways, when the supplying company loses interest or
just can't be persuaded to do so for a particular market) by anyone
needing to do so. Of course, this has been one of the enduringly
effective pro-open-source arguements for many years, but Java on AMD64
once again demonstrates it.

There IS a server edition Java VM available for AMD64. There is NO full
desktop edition Java for AMD64. Why? Both Blackdown and Sun itself
refuse to port it, saying there's not enough interest. The server edition
VM is great for long-running Java applications, where the native code is
compiled and remains in memory, because it does a lot of time-intensive
efficiency optimizations that make the code once compiled smaller and
faster. GREAT -- IF your application IS a long-running probably server
application. NOT SO GREAT, if it's a trivial web applet or other trivial
use of the technology, where compile/start-up time, what a desktop edition
is optimized for, matters far more than run-time efficiency, what the
server edition is optimized for.

If Java was open source, both editions would be freely available for new
platforms, as those using them could if necessary download the code and
compile it themselves, instead of them being ignored by the
proprietary-ware vendors, as is happening in the case of desktop AMD64
users. It's a VERY REAL issue, and a VERY REAL win for open source.

To bad Sun is being stubborn about it. By the time it decides to open
source it, Sun will have likely squandered away the opportunity they had.
IBM recognizes this, probably because of intensive open source involvement
of their own. Sun apparently doesn't.

Duncan

(Log in to post comments)

When Sun finally DOES open source Java, it'll likely be to late.

Posted May 20, 2004 21:43 UTC (Thu) by roelofs (subscriber, #2599) [Link]

> [V]ery real concerns about keeping Java portable.

As mentioned recently in discussion following a C#/Mono story, Open Source doesn't need to worry about "portable", because open source is /by/ /definition/ portable.

I may have misunderstood either what he said or what you wrote (or both!), but it appears to me that he's talking about Java-the-language and you're talking about Java-the-VM. Certainly an open-sourced JVM is fairly portable (although I think you understate it considerably by referring only to word-size issues--threading models have traditionally been a much bigger headache, IIRC), but if different flavors of that JVM start diverging, then how portable is the Java code that's supposed to run on top of it?

That said, I'm fully aware (as is, no doubt, Herr CEO) that there are already a number of JVMs, so it's hard to see where things would be massively worse if Sun's were open-sourced. That is, there's already divergence, and one could certainly argue that, to the extent the divergence stems from actual bugs, having more open source can only help (i.e., they'd get fixed more quickly by those who have been bitten by them and who therefore have a vested interest in doing so). So overall, I tend to side with you on the matter...

Greg