Just saying "Kaffe sucks" is not very enlightening. Last I heard it wasI've gone through the code and really Kaffe hasn't passed beyond a simple Java interpreter in the like 5 years (at least when I encountered it) of its development. Looking through the code I can't see how without some serious rearchitecting that it ever will. Honestly it would be easier to start from scratch to get to a modern JVM/JIT than to start with Kaffe.
able to run "real world" Java apps such as Eclipse. Not sure if any
of the F/OSS VM's are running JBoss yet or not, but I wouldn't be surprised. What doesn't Kaffe currently do that it needs to? Or
is the "suckage" just in terms of performance or what?
Its irrelevant. You can't do any of that chocoletty VM goodness. If I want a compile-time language, quite frankly it won't be Java.> GCJ is not a VM and Java as a langauge without the > runtime environment is not a very compelling story.
Ok, point taken that GCJ isn't a VM. But I've heard good reports about the ability to run "serious" java apps compiled with GCJ. Unfortunately I haven't made it a point to dig into the subject, so I can't quote specifics.
> And none of the > others actually are much of an effort.
The guys writing SableVM might disagree with that. It has an active, if small, community around it and has made great progress from what I've seen.
Actually sable looks pretty good compared to last time I looked. It still doesn't matter.
Interesting. Everything I've been hearing from the guys writing open source VM's indicates that the VM itself isn't the hardest part, and that the class library is the biggest stumbling block remaining. Luckily, it appears that most (all) of the F/OSS java projects have pretty much unified around GNU Classpath as the Class Library project, which simplifies things somewhat.
Again... its the TCK and sun's licensing and control over Java. An open source Java and open source TCK would allow the platform to evolve itself.
I disagree with your above characterization. Sable isn't near something I'd put in production (it claims to run poorly on SMP and has a "first implementation near completion" JIT). I think classpath is just plain tedious. Writing a JIT...thats actually technically difficult and to do so accross platforms is a large effort.
-andy
TTYL,
Phil
_______________________________________________ Juglist mailing list [email protected] http://trijug.org/mailman/listinfo/juglist_trijug.org
