2009-09-11T00:29:48 *** MichaelRaskin has left #glendix 2009-09-11T00:31:06 *** MichaelRaskin has joined #glendix 2009-09-11T04:08:09 *** FragmentedCurve has joined #glendix 2009-09-11T05:40:33 *** MichaelRaskin has left #glendix 2009-09-11T06:41:20 *** FragmentedCurve has quit IRC 2009-09-11T06:43:50 *** FragmentedCurve has joined #glendix 2009-09-11T07:21:50 *** MichaelRaskin has joined #glendix 2009-09-11T10:29:05 *** MichaelRaskin has left #glendix 2009-09-11T10:30:09 *** MichaelRaskin has joined #glendix 2009-09-11T10:54:02 *** FragmentedCurve has quit IRC 2009-09-11T10:55:10 *** FragmentedCurve has joined #glendix 2009-09-11T10:57:10 *** FragmentedCurve has quit IRC 2009-09-11T10:57:16 *** FragmentedCurve has joined #glendix 2009-09-11T10:58:43 *** FragmentedCurve has joined #glendix 2009-09-11T15:15:25 *** raph_ael has quit IRC 2009-09-11T16:05:47 *** KillerX has joined #glendix 2009-09-11T17:22:47 *** KillerX has quit IRC 2009-09-11T19:10:16 *** MichaelRaskin has quit IRC 2009-12-11T00:13:39 *** Murmuria1 has joined #glendix 2009-12-11T00:13:40 *** Murmuria has quit IRC 2009-12-11T00:20:28 *** solidsnack has joined #glendix 2009-12-11T00:22:51 *** Murmuria1 has quit IRC 2009-12-11T00:22:59 *** Murmuria has joined #glendix 2009-12-11T00:26:22 *** Murmuria has quit IRC 2009-12-11T00:26:27 *** Murmuria has joined #glendix 2009-12-11T00:29:04 *** Murmuria has quit IRC 2009-12-11T00:29:10 *** Murmuria has joined #glendix 2009-12-11T00:34:12 *** Murmuria has quit IRC 2009-12-11T01:09:03 *** Murmuria has joined #glendix 2009-12-11T01:39:36 *** solidsnack has quit IRC 2009-12-11T01:57:01 *** MichaelRaskin has left #glendix 2009-12-11T01:57:19 *** MichaelRaskin has joined #glendix 2009-12-11T03:11:25 *** KillerX has quit IRC 2009-12-11T04:03:33 *** solidsnack has joined #glendix 2009-12-11T05:28:19 *** anomit has quit IRC 2009-12-11T05:35:11 *** anomit has joined #glendix 2009-12-11T05:42:56 *** MichaelRaskin has quit IRC 2009-12-11T06:40:38 Murmuria, around? 2009-12-11T06:41:31 anomit, right here! 2009-12-11T06:41:45 :) 2009-12-11T06:41:59 *** MichaelRaskin has joined #glendix 2009-12-11T06:41:59 wassup? 2009-12-11T06:42:23 so J.R. Mauro thinks a FUSE filesystem for the cs is a bad idea 2009-12-11T06:42:55 yes... 2009-12-11T06:43:22 the argument seems valid... 2009-12-11T06:43:46 and there is no point of something temporary... we already have plan9port for that.. 2009-12-11T06:44:04 *** solidsnack has quit IRC 2009-12-11T06:44:11 right, I haven't been able to think of anything else though :( 2009-12-11T06:44:23 hmm, so, the point is, how else to do this in userspace? :-/ 2009-12-11T06:44:36 exactly 2009-12-11T06:44:55 my knowledge fails me 2009-12-11T06:45:17 well, it can be done... 2009-12-11T06:45:42 I think what we can do is have the files served from kernel space 2009-12-11T06:45:51 but send the DNS queries to userspace 2009-12-11T06:46:04 NFS does a similar mapping of userid to usernames 2009-12-11T06:46:27 it queries /etc/shadow for doing that conversion 2009-12-11T06:46:37 because on the network we use usernames 2009-12-11T06:46:47 *** solidsnack has joined #glendix 2009-12-11T06:46:55 and locally, the kernel only understands users as integer numbers 2009-12-11T06:47:08 the same goes for ip Vs urls now 2009-12-11T06:47:10 don't know much about NFS but yeah, read that in a book 2009-12-11T06:47:16 same here 2009-12-11T06:47:42 let me respond to his email so Anant or Mauro can shed more light 2009-12-11T06:48:03 I have been hoping Mauro more actively contributes soon though 2009-12-11T06:48:38 he has been around, and has good experience.. we roped him in from the Plan9 mailing lists... 2009-12-11T06:49:03 anyways... I am writing a response 2009-12-11T06:49:15 lets see if either of the two help you further 2009-12-11T06:57:41 What is the point of "Not everyone can run FUSE"? 2009-12-11T06:58:22 Technically, running FUSE is easy; administratively, I would be more cautious when enabling Glendix patches than when enabling FUSE. 2009-12-11T06:59:13 If admin is too careful, nobody forbids enabling FUSE only for root-owned filesystems - with user access enabled 2009-12-11T07:08:51 MichaelRaskin, fuse is a dirty way of implementing this 2009-12-11T07:09:04 Why? 2009-12-11T07:09:26 well, I guess Mauro should answer this... 2009-12-11T07:09:30 It is more or less designed for such tasks 2009-12-11T07:09:33 I am nto sure if I can be convincing 2009-12-11T07:10:43 I am thinking though, that we can continue to stay in kernel space for /net/cs 2009-12-11T07:11:41 Does Glendix aim at just being a one more Linux fork, like early Xen dom0? 2009-12-11T07:11:42 fuse will make things harder... for example... if you were to use URLs early during boot... 2009-12-11T07:12:21 Or does Glendix want to be "trivially available" in some non-Glendix-devoted distributions? 2009-12-11T07:12:57 Early during boot the DHCP has not yet run - usually 2009-12-11T07:13:26 And if you change that and configure network early, mounting a simple fusefs in early boot stages is not hard 2009-12-11T07:13:50 maybe... so you think fuse can be used for this ? 2009-12-11T07:14:40 I think yes 2009-12-11T07:15:15 You want binary-compatible Plan9 reimplementation that can be pushed into distributions 2009-12-11T07:15:53 The less you depend on kerenel code, the better chances you have for inclusion 2009-12-11T07:16:07 And never forget the power of symlinks 2009-12-11T07:16:20 well.. most of what we plan to do in kernel, we what to have it reflect upstream 2009-12-11T07:16:43 You mean - you hope it will be accepted? 2009-12-11T07:16:59 the /proc made it to the kernel eventually... 2009-12-11T07:17:08 Fortunately yes. 2009-12-11T07:17:09 someday maybe /net would be there too 2009-12-11T07:17:24 /proc uses in-kernel data structures. 2009-12-11T07:17:27 if we can show it works, it will get acepted 2009-12-11T07:17:32 It has to be in-kernel 2009-12-11T07:17:58 You need some propaganda to push sane per-process mount semantics into kernel 2009-12-11T07:18:02 Ashwin Ganti implemented the p9auth driver last year, which got into staging 2009-12-11T07:18:21 Accumulate goodwill 2009-12-11T07:18:37 well, those things can be thought over later... 2009-12-11T07:18:45 if and when we complete /net 2009-12-11T07:18:48 So you can say: we use FUSE and CUSE and symlinks to do what is sane to do out-of-kernel 2009-12-11T07:19:10 Out-of-kernel /net FUSE code can be reused by an in-kernel FS 2009-12-11T07:19:23 yes that we can... 2009-12-11T07:19:30 like I said, this is a clean slate 2009-12-11T07:19:31 It is simpler to do out-of-kernel 2009-12-11T07:20:01 no idea... I have not seen all this in code... thats why I was waiting for Anant or Mauro to elaborite 2009-12-11T07:20:42 MichaelRaskin, do you know how NFS works? 2009-12-11T07:20:58 Not much.. 2009-12-11T07:21:00 cos even they had to handle the problem of userid <--> username 2009-12-11T07:21:17 which is more or less the same domain as our problem here 2009-12-11T07:21:30 Well, they seem to have a userspace daemon 2009-12-11T07:21:38 ya... 2009-12-11T07:21:38 The point is - they are way older than FUSE 2009-12-11T07:21:50 They had to go the hard way 2009-12-11T07:22:32 Performance issues may lead you to writing in-kernel code, but then you will be able to show the numbers 2009-12-11T07:22:50 This issue is easy to measure, after all 2009-12-11T07:23:37 So if you have a FUSE implementation, an in-kernel implementation and benchmarks, you show - we have users, we have applications, we have benchmarks, please tell us what we should clean up in he patch. 2009-12-11T07:24:02 Otherwise you may get a "use FUSE" reply from the beginning. 2009-12-11T07:24:38 (In the first case, you just use FUSE version while patch is changing during inclusion) 2009-12-11T07:24:39 touche 2009-12-11T07:25:13 Look at Con Kolivas and his scheduler 2009-12-11T07:25:28 Ingo misses the point at point blank - repeatedly. 2009-12-11T07:25:55 anomit, so we might do this in different ways, but do you have a design in mind? 2009-12-11T07:26:20 why dont you proceed with your design, and come back with more detailed paperwork? 2009-12-11T07:26:52 guess I went too much "into the box" thinking about FUSE to come up with other approaches 2009-12-11T07:27:41 thats good, at least do what you can.. and give us a design... 2009-12-11T07:27:48 we can start building it your way, 2009-12-11T07:28:16 until such time as someone points us to a more beautiful solution 2009-12-11T07:28:36 okay, I'll post whatever I have done till this point 2009-12-11T07:28:56 presently we are only building castles in the air 2009-12-11T07:29:04 heh 2009-12-11T07:30:25 anomit, MichaelRaskin, I have been exploring the p9auth driver and Linux Capabilities last couple of months 2009-12-11T07:30:38 trying to see how they go together... 2009-12-11T07:30:48 seems like a neat solution 2009-12-11T07:31:40 either of you know much more in that area? 2009-12-11T07:32:08 Murmuria, not really, just know what p9auth is there for 2009-12-11T07:33:00 its a very interesting topic... I am going to write up a report on my findings so far... and then have Anant, Mauro and all of you read it... 2009-12-11T07:33:24 p9auth project will feel at home in the Glendix world... 2009-12-11T07:34:13 I see 2009-12-11T07:46:09 *** Murmuria has quit IRC 2009-12-11T07:46:22 *** Murmuria has joined #glendix 2009-12-11T08:28:26 *** Murmuria has quit IRC 2009-12-11T08:35:56 *** solidsnack has quit IRC 2009-12-11T10:00:21 *** solidsnack has joined #glendix 2009-12-11T10:44:00 *** solidsnack has quit IRC 2009-12-11T13:01:10 *** anomit has quit IRC 2009-12-11T15:52:37 *** Murmuria has joined #glendix 2009-12-11T16:13:22 *** MichaelRaskin has left #glendix 2009-12-11T16:13:41 *** MichaelRaskin has joined #glendix 2009-12-11T16:49:20 *** Murmuria has quit IRC 2009-12-11T17:19:22 *** Murmuria has joined #glendix 2009-12-11T18:16:00 *** KillerX has joined #glendix 2009-12-11T18:16:31 *** KillerX has quit IRC 2009-12-11T18:16:51 *** KillerX has joined #glendix 2009-12-11T19:16:56 *** MichaelRaskin has quit IRC 2009-12-11T20:48:11 *** KillerX has quit IRC 2009-12-11T20:52:11 *** KillerX has joined #glendix 2009-12-11T20:56:50 *** Murmuria has quit IRC 2009-12-11T21:24:06 *** solidsnack has joined #glendix 2009-12-11T21:35:00 *** MichaelRaskin has joined #glendix 2009-12-11T21:35:22 *** MichaelRaskin has left #glendix 2009-12-11T21:36:23 *** MichaelRaskin has joined #glendix 2009-12-11T22:08:57 *** KillerX has quit IRC 2009-12-11T22:33:07 *** solidsnack has quit IRC 2009-12-11T22:43:22 *** solidsnack has joined #glendix 2009-12-11T22:59:07 *** solidsnack has quit IRC 2009-12-11T23:33:30 *** tush726 has joined #glendix