2009-05-07T00:02:56 I can't imagine there being any difference in performance at the minute. But because Linux is Linux, I'm guessing the difference in performance will become greater (but still only slight) as Glendix gets ore complex. 2009-05-07T00:03:19 In any case though, Plan 9 programs are small enough that they wouldn't suffer. 2009-05-07T00:04:08 To give a comparison, there would almost certainly be less overhead than say WINE. 2009-05-07T00:04:41 I've never used WINE but I hear it isn't slow - I've heard Firefox in WINE runs faster than Linux native Firefox. 2009-05-07T00:05:21 Probably. I've tested Wine with some Windows applications and was really impressed about the speed. 2009-05-07T00:05:28 Even though my CPU only has 500 MHz. 2009-05-07T00:05:56 There is a project that aims to port Wine to the Linux kernel. 2009-05-07T00:06:23 WINE gives you fake Windows syscalls in user space, Glendix gives you plan9's small number of syscalls in kernel space, plus the filesystems that Plan 9 programs expect to find. 2009-05-07T00:06:33 Yep. 2009-05-07T00:06:37 I see no reason for Glendix to be slow. 2009-05-07T00:08:41 Glendix sounds really interesting to me. In the beginning I was sceptical about the binary compatibility but the current concepts in Linux don't seem to be well-thought like the permission issues you mentioned. 2009-05-07T00:09:17 What I like most is that it might be possible to mix ``Linux'' and Plan 9 applications. 2009-05-07T00:09:36 oh. The binary compatibility works as of now, it's just the system calls and synthetic filesystems that aren't there yet. 2009-05-07T00:09:45 Okay. 2009-05-07T00:10:37 Does there need to be a /completely/ new file system or will it be a wrapper for e.g. Ext4 with additional functionality? 2009-05-07T00:11:06 Plan 9 file systems have stuff that Unix does, like append-only mode. 2009-05-07T00:11:15 Hm. 2009-05-07T00:11:32 But really by 'Glendix needs to implement file systems' I meant the virtual filesystems that Plan 9 has. 2009-05-07T00:11:50 Like /net 2009-05-07T00:12:00 Or like #m for mouse stuff. 2009-05-07T00:12:10 Those things. 2009-05-07T00:12:25 Append-only mode means that files can contain the attribute to only allow appendings instead of file creation and deletion? 2009-05-07T00:12:47 yeah 2009-05-07T00:12:51 Ah, okay. 2009-05-07T00:13:16 Well, I like the idea of this file system thing but does it really scale well? 2009-05-07T00:13:18 Unix has something like it, but only for directories (the sticky bit). 2009-05-07T00:13:32 Append-mode for directories? 2009-05-07T00:13:35 Well, it scales across multiple computers is one thing. :-) 2009-05-07T00:13:36 *** TheAppleMan has joined #glendix 2009-05-07T00:13:51 Append-mode is something like echo test >> file, right? 2009-05-07T00:13:57 yes 2009-05-07T00:14:06 Ok. 2009-05-07T00:14:56 Yep, that was what really impressed me about Plan 9. E.g. the simplicity of using sound cards *remotely*. 2009-05-07T00:15:07 Anyway. Just speaking of stuff that Plan 9 filesystems have like append-only mode, the more I think about it the more I think that existing Linux filesystems can accomodate that without modification. 2009-05-07T00:15:18 It's just a flag in the st_mode. 2009-05-07T00:15:26 Ah. 2009-05-07T00:16:03 That would mean that Plan 9 programs can set the flag, and that other Plan 9 programs can make sense of it, but that Unix programs would see it and interpret it completely differently. 2009-05-07T00:16:21 * stu8ball checks stat(3) just to see what the mix-up would be exactly. 2009-05-07T00:16:34 (2) even 2009-05-07T00:17:01 What I disliked most about Linux wasn't the kernel itself -- it was the whole user space. Look at the GNU tools. All these applications are way too bloated. You cannot run a modern Linux desktop on a 500 MHz with decent performance nowadays. 2009-05-07T00:17:16 syslogd: look at the compiler... ugh. 2009-05-07T00:17:29 There are also too many abstraction layers in all applications. 2009-05-07T00:17:45 jrm8005: Yes, or glibc. 2009-05-07T00:17:54 oh. Nothing. The plan9-special mode flags are quite high values. 2009-05-07T00:18:03 syslogd: yeah, yet another 'ugh; 2009-05-07T00:18:11 stu8ball: ha, readline! 2009-05-07T00:18:20 Sure, there are alternatives but they only re-implement glibc in a lighter way but aren't any better because they are using the same crappy API. 2009-05-07T00:18:23 What about it? 2009-05-07T00:18:29 oh right, you meant syslogd 2009-05-07T00:18:39 whoops, yes, I did. 2009-05-07T00:18:44 :) 2009-05-07T00:19:02 window(1) is like having readline transparently in all programs. :-) 2009-05-07T00:19:49 * stu8ball is pretty convinced Plan 9 is the best OS for programs not having to deal with stuff that the OS should have done for them. 2009-05-07T00:21:00 I hope nobody will /kick me when saying that I don't like C. 2009-05-07T00:21:09 In my opinion C is too procedual. 2009-05-07T00:21:21 syslogd: right tool for right job. 2009-05-07T00:21:26 C++ is an advancement but too bloated. 