Draft 0.2 plan

I'm glad to see the targets set so clearly above. However I think there are at least two notable features missing from the list.

1. R/W vfat, fat32 filesystem support. This is the most used filesystem type on current desktops. It would be really important to be supported as soon as possible. Maybe we should post a TODO item for this.

2. While I was making charva work on JNode I found a big problem. At the moment only one charva instance can be run in Jnode at once or only one charva based application. It is so because charva relys on static variables and singletons while dealing with a console, and background threads. It is also hard to exit a charva application. I I tried a workaround but since this problem should be addressed on the kernel and VM level it doesn't make much sense further hacking it. This is only one little example of a complex problem.

JNode needs isolates. Most of the Java applications assume that they are alone in the JVM and they are written like that. Further when you call System.exit() the aplication should exit (all related threads die) and when you don't call system exit the JVM exits when the last non-demon thread terminates by itself.

The java shell command would need to start a seprate isolate where the static data in the classes is held separately per isolate and System.in, System.out, System.err, System.exit() and so on have a meaning similar to that in a JVM instance.

In case of JNode as an OS this is much more important than in the case of a JVM running as a separate OS process. So much that the lack of it makes development harder already.

WHEN will these issues be addressed? What are the current plans for it?

Levente

The ServiceManager is nearing an initial release.
The SM handles install and removal of services at runtime. Removal is synchronous an very clean. The ServiceManager also handles the issue of service versions.

Security was a design isue for the SM from the begining, so before I wrote it I took the liberty of designing a user security model. The primary classes in the model are User and UserManger.

UserManger acts something like a certificate athority. You give it a username and password and it gives you a certified User object.
User is final and not serializable. Any client that recieves a User object as credentials need not wory about verifying that such a User exists, because the UserManager has already verified it.

User contians name/id pairs for the user, primary group, and alternate groups.

This user model does not have any connections to java.security.

It may be that my user system is insufficient for use in Jnode, but it has several ineresting benifits that I think should be considered.

its a strange world to be starting an OS from scratch again.

What are the plans for multiuser support? How multitasking is it?

I think JNode is a chance to do much more than to reinvent linux in java. The Association for Computing Machinery (ACM.org) has a couple decades of some very good papers of Operating System topics that havent seen the light of day.

Id like to see JNode grow with a distributed operating system frame of mind. I would define that as datastructures and processes that can easily migrate from JVM to JVM and machine to machine. Java, I feel, leads the way in this area with is serializability and virtual machine.

Also a graphical environment that tries something new would make JNode stand out. Something completely different than the tired paradigm of simulating pieces of paper in a 1970's office (drawers, a desktop, etc).