**** BEGIN LOGGING AT Sun Nov 19 17:08:52 2000
17:08:52 --> RobertFit (bob@chi-tgn-guy-vty26.as.wcom.net) has joined #jos
17:09:30 --- RobertFit has changed the topic to: The JOS Project
17:09:35 --- RobertFit sets mode +t
17:09:37 --- RobertFit sets mode +n
17:29:05 --> jonh (~jonh@e-airlock224.esatclear.ie) has joined #jos
17:29:23 Hello Jonh.
17:29:33 hi
17:31:50 Are you interested in JOS?
17:32:23 yes i've bben following it for a bit
17:32:41 but not very closely
17:33:59 Have you tried compiling any of the source code?
17:34:08 i was online so i thought i'd drop in
17:34:24 no but it's on my to do list
17:35:01 Which part of Ireland are you from?
17:35:10 actually been spending time messing with atheos recently
17:35:25 i'm living in dublin
17:35:41 Cool me too.
17:35:59 Though i'm in Chicago at the moment.
17:37:07 I thought there would be more people then us two
17:37:45 Yeah, sometimes there are, not everybody comes at the same time.
17:38:22 is there much support, i vaguely read the mailing lists and haven't seen much trafic recently
17:40:17 I think there is a lot of interest, but very few people are willing to give up there time and help develope it.
17:41:23 how much code had already been writen ?
17:43:08 We currently have three jvm's and one kernel.
17:43:49 We are starting to move over to the teaseme jvm which uses the linux kernel.
17:47:05 so basically jos would boot the linux kernel before launching
17:47:41 Well, at the moment yes.
17:48:17 are we talking 2.2 or 2.4 ?
17:48:25 2.2
17:49:54 and the present kernel what is it based on ?
17:50:34 It's written in C++ and assembly.
17:51:38 so it was built for jos ?
17:52:12 Yes, it is called JJOS and it runs the decaf jvm.
17:53:21 is it still in alpha ?
17:54:00 Yes, pre-alpha would be better.
17:56:12 is the source for it tarballed or is it only on CVS
17:56:57 CVS is your best bet.
17:59:44 --> iain (~iain@ip170.san-antonio16.tx.pub-ip.psi.net) has joined #jos
17:59:54 Hello Iain.
17:59:59 hi
18:00:42 Hi
18:00:52 Finally made it to a jos meeting! :)
18:00:59 ;)
18:01:03 wb
18:02:19 --> stillzero (Stillzero@AC822F12.ipt.aol.com) has joined #jos
18:02:36 sup gusy
18:02:48 Hello stillzero
18:02:57 hi
18:03:19 whats the jos project?
18:03:44 We're trying to create a java based operating system.
18:04:11 --> madman1 (~madman1@194.165.170.241) has joined #jos
18:04:11 is anyone here running a server?
18:04:32 ook
18:04:37 hi all
18:04:40 and whats your end goal?
18:04:48 Hello madman1.
18:04:54 hi
18:04:57 <-- stillzero has quit (Leaving)
18:05:02 ls
18:05:10 wrong window.
18:05:14 hi madman
18:05:27 m\23\ireland
18:05:48 What's the story with all the Irish people today? ;)
18:05:53 hi iain
18:06:07 why who else
18:06:19 Jonh and myself.
18:06:42 egads i'm surrounded ... :)
18:06:48 small world and getting smaller i guess
18:07:08 where ye from
18:07:12 I think we ran out of drink.
18:07:36 <-- jonh has quit (Ping timeout for jonh[e-airlock224.esatclear.ie])
18:09:09 * RobertFit is working to create a demo bootdisk for teaseme.
18:11:51 <-- iain has quit (Ping timeout)
18:12:31 --> jonh (~jonh@e-airlock224.esatclear.ie) has joined #jos
18:13:43 Madman, i'm from Dublin.
18:14:24 But i'm currently working in Chicago.
18:17:23 --> iain (~iain@ip138.san-antonio16.tx.pub-ip.psi.net) has joined #jos
18:17:33 wb iain
18:17:45 aaargh. computer crashed.
18:18:55 Windows?
18:19:19 Yup. :)
18:19:34 i was going to ask the same question ...
18:20:43 So how's it going Iain? You been a little quite lately.
18:21:23 Yeah. Been out of town here and there and working to get something released this monday.
18:21:39 I know the feeling.
18:22:07 I have noticed a lot more kernel activity on the three candidates though.
18:23:09 Yeah, i'm still working on ujvm, but at the moment i'm creating a bootdisk for teaseme.
18:23:17 er, at least ujvm and teaseme. I think todd (jquinn) has been working on the new wiki.
18:23:40 Oh cool. I think a boot disk is essential so I'm glad to see that happening.