2009-05-07T00:21:29 Yep. 2009-05-07T00:21:53 For hardware-close things I admit that C would be the best choice. 2009-05-07T00:22:20 But I'd always prefer a object-oriented language for user applications because of maintainability. 2009-05-07T00:24:18 I am planning to write a lightweight libcxx which is completely object-oriented. 2009-05-07T00:25:27 oh well, here we go again! 2009-05-07T00:25:44 I dont think you will be the first to try that 2009-05-07T00:25:44 Java has some nice OO APIs. 2009-05-07T00:26:05 the way you spoke about C, I would like to speak about Java 2009-05-07T00:26:08 I dont like it 2009-05-07T00:26:11 *** kristoffer has quit IRC 2009-05-07T00:26:14 Hehe. 2009-05-07T00:26:30 I don't like Java either but some ideas were really nice. 2009-05-07T00:27:01 If you don't use too many design patterns, your applications are still maintainable and you can benefit from them. 2009-05-07T00:27:03 I dont see a requirement of any sort of Java like concepts in my /net code 2009-05-07T00:27:10 I dont see how Java can help 2009-05-07T00:27:21 C's creators disliked it enough to try again with Alef, then again with Limbo. But C isn't bad. 2009-05-07T00:27:22 Nah, I don't want to use any OO for the kernel-space. 2009-05-07T00:27:30 if you talk about simple user-spacy things, I like Python's simplicity 2009-05-07T00:27:31 Just for the user application. 2009-05-07T00:27:32 +s 2009-05-07T00:27:56 Python is nice though I prefer a C-like syntax. 2009-05-07T00:28:44 I am not a fan of using indentions for scopes. 2009-05-07T00:29:50 that might be partly because your choice of text editor bungles things up for you? 2009-05-07T00:29:50 stu8ball: What are Alef and Limbo? Cannot find anything via Google. 2009-05-07T00:30:13 Alef is basically C with concurrency builtins. 2009-05-07T00:30:32 Limbo is much mroe different than C than that. 2009-05-07T00:30:51 I don't know either Limbo or Alef. 2009-05-07T00:31:15 Murmuria: I am using VIM so there should not be any problems. What I don't like is simply the concept of the (mis)usage of indentions in Python. 2009-05-07T00:31:27 Limbo's the language for the userspace in the Inferno OS, it's good for concurrency, it compiles to JVM-like bytecode 2009-05-07T00:31:36 Ok. 2009-05-07T00:31:47 syslogd, http://en.wikipedia.org/wiki/Limbo_(programming_language) 2009-05-07T00:31:48 I think I've read of Inferno OS on OSNews. 2009-05-07T00:31:49 It doesn't have true pointers is another thing I can recall abou tit. 2009-05-07T00:31:53 Thanks! 2009-05-07T00:32:05 Ok. 2009-05-07T00:32:17 Well, C pointers aren't "true pointers" either I suppose ... but meh, needless detail. :-P 2009-05-07T00:32:21 I can live without pointers. 2009-05-07T00:32:59 So, what's wrong with Limbo? What they were trying to achieve sounds interesting. 2009-05-07T00:33:15 Nothing's wrong with it as far as I know. 2009-05-07T00:33:28 I just never learned it is all. 2009-05-07T00:34:43 Hm, okay. 2009-05-07T00:34:53 syslogd, people who try Limbo once fall madly in love with it 2009-05-07T00:35:04 I am hoping to use it one day myself 2009-05-07T00:35:12 :) 2009-05-07T00:35:25 I wonder in what way it is being used currently... 2009-05-07T00:35:35 but it has many strong lobbiests 2009-05-07T00:35:43 Ok. 2009-05-07T00:35:59 The syntax is too Pascal-like for my taste. 2009-05-07T00:36:08 := 2009-05-07T00:37:00 :D 2009-05-07T00:38:13 Got to go. Thanks for the awesome talk. :) 2009-05-07T00:48:13 ah... question... 2009-05-07T00:48:43 After just recalling giving #m as an example, I'm left wondering how the #-filesystems will work in Glendix. 2009-05-07T00:48:51 VFS hack is it? 2009-05-07T00:49:53 I don't mean how will they be implemented, that's obviously libfs, I mean how can there be paths starting with '#' in Linux. 2009-05-07T01:11:37 *** TheAppleMan has quit IRC 2009-05-07T01:24:24 *** syslogd has quit IRC 2009-05-07T01:27:27 *** syslogd has joined #glendix 2009-05-07T03:22:32 stu8ball: we're going to have to do multiple things. 2009-05-07T03:23:20 some things will be done by having the kernel check and make the right files available, other things might use union mounting. 2009-05-07T03:51:23 *** jrm8005 has quit IRC 2009-05-07T04:01:16 *** jrm8005 has joined #glendix 2009-05-07T07:10:04 *** jrm8005 has quit IRC 2009-05-07T08:44:04 *** kristoffer has joined #glendix 2009-05-07T11:26:43 *** synchris_ has quit IRC 2009-05-07T11:29:08 *** synchris_ has joined #glendix 2009-05-07T12:51:47 *** KillerX has joined #glendix 2009-05-07T13:32:34 *** jyujin has joined #glendix 2009-05-07T14:10:35 *** synchris_ is now known as synchris 2009-05-07T14:53:51 *** dagle2 has joined #glendix 2009-05-07T15:05:36 *** dagle1 has quit IRC 2009-05-07T15:32:48 *** uriel has quit IRC 2009-05-07T15:32:52 *** uriel has joined #glendix 2009-05-07T16:00:36 *** KillerX has quit IRC 2009-05-07T17:03:55 *** dagle2 has quit IRC 2009-05-07T17:03:59 *** dagle2 has joined #glendix