| 00:33.39 | *** join/#edev wing0_ (~trowa@187.39.184.132) |
| 01:18.22 | *** join/#edev wing0__ (~trowa@187.39.184.132) |
| 01:44.46 | *** join/#edev djerome (~djerome@ip24-251-138-141.ph.ph.cox.net) |
| 01:45.23 | *** join/#edev djerome_ (~djerome@ip24-251-138-141.ph.ph.cox.net) |
| 01:52.15 | *** join/#edev miltonm (~miltonm@32.97.110.51) |
| 01:54.01 | *** join/#edev djerome (~djerome@ip24-251-138-141.ph.ph.cox.net) |
| 02:07.01 | *** join/#edev wing0 (~trowa@187.39.184.132) |
| 03:05.17 | *** join/#edev melchior (~melchior@AStrasbourg-552-1-82-176.w92-141.abo.wanadoo.fr) |
| 04:18.23 | *** join/#edev djerome (~djerome@ip24-251-138-141.ph.ph.cox.net) |
| 05:39.25 | *** join/#edev miltonm (~miltonm@32.97.110.51) |
| 05:49.21 | *** join/#edev mbuf (~user@nat/redhat/x-objyfcqkjiwwswbt) |
| 05:59.09 | *** join/#edev gvandenbosch (~gvandenbo@86.81.67.223) |
| 06:22.55 | *** join/#edev djerome (~djerome@ip24-251-138-141.ph.ph.cox.net) |
| 06:38.03 | *** join/#edev landley (~landley@swsoft-msk-nat.sw.ru) |
| 06:38.03 | *** mode/#edev [+o landley] by ChanServ |
| 06:39.15 | *** join/#edev risca (~risca@130.236.250.39) |
| 07:42.09 | *** join/#edev wing0 (~trowa@187.39.184.132) |
| 08:19.12 | *** join/#edev florian_kc (~fuchs@port-217-146-132-69.static.qsc.de) |
| 08:19.18 | *** join/#edev florian_kc (~fuchs@Maemo/community/contributor/florian) |
| 09:36.24 | *** join/#edev wing0 (~trowa@187.39.184.132) |
| 09:48.34 | *** join/#edev CcSsNET (~user@c-98-216-138-179.hsd1.ma.comcast.net) |
| 11:02.21 | *** join/#edev chouimat|away (~mathieu@kde/developer/chouinard) |
| 11:47.42 | *** join/#edev gustavoz (~gustavoz@host93.190-31-128.telecom.net.ar) |
| 11:51.14 | landley | I really, really, really hate nfs. |
| 11:51.31 | landley | And I hate rpc.lockd, and rpc.statd... |
| 12:39.55 | dalias | rpc.* should DIE |
| 12:40.10 | dalias | they were traditionally the source of endless root vulns |
| 12:41.59 | *** join/#edev RushPL (~quassel@89-78-81-193.dynamic.chello.pl) |
| 12:42.26 | CcSsNET | lol |
| 12:43.52 | *** join/#edev n900evil (~Speedevil@tor/regular/SpeedEvil) |
| 12:44.31 | dalias | and nfs is...well, hopeless |
| 12:44.48 | dalias | sharing a filesystem in cleartext is a complete joke |
| 12:45.01 | dalias | sadly even MS technology is better.... |
| 12:52.50 | CcSsNET | dalias, im sure proper security hardening precautions on a filesystem, userbase, and kernel can secure that situation |
| 12:52.52 | CcSsNET | ;) |
| 12:53.16 | CcSsNET | but its still rather dumb in the first place |
| 13:05.52 | dalias | unencrypted, nfs is only useful on localhost, which is a pretty dumb use.. :) |
| 13:05.56 | dalias | you can tunnel it over ssh |
| 13:06.14 | dalias | but that's going to be fun when a momentary network outage causes the ssh connection to drop :) |
| 13:06.32 | dalias | personally i would never recommend nfs for any use |
| 13:06.45 | dalias | i would actually recommend samba over nfs, even for 100%-unix networks |
| 13:16.06 | *** join/#edev TemporalBeing (~bmeyer@12.52.185.66) |
| 13:19.40 | CcSsNET | i recommend sshfs |
| 13:19.44 | CcSsNET | nothing else |
| 13:21.52 | CcSsNET | although using a storage media and moving it by hand has its obvios benifits lol |
| 13:27.50 | dalias | lol |
| 13:28.16 | dalias | imagine you're running a corporate or university network where home dirs should be shared across thousands of machines |
| 13:28.50 | dalias | (i've seen this done with nfs, where stealing anyone's private data is as easy as plugging a device into a network jack in an empty classroom...) |
| 13:52.10 | CcSsNET | heh |
| 13:52.24 | CcSsNET | sshfs to the rescue |
| 13:55.35 | dalias | sloooowwww |
| 13:56.11 | dalias | do you really want your fileservers having thousands of highly stateful ssh connections over tcp? |
| 13:56.53 | dalias | also isn't sshfs some ugly FUSE hack? :) |
| 13:57.13 | dalias | and does it offer correct locking semantics? |
| 13:57.25 | CcSsNET | eh i dont use fuse |
| 13:57.27 | CcSsNET | so idk |
| 13:57.50 | CcSsNET | all them virtual userland filesystems i dont use |
| 13:58.02 | CcSsNET | gnome vfs is another? |
| 13:58.03 | CcSsNET | lol |
| 13:58.19 | dalias | sshfs looks like a toy for making sftp servers look like part of the local fs |
| 13:58.24 | *** join/#edev kergoth (~kergoth@ip24-251-170-95.ph.ph.cox.net) |
| 13:58.25 | dalias | not something you could actually use |
| 13:58.28 | CcSsNET | well yea |
| 13:58.31 | CcSsNET | its useable |
| 13:58.39 | dalias | i mean use as / or /home |
| 13:58.42 | CcSsNET | would require some scripting |
| 14:04.03 | *** join/#edev GPSFan (~kenm@64.92.145.112) |
| 14:04.03 | *** mode/#edev [+o GPSFan] by ChanServ |
| 14:12.25 | *** join/#edev Zygo (startkeylo@startkeylogger.hungrycats.org) |
| 14:13.31 | *** join/#edev risca (~risca@81-233-40-5-no18.tbcn.telia.com) |
| 14:16.59 | *** join/#edev n900evil (~Speedevil@tor/regular/SpeedEvil) |
| 14:22.09 | landley | dalias: sshfs is fuse, yes, but it's actually reasonably implemented. Yes it's on top of sftp. |
| 14:23.03 | landley | The remote filesystem with the brightest future is p9 but the only reasonable server I've seen for it is the virtfs in qemu, which hasn't been genericized and broken out yet. |
| 14:23.17 | landley | They're still working out what the Linux extensions should be. Nice design otherwise, though. |
| 14:24.04 | landley | (Plan 9 doesn't use uids, its usernames are actual names. Linux needs an integer, what /etc/passwd says isn't the kernel's problem. So it needs extensions to deal with things the way Linux sees them. Things like the suid bit that p9 didn't have.) |
| 14:24.16 | landley | NFS is the cobol of filesystems. |
| 14:30.12 | dalias | the client and server could just translate usernames and group names |
| 14:30.24 | dalias | and suid bit should be considered harmful anyway :) |
| 14:31.22 | landley | The server is in userspace but the client is in kernel. |
| 14:31.38 | landley | dalias: capabilities are worse than suid. |
| 14:31.43 | dalias | oh i agree |
| 14:31.47 | dalias | i think both are harmful |
| 14:31.53 | landley | Anybody who thinks capabilities improve security, I don't want them involved with any systems I'm dealing with. |
| 14:31.56 | dalias | there should simply be no way to elevate privilege |
| 14:32.01 | dalias | once you've given it up |
| 14:32.26 | landley | dalias: you are an idiot. |
| 14:32.43 | dalias | glad to see we get along as well as always :) |
| 14:32.46 | landley | (I apologize, I'm really jetlagged right now, but DUDE...) |
| 14:33.02 | landley | You know about the tradeoff between security and usability? |
| 14:33.17 | landley | Feel free to produce a system nobody but you personally will ever use. |
| 14:33.26 | landley | Hiya prpplague. |
| 14:33.31 | prpplague | landley: greetings |
| 14:33.37 | prpplague | landley: whats cookin? |
| 14:33.48 | *** join/#edev prpplague^2 (~prpplague@ppp-70-249-145-165.dsl.rcsntx.swbell.net) |
| 14:33.51 | dalias | most of the stuff done by elevating privileges could be done better, and more securely, with simple ipc and privileged processes |
| 14:34.09 | dalias | if you start a suid binary, there are all sorts of potentially-dangerous inputs you control |
| 14:34.17 | landley | prpplague: in Russia. About to fall asleep. Heading back to the hotel in half an hour, plane home leaves at noon tomorrow. |
| 14:34.25 | dalias | resource limits, environment, signal dispositions, ... |
| 14:34.37 | dalias | not to mention all kinds of linux extensions that could affect security |
| 14:34.42 | landley | dalias: you used "simple" and "ipc" in the same sentence. |
| 14:34.42 | prpplague | landley: ahh dandy! |
| 14:34.52 | landley | Have you ever actually deployed a system that's seen real-world use? |
| 14:34.55 | dalias | if you open a unix socket to a daemon, there's only _one_ channel of input |
| 14:35.26 | prpplague | landley: that question for me? |
| 14:35.33 | dalias | i think it was for me :) |
| 14:35.51 | landley | Because nobody's ever found a security flaw in a server you talk to through a TCP/IP socket. Even the ones that _aren't_ supposed to be performing system administrative tasks. |
| 14:36.28 | landley | prpplague: I'm arguing with dalias, who thinks that talking to servers running as root via an IPC mechanism is an improvement over suid binaries. |
| 14:36.34 | dalias | i didn't say a daemon will automatically be secure |
| 14:36.50 | dalias | i said it only has one channel of attack you have to consider |
| 14:36.53 | dalias | as opposed to many |
| 14:36.54 | landley | I'm pondering asking him if he's ever heard the phrase "denial of service attack", but honestly it's not a conversation worth having. |
| 14:37.06 | prpplague | landley: ahh |
| 14:37.35 | dalias | opening many unix sockets to a local daemon has no much attack potential than a fork bomb |
| 14:37.49 | dalias | s/no much/no more/ |
| 14:38.58 | dalias | if the daemon likes it could even kill-9 all processes by a user that floods it :) |
| 14:39.59 | landley | dalias: I've been working on linux containers (lxc.sf.net) since December. |
| 14:40.32 | landley | Even listing the various security implications of things you're handwaving past would take longer than I'm likel to be awake. |
| 14:42.18 | landley | To summarize: this new mechanism I love isn't widely used and people haven't published 8 gazillion failure modes for it found by real world testing yet. Therefore it must be more secure. |
| 14:42.42 | dalias | for now :) |
| 14:42.51 | landley | Bruce Schneier has a rant about amateur cryptographers. |
| 14:43.21 | dalias | btw how many suids/sgids are on a normal system, and how many are for actually-useful things rather than crap like helping other users stalk you by writing to utmp? |
| 14:44.31 | landley | Why do you ask questions you could find the answer to by yourself with the "find" command? |
| 14:44.33 | dalias | the only useful suid i can think of is chome (which creates sandboxed users) |
| 14:44.42 | landley | If you don't KNOW, and you're confidently designing an alternative... gaaah. |
| 14:44.48 | dalias | i wasn't asking you to answer it, just think about the answer ;) |
| 14:44.50 | landley | I'm too sleep-deprived for this. |
| 14:45.13 | dalias | *nod* fair enough |
| 14:45.15 | landley | dalias: Hi, I've been developing and deploying embedded systems professionally since 2001. |
| 14:45.33 | landley | I've _IMPLEMENTED_ the answer, repeatedly, multiple times. |
| 14:45.55 | dalias | yes, but not with tools that make it easy to reduce the number |
| 14:46.02 | dalias | (like openwall tcb passwords) |
| 14:46.03 | landley | Busybox? |
| 14:46.09 | landley | openwall is crap. |
| 14:46.16 | landley | It's security masturbation. |
| 14:46.33 | dalias | some good designs come out of it |
| 14:46.37 | dalias | tcb being one of the best |
| 14:47.41 | dalias | no special perms are needed to read or change your password. (if you want to enforce password strength, just enforce it at password-authentication time, i.e. refuse to match weak passwords even if the hash matches) |
| 14:47.44 | landley | Ah, "trusted computing". |
| 14:48.23 | landley | http://www.fool.com/portfolios/rulemaker/2000/rulemaker000127.htm |
| 14:48.26 | landley | They never learn... |
| 14:48.52 | dalias | huh? |
| 14:48.55 | landley | Eh, meant to link to http://www.fool.com/portfolios/rulemaker/1999/rulemaker990621.htm |
| 14:48.59 | landley | I mentioned the sleep deprivation. |
| 14:49.23 | landley | Same general idea behind "trusted computing". We're going to program this device such that when we physically hand it to you, we control what it does and you don't. |
| 14:49.36 | landley | Because nobody ever modchips an xbox or jailbreaks an iphone. |
| 14:50.03 | dalias | dunno how we got into this dicussion |
| 14:50.26 | landley | Sleep deprivation and caffeine. |
| 14:50.51 | landley | Plus it's this or star at the rpc.lockd and rpc.statd code some more, which are just cthulian abominations. |
| 14:50.58 | dalias | maybe you looked up tcb and found some unrelated definition related to "trusted computing" ? |
| 14:51.12 | landley | I've come to the conclusion the easy way to make them work in container network context is to RIP THEM OUT AND WRITE NEW ONES IN USERSPACE. |
| 14:51.27 | landley | http://en.wikipedia.org/wiki/Trusted_computing_base |
| 14:51.47 | chouimat|away | landley: I now believe that trusted computing is an oxymoron |
| 14:51.52 | dalias | the relationship to the malicious use of the phrase "trusted computing" is minimal |
| 14:52.03 | landley | I've _always_ reduced the amount of crap installed on systems I was securing. Since the 90's. |
| 14:52.13 | landley | Removing everything the system shouldn't be doing is an important part of security. |
| 14:52.36 | dalias | *nod* |
| 14:52.38 | landley | The reason busybox ran the display at Cheyene mountain (before the shut down norad) was because they could visually audit the entire codebase, which they couldn't do with the gnu tools. |
| 14:52.54 | landley | One of the more weird emails I got as project maintainer, that. |
| 14:53.05 | dalias | btw what happened to toybox? :( |
| 14:53.16 | dalias | i was looking forward to it |
| 14:53.20 | *** join/#edev chouimat (~mathieu@kde/developer/chouinard) |
| 14:53.22 | landley | I got fan mail about it yesterday, which I found weird. |
| 14:53.32 | landley | I believe I blogged about it, hang on... |
| 14:53.45 | dalias | busybox is actually pretty crappy, but still better than the alternatives :) |
| 14:54.05 | *** join/#edev chouimat (~mathieu@CPE002129b5a060-CM0011e6c40c15.cpe.net.cable.rogers.com) |
| 14:54.13 | *** join/#edev chouimat (~mathieu@kde/developer/chouinard) |
| 14:56.08 | landley | dalias: I stopped working on toybox due to http://lists.busybox.net/pipermail/busybox/2010-March/071783.html and yesterday I blogged about reconsidering it at http://landley.net/notes.html#31-03-2011 |
| 14:56.39 | landley | Busybox is getting crappier. |
| 14:56.52 | landley | Not working on toybox hasn't made me more inclined to work on busybox. My enthusiasm for that project continues to go downhill over time. |
| 14:57.17 | chouimat | hehe |
| 14:59.05 | dalias | my feeling is that busybox is too cluttered with complexity by ugly micro-optimizations, an abundance of build options, etc. |
| 14:59.29 | landley | I blogged about the problem. |
| 14:59.39 | landley | My priorities were: 1) simple, 2) small, 3) fast. In that order. |
| 14:59.43 | dalias | rather than just letting almost-optimal size and performance come naturally from simplicity |
| 14:59.51 | landley | Deny's prioroties are 1) small, 2) fast, 3) simple. In that order. |
| 14:59.56 | dalias | :) |
| 15:00.49 | dalias | gnu's priorities are 1) gigantic nonstandard features, 2) lines of code (maximizing) so redhat will pay us more, 3) fast |
| 15:00.51 | landley | I've actually been pondering porting busybox applets to toybox commands. |
| 15:01.16 | landley | Leverage the busybox dev community that way, but do a scrub and disinfect of each command and a massive #ifdefectomy. |
| 15:01.50 | landley | gnu has the buffet mentality. "It's free, I must eat it all." |
| 15:01.52 | chouimat | landley: I think gnu stuff need one too |
| 15:02.08 | dalias | one thing you'll probably hate, but that i'd really like to see, is a clear distinction between the "sus tools" and "linux admin tools" sections |
| 15:02.15 | landley | The gnu stuff needs the old yeller treatment. Take it out back and shoot it, then start over. |
| 15:02.29 | dalias | where the former are written to be completely portable to any moderately posix-conformant system |
| 15:02.38 | dalias | and the latter of course require linux or a close cousin |
| 15:02.43 | landley | I have no interest in bsd, and macosx can pay for its own software. |
| 15:03.28 | landley | Posix is nice because it's what Linux implements. It's good documentation. |
| 15:03.31 | chouimat | landley: my experience with embedded freebsd is the source code is a lot cleaner that some of the linux stuff |
| 15:03.56 | dalias | indeed |
| 15:04.08 | dalias | i'm not a bsd person at all |
| 15:04.14 | landley | Windows claiming to be posix is moderately adorable in a chibi-cthulu sort of way, but you do not pick up the cute wild animal because it will bit you and giv eyou rabies. |
| 15:04.18 | dalias | but i have read source from time to time, and it's a hundred times cleaner |
| 15:04.30 | *** join/#edev sndcrb (~sndcrb@dsl-hkibrasgw2-fef8de00-126.dhcp.inet.fi) |
| 15:04.34 | dalias | windows claiming posix is just a lie |
| 15:04.34 | landley | chouimat: if I was going to mess with stuff that nobody will ever use, I'd play with Plan 9. |
| 15:04.40 | landley | But years on OS/2 pretty much got that out of my system. |
| 15:04.46 | chouimat | lol |
| 15:04.57 | landley | And yet the posix guys believe them and let them participate in the standards committee. |
| 15:04.58 | chouimat | landley: I still have a box running os/2 here |
| 15:05.19 | landley | I installed it under qemu circa 2006. |
| 15:05.24 | dalias | i don't think there's been a certified version of windows since some ancient NT, has there? |
| 15:05