18:24:50 <-- jonh has quit ([x]chat)
18:24:51 Yeah, when I tried to get it running on my laptop I needed a lot of help from John (jewel), so I don't think everbody will be able to get the same level of support.
18:25:42 :) Yeah. Unfortunately os's require so much upfront before you can even see ANYTHING at all.
18:27:18 How does teasme look?
18:28:25 <-- madman1 (~madman1@194.165.170.241) has left #jos
18:29:25 It's pretty good, it's a full jvm and it didn't seem to be that hard to port to linux. Though the code really needs to be cleaned up.
18:29:50 is it C or assembly/c or C++?
18:31:01 C though I think it has some assembly as well.
18:31:41 what is it using for a class library... or does it have one?
18:31:55 Classpath.
18:32:01 cool.
18:33:05 what about ujvm? what's it written in and how far along is it?
18:35:44 It's written in C and i'll be adding some assembly aswell. Currently it can load a classfile i'm starting to work on executing bytecode now, i'm not sure if I should go with and interpreter or go stright to a jit compiler.
18:37:08 is it designed to be hosted on another kernel like linux or is it standalone?
18:38:31 At the moment it is just a native linux application but i'm going to port it to use the kernel interface, and then port the kernel interface to work with the linux kernel (maybe).
18:38:54 Cool.
18:39:04 I think you should go with an interpreter first.
18:39:23 Yeah, I was favoring that too.
18:39:32 baby steps. plus jits are notoriously complicated and messy.
18:42:13 With the previous version of my jvm it had a precompiler, that wasn't to bad to write it was tricky to write support for multipule java processes.
18:42:39 Yeah. That would be cool.
18:42:59 Actually, I was thinking that for an os, perhaps a multipass scheme would be best.
18:43:06 run straight interpreter on "new code".
18:43:09 watch it
18:43:25 precompile "hot spots"
18:43:30 watch it,
18:43:34 recompile hot spots again
18:43:36 etc.
18:44:17 Yeah, but hard to write.
18:44:20 :(
18:44:45 I've heard of people discussing something similar for enterprise jvm's for super optimizing ejb's etc over long periods and for particular loads.
18:44:49 Yeah. Very hard to wrte.
18:45:43 But that's why we're here right? The love of a good challenge! If we wanted easy, we'd be writing scrolling text applets! :)
18:45:43 --> jonh (~jonh@d-airlock123.esatclear.ie) has joined #jos
18:45:58 hi jonh again!
18:46:12 Hi
18:46:32 hi having trouble with my isp
18:46:58 Yeah, but it could be a realy cold scrolling text applet. ;)
18:47:04 Yeah. I've had the same. Sometimes they just don't want me logged in for more than 5 minutes it seems.
18:47:42 Robert: you betcha. scrolling text with its own custom tcpip stack. :)
18:48:32 :)
18:48:53 Isn't it funny how programmers can't help re-inventing the wheel?
18:49:11 I hate replicating work and yet here I am trying to build yet another operating system... :)
18:49:18 What it's been done already?
18:49:29 Well, os's that run java has.
18:49:39 :)
18:49:51 But that's not good enough for us. No, we have to reinvent one so it does it better. :)
18:51:16 I always say to myself what's wrong with reinventing the wheel
18:52:22 Well, for me personally, after reinventing the wheel several dozens of times, I've become tired of it.... just gets depressing.
18:52:40 But still, I don't learn.
18:52:47 yeah , but that life isn't it ...
18:53:24 Actually, I love starting to reinvent a wheel, its just 3-4 weeks of hard work into it that I start dispairing! :)
18:55:19 Is that when you realize that it's going to take a lot more effort then you thought it would ?
18:56:15 yes and no. I do embedded development. And that's when I usually say, dang, we should have paid the $5k up front and the 25 cent royalties and just bought a tcpip stack. :)
18:57:04 Cost vs pain usually goes in favor of cost during proposal and design, and then shifts to pain when implementation is grinding away... :)
19:02:31 i'm off, bye
19:02:35 <-- jonh has quit ([x]chat)
19:10:16 --> Tubez (Xner@7dyn237.delft.casema.net) has joined #jos
19:10:23 Wow. People.
19:10:29 --- Tubez is now known as Ranieri
19:10:34 yes, people.
19:10:37 Hi.
19:10:39 Hello Ranieri.
19:10:44 We should consider registering with cservice
19:10:45 And were alive too.
19:10:51 Hi iain, hi robert!
19:10:57 huh?
19:11:08 Not bad :)
19:11:38 See, i missed the move to the sourcefourge mailing lists, so i though the whole thing had just died :)
19:12:20 yes. That was an unfortunate result of how things played out... *sigh*
19:12:49 Aww, well it happens. But i noticed a lot of things have been happening in the meantime, esp in the JVM arena.
19:13:23 yes. We have 3 going now.
19:13:47 Not bad .. how about the kernels?
19:14:24 just jjos and linux. and jjos is dormant so the linux avenue is really the active development branch.
19:15:29 Hmm. That might not be so bad considering that's sort of the way i was doing it all along :)
19:16:21 Do any of the JVM/classthings do 1.3 ? That would be sort of important for me
19:16:42 Well the only class library even mentioned is classpath and that's 1.1
19:16:59 the jvm's themselves really don't have to change much to support any 1.x version.
19:17:08 Hmm.
19:17:39 the class library issue will be a major problem though.
19:17:43 See my design all along has been to subclass SocketImplFactory and DatagramSocketImplFactory, btu the last one is only available in 1.3
19:17:51 It seems to me that classpath is pretty much dead.
19:18:09 That would be a bad thing. I do not presume we can just use sun's.
19:18:35 yeah. Well, if you need a specific 1.3 library feature, it can probably be added. :) We are after all probably going to have to write/rewrite much of it anyhow.
19:19:14 We can use sun's if we adopt the scsl. Many people seem opposed to that (that's my general impression anyhow).
19:19:19 Well, what i need should have been in 1.0 but noone needed it so noone ever noticed it was missign till 1.2 so that's why it went in 1.3 :)
19:19:47 I'm not really into licensing but scsl requires ownership is transferred to sun right?
19:19:47 :) Yeah, I know what you mean.
19:20:13 OTH Sun gets screwed both ways because people say, "this has to be in the librares" and the other group says, "the libraries are too big".
19:20:26 :)
19:20:47 yes and no. scsl say you have to share with sun and patches and bug fixes belong to sun.
19:20:53 and commercial distribution must be licensed with sun.
19:20:55 That's what you get when you propose a such a scheme where ppl can only rely on your code to do stuff :)
19:21:02 and shipping versions must pass sun compliance
19:21:34 Aha. What about separating JOS and the classpath? Is it possible licensing-wise?
19:21:55 Like JOS is a GPL product that relies on an SCSL product?
19:22:00 separating? They're not tied together.
19:22:23 THat's what i mean :)
19:22:44 Oh yes, you can do that. But its illegal to integrate gpl with scsl because gpl doesn't allow it.
19:23:20 Heh, good old GPL :)
19:23:42 Well, suppose i want to catch up the months i have missed as quickly as possible, what do you reccomend i check out?
19:24:24 I think most of it's list on the home page, also read the mailing list archive.
19:25:01 Yup. Those two will probably do it.
19:25:04 Ok, just for the record .. what "the" home page? jos.org? metamech? jos.sourceforge.net?
19:25:18 jos.org
19:25:38 Ok :)
19:26:08 I'm tryign to work through the archives right now. It's a lot of stuff but i will get to the end of it eventually :)
19:26:12 yes www.jos.org will always point to the current jos web home.
19:27:33 * Ranieri is off reading the home page :
19:43:40 --- Ranieri is now known as Tubez
19:46:00 Hmmmm, where can i find info on the linux kernel?
19:46:43 What type of info are you looking for?
19:47:22 Dunno, the code would be nice :)
19:47:32 Can't find it's wiki topic if any
19:47:38 http://www.kernel.org/
19:48:14 Oh, is it just THE linux kernel?
19:48:26 Yes.
19:48:49 I though people were building a java shell around the linux kernel or something.
19:49:15 Yes, we are but at the moment it just needs the standard kernel.
19:50:07 I'm working on a bootdisk at the moment which will allow people without access to linux to try teaseme.
19:50:20 Ok this is the deal: i got some broken and primitive version of a network stack in java that runs in linux, and i would like to get it to run in a more JOS-like environment so i can get busy with network driver and stuff. What should i do?
19:50:39 I was planning on using JJOS but i hear it's been put on ice.
19:50:58 Or are we goign to skip the whole silly kernel idea and just use linux?
19:51:05 And it's drivers?
19:51:23 Does this mean we are goign to use its network stack too?
19:51:34 No, I want to use our own kernel and drivers.
19:51:53 Linux is just a base to work off.
19:52:16 SO the most advanced codebase for an independent kernel is still jjos?
19:52:20 yes, and although jjos is dormant decaf is still under some development
19:53:03 Interesting.
19:53:35 Where are all the things (going to be) handled like multiple processes/fork/etc ? In the JVMs?
19:53:42 I think the goal is still that the eventual place we want to be is an ultra minimalistic kernel, with a jvm on top of that, and most of the os on top of that.
19:53:55 In a java os on top of a jvm
19:54:03 I could deal with that i think.
19:54:06 think of the pico kernel and the jvm as a processor.
19:54:18 so we need a java os on top of that "processor".
19:54:31 the rheise.os project is an attempt at an java os on top of any jvm.
19:54:48 Where do the network drivers belong? In the picokernel or in the OS?
19:55:27 In the os. The only trick is how to let the drivers efficiently drive the hardware through the os+jvm+picokernel.
19:55:34 Hmm, i should have a look at that.
19:56:07 Yes. We've bashed around the ideas of using some special api + real-time java spec to provide that interface for drivers (a jos driver spec).
19:56:08 iain: yeah, that's what has been bugging me too. I really need a way to access memory as java types or i spend too much time copying stuff around.
19:56:39 Yup. First read the real-time java spec (books or online from sun).
19:56:41 Parsing packet headers is hell.
19:56:58 Then your thoughts are going to be essential. How to do this in a clean way while still being efficient.
19:57:08 java.sun.com? Okie.
19:57:09 Yes, I know what you mean. :)
19:57:28 yup.
19:58:23 I think you know the problem. I'm workign from mostly C examples (like all of us i guess) and there they do (float)buffer[3]=0.34 or something and that just drives me nuts :)
20:00:07 brb
20:02:42 ok. Yes, I know exactly what you mean.
20:03:49 I'll brb, i'm switching back to my cabel modem. Don't talk for a minute.
20:03:53 I've been working on a little C debugger in java and casting data is a major pain (esp with its signed data and sign extended shifting).
20:03:59 ok
20:04:03 --- RobertFit gives channel operator status to iain
20:04:06 --- RobertFit gives channel operator status to Tubez
20:04:10 --- You have left channel #jos
**** ENDING LOGGING AT Sun Nov 19 20:04:10 2000
**** BEGIN LOGGING AT Sun Nov 19 20:05:22 2000
20:05:22 --> RobertFit (bob@c851848-b.mntp1.il.home.com) has joined #jos
20:05:22 --- Topic for #jos is The JOS Project
20:05:22 --- Topic for #jos set by RobertFit at Sun Nov 19 17:12:28 2000
20:05:29 What i'm doing now is reading the whole header as an imput stream of some sort and calling the appropriate methods for every value
20:05:31 --- Tubez gives channel operator status to RobertFit
20:05:50 Boy that spec is large/my modem is slow :)
20:06:04 yes at least to the first
20:06:06 :)
20:06:15 --- Tubez is now known as RanieriA
20:07:11 I think the best approach is the simplest (which it sounds like you're taking).
20:07:34 I'd do optimization once things are working reliably. Plus profilers often surprise you about where your bottlenecks are.
20:08:01 esp in java
20:08:25 BUT i requires copyign most of the fields for parsing incoming packets and then again for outgoing packets, not to mention the fact that you probably end up copyign the whole packet once for every layer (link/network/transport)
20:09:00 oww well, just make sure you got a fast processor to go along with that fast network adapter! :)
20:09:24 :) Some times that's all you can hope for.
20:10:04 Well, speed is not a concern yet. Designing a proper OO TCP finite state machine is :)
20:10:50 Yippie, the specs are in.
20:11:09 Yes. I think that will be a very interesting challenge.
20:54:08 I've finally got the Linux bootdisk to work, now to get teaseme working.
20:55:14 Cool. I think I may have to leave pretty soon.
20:55:27 Giggle. Good luck. Do all the classpath things fit on a floppy?
20:56:06 I don't think so, it's 1800k.
20:58:12 Then how are you going to do it? Just stick the vfat driver in the kernel and load it from a HD partition?
20:59:16 I'm going to see what I need from classpath and delete the rest. I've already got the vfat driver installed. ;)
21:01:04 giggle.
21:03:24 --> Tubez_ (Xner@7dyn208.delft.casema.net) has joined #jos
21:05:19 <-- RanieriA has quit (Ping timeout for RanieriA[7dyn237.delft.casema.net])
21:07:09 I gotta go guys. I'll see you later.
21:07:19 Bye Iain.
21:07:31 bye
21:07:33 --> popeye- (POUH_@Montreal-ppp93297.qc.sympatico.ca) has joined #jos
21:07:36 <-- iain has quit (Read error: Connection reset by peer)
21:08:03 salut ca va
21:08:15 Hello popeye.
21:08:38 salut
21:08:47 vs etes de ou
21:10:20 Sorry I don't speak French.
21:11:01 where you from
21:11:22 I'm from Ireland, but i'm currently in Chicago.
21:11:59 well i'm in montreal
21:12:52 I haven't been to Canada yet, but I hope i'll be able to go sometime.
21:14:02