**** BEGIN LOGGING AT Sat Dec 2 18:21:47 2000 18:21:47 --> RobertFit (bob@c851848-b.mntp1.il.home.com) has joined #jos 18:50:51 --> COUSE (PlUsH@ts1-21.f1931.quebectel.com) has joined #jos 18:50:58 hello 18:51:06 wat you name 18:51:28 fveu tu maite @ my 18:52:31 <-- COUSE (PlUsH@ts1-21.f1931.quebectel.com) has left #jos 20:08:16 --> bob44 (bob@87-67-dialup-mtl.axess.com) has joined #jos 20:17:59 Hello Bob. 20:19:58 --- RobertFit sets mode +t 20:20:01 --- RobertFit sets mode +n 20:20:19 --- RobertFit has changed the topic to: The JOS Project 20:24:53 <-- bob44 has quit () 00:46:43 --> Keonfor (optimusb@pm17d128.iae.nl) has joined #jos 00:46:56 Hello Keonfor. 00:47:23 Hello. 00:47:35 So, what happened? 00:48:26 I'm catching up on some email and waiting to see if anybody would join the channel. 00:48:58 Uhm... Isn't the meeting supposed to be over by now? 00:49:51 No it's not until tommorow, the 3rd of December. 00:50:14 1700 UTC 00:50:49 Ihoulé man, I'm doing too many things at once here! 00:51:18 I'm playing a memory game of Minesweeper and transposed thee meeting-date by 24 hours. 00:52:20 I've done things like that. ;) 00:53:25 Not that I have any right to speak; I'm not a part of the JOS project, just considering joining. 00:54:32 Everybody has a right to speak at the meetings, including non members. 00:55:32 Shoot, just blew myself up.... 00:55:55 So what would you like to work on? 00:56:10 How's the project going, anyway? I can't get a really good picture, what with the mailing lists being down. 00:57:03 We've moved to a new mailing list, it's been fairly busy. 00:58:15 Sorry for the delay (still teasing my memory here). 00:58:39 We currently have two supported JVM's, teaseme and ujvm, the other JVM decaf is not really being worked on at the moment. Also there is CJOS but I haven't looked at it yet. 00:58:51 I'm not sure. I am a CS student, so everything seems attractive at some level (professional handicap). 00:59:22 I've noticed the traffic from on the lists from the archive. Growing at an exponential rate, it seems... 00:59:45 It comes and goes. 01:01:23 Is developing a multi-user OS still a long-term goal (some of the web-pages seem rather old, so I thought I'd ask). 01:01:48 Excuse me, that was supposed to be a question mark. 01:03:12 Yes. 01:03:28 The wiki web is very out of date. 01:03:59 I had a hunch.... 01:04:48 Too bad about the multi-user thing. That doesn't seem like something you'd want to build into a design after the major work has been done already. 01:05:21 There is currently work going on to develope a new wiki web which will be used by JOS (sfwiki.sourceforge.net). 01:07:00 Not really there isn't even a single user in the design yet. We are still working on getting threads/processes working. 01:08:57 I see. I assume you've abandoned Sun's approach of a 1-program VM? 01:11:18 Yes. Getting many processes in one JVM, seems to be one of the biggest features that JOS will need to work. 01:14:08 I agree. Unless you envision rebuilding MS-DOS, you cannot have an OS-core that is limited to running one major program at a time. It is essential to multi-user systems, systems with background processing, anything that requires separate threads of control in separated memory areas. 01:17:59 However, I do wonder WHY you haven't included multi-user capabilities in the OS spec. If you are going to have full multi-processing in the OS anyway, why not extend that to multi-user environment management and be done with it? 01:20:31 And I just blew myself up again... :-( 01:21:04 We did talk about this before (years ago), our kernel will be a micro kernel (kernel + jvm). We will build multi-user support on top of this and keep the kernel small. 01:24:19 That's what I was thinking. I just meant that, if you have a kernel capable of managing multiple processes, then why not go directly for multi-user support? After all, the user environment is a process and multi-processing then removes the difference between single-user and multi-user (if you discard securty issues for the moment). 01:26:00 I see a process in JOS being very different to a Unix type process, it'll be very light weight more like a thread. 01:26:53 I don't think user support will make sense with this type of process. 01:29:56 Though maybe i'm wrong, I didn't study CS. ;) 01:30:48 Do you envision JOS being a multi-tasking OS? 01:32:33 Yes. 01:33:58 Do you envision JOS as a REAL multi-tasking OS, or do you feel JOS should be like Windows? 01:35:57 It should do real multi-tasking, give a time slot to each thread/process, use multiple processors, etc. We may be able to make it a real time OS. :) 01:39:06 I wouldn't know what sort of advantages that would bring. But if you are going to expend the effort to create an OS with REAL multitasking capabilities, I think it would be a waste not to include multi-user support. After all, with all the support in place for real multi-processing, you're halfway there already. 01:40:39 I'm not sure what's needed for a multi-user OS, can you explain what's needed? 01:41:29 I can try, but please allow for the fact that it's 2:45 AM over here. :-) 01:41:44 Sure no problem. 01:41:55 First of all, you need multiple user environments, with time-slots for each. 01:42:34 In principle that isn't so hard to do if you already envisioned an OS with a GUI in which a single user can run multiple sheels. 01:43:06 Second, you need a security system to keep people out of resources that aren't theirs. 01:43:21 (That's like Unix's access rights). 01:44:08 Uhmm.... Let me think here.... 01:46:02 You have to build the OS so that access to common devices can be shared reliably (modems and such) without jamming everything up. But you'd probably get most of that anyway (two Net-aware applications running at the same time with one user have the same requirements, etc). 01:46:35 Lessee, anything else? 01:47:31 No, nothing comes to mind offhand. Basically, you need reliable multi-processing and a security system for access rights management. 01:50:09 And I just blew myself up again. 01:51:37 Yes, you need something to keep users from blowing eachother up! :-) 01:51:55 Ok, as I see it you don't need the kernel to do any of this, you just need a root process (very like a kernel ;) which will manage all the resources. The kernel will provide the basic services, access to hardware, virtual memory, threading, but it doesn't have to know that Joe User and Jane User are using files of a server. 01:53:59 In JOS we will have a system process which control the OS's resources. 01:54:28 Sounds right. However, the kernel MUST know that process A018905qhpafd (= Joe at his terminal) is not allowed to access resources marked as the private property of some authentication entity. 01:56:05 Ah, you've got some implementation of the Banker's Algorithm up? Good. Good place to start. 01:56:31 The real question is. How do you do that in Java? We don't plan to support native code outside of the kernel and system process. 01:56:56 Sorry how do you access the private memory in Java. 01:58:48 Sorry, I don't follow what you mean. Do you mean "how do you access the private resources"? 02:01:06 Yes, other OS's have to protect them selfs from the code the user is using. With a Java OS only running Java the OS doesn't have to try and protect it's self as much, the Java language does it for you (i.e. no pointers, etc). 02:03:47 Ah, I see.... You don't want to integrate the concept of access rights into the kernel... 02:04:39 Which then leaves you with the problem of how you protect memory blocks from wrongful access by different users. At least, that IS what you mean, right? 02:07:12 It seems to me that you can get a long way by making use of the memory protection you must already have for processes. After all, you don't want to have your wrdprocessor writing in you spreadsheet's memory, right? 02:07:32 The user first has to get access to the memory, the JVM doesn't allow you to do that. 02:07:54 Java doesn't allow that to happen. 02:08:38 Yes, exactly. So now I don't see the problem anymore. 02:09:53 Oh, wait. You mean "why do you need access right for resources if Java doesn't allow memory access"! 02:11:03 With "resources" I was thinking more along the lines of files and directories and stuff. Those would have to be protected from access by unauthorized figures. 02:11:55 Ok. The kernel doesn't have to do that, I would leave that up to the system process and class libraries. 02:11:59 Sorry, that's Dutch. From unauthorized people, I mean. 02:13:01 Indeed, you're right. The kernel wouldn't be responsible for everything, just parts of the security mechanism. 02:14:34 You could integrate that mechanism into the system process and class libraries. You could also think of including a security manager into the file system that matched access requests to access rights. 02:15:53 Again, I still don't see what the kernel has to do with this. 02:16:41 The filesystem is not part of the kernel it'll be in the system process. 02:18:16 The kernel is what makes it all possible. My entire point is that not including multi-user support right from the start would be a shame since you are already planning a kernel that has most of the basic ingredients aboard. 02:23:46 The OS will be multi-user but the kernel doesn't have to have any special support of this. A JOS process is not the same as a Unix process, more like a thread very light weight. JOS will have a system process which will control user access and resource allocation, even some of the drivers/services will be running as part of the system process. 02:28:02 Eh? I thought you said that multi-user support was still only a long-term goal. I know that the kernel doesn't require any special support for this other than support for multi-processing, which is why I wondered why multi-user support was only a long-term goal. 02:29:50 Let's just say everythings a long-term goal. ;) JOS is still a long way off from being a real OS, single or multi-user. :( 02:31:13 Ah, I see. As Pitr would say, "Is misunderstandink". 02:31:26 :) 02:33:06 Ok. Look, it's 3:36 AM over here, so I'm going to leave. I'll think about joining the project. 02:33:21 Ok, thanks for the chat. 02:34:44 Thank's for the info. CU, maybe. 02:34:48 <-- Keonfor (optimusb@pm17d128.iae.nl) has left #jos 05:06:45 --> Jasmin (lRcap69@200.32.116.211) has joined #jos 05:06:55 <-- Jasmin (lRcap69@200.32.116.211) has left #jos 09:55:45 --- You are now known as rf-away 10:35:43 --> popeye- (POUH@64.228.225.10) has joined #jos 10:47:28 <-- popeye- has quit (Passer 1 beau Nowell :)) 14:22:02 --- You are now known as RobertFit 14:35:00 --> dapustule (~risingbil@HSE-Toronto-ppp167893.sympatico.ca) has joined #jos 15:11:46 <-- dapustule has quit (Read error: Connection reset by peer) 16:19:36 --- Disconnected (). **** ENDING LOGGING AT Sun Dec 3 16:19:36 2000