Home   Info   DevZone   Wiki  
UsersWeb  |  MainWeb  |  InfoWeb  |  DevZoneWeb  |  SupportWeb
[ HenriTorgemane ] [ not logged in ] [ Web: Imported ] goto:  options
[ get info on or edit ] [ login or new user ] [ list of topics, hubs & nodes, or recent changes ]

Henri Torgemane henri.torgemane@etu.utc.fr ( kind of ugly web page around http://wwwhp.utc.fr/~htorgema/ )

Computer (science|engineering) student somewhere in France..

JavaScript addict, java lover..

More about me later..

First, I'm still diggin the site to find the answer to my questions that aren't in the faq.. Roughly, - does it work on one machine, many or all? - will it be a mix of C and Java, pure Java, mix of asm and Java? - will it execute bytecodes and/or compile them? Are some insane hackers going to write a cross-platform JIT compiler for this purpose? - if not, how can Jos ever hope to be kind of fast? - Is jos going to be based entirely on sun implementation of java.*, or will we rewrite most of it? - The timeline seems to have a few projects started. Where are the interfaces these piece of code are going to follow? Do we code first, and then try to guess what kind of interface would still work with the existing code?

Anyway, if I can't find, say, 30% of these answers, in the next two days, then I'm probably going to spam a list of two in my futile attempt to find answers..

Hmm. Well we could actually use help to answer some of these questions. For some of them, I guess the best answer... we will know when we get there. As for what language it will be implemented in... most likely C and assembly although there is a project somwhere to write a Java to C translator. When we finally wressle loose Stephen's code we will have a clue what the pico kernel is written in. (although there is nothing saying that we have to use it if someone else writes a better one.).

To help answer some of your questions.. there were tons of discussion about many of these points in the mail archives...

Here is my take

It will initially work on a 486 processor... and any other processors people port it to.

Most likely a mix. As much in Java as possible... then C, then assembler when necessary. A Java to C translator of some sort might be used as well as the GNU C compiler.

insane hackers going to write a cross-platform JIT compiler for this purpose?

Hopefully we can write the VM so that this design decision can have more than one implementation. Kaffe does JIT and a few people have been thinking that we will port this. Although I see the classes being compiled into native code when they are "loaded" into the system... not necessarly ever time an object instance is made.

Hmm. There is a disucssion about JavaSoft's hot spot technology buried somewhere... I don't think that anything is a silver bullet. It will most likely be an effective mix of technologies that people are comfortable with.

java.*, or will we rewrite most of it?

We will have to rewrite all of it.... otherwise we cannot distribute the operating system. There are other organizations starting to do this... hopefully we will be able to leverage the best of their work.

Good question. There is no right answer or central athority here. I think it would be great to get a design repository going and a standard modeling tool... Playground seems good enough. Alghough there is often the risk of designing/talking too much and not getting anything done.. or producing an acedemic design that has no practial chance of being implemented. I have often done the latter so feel that digging into code for discovery purposes has merit.

Depends on which group you are working in and how those people work. Bill and I are trying to work on a VM design... others are porting Kaffee, others have already exsting shells (jsh) waiting to be bolted on and others are busily brainstorming. There is no right answer here.

If you are intrested in helping organize, we could always use the help..... Help with the FAQ?

- ClarkEvans



Content of these pages are owned and copyrighted by the poster.
Hosted by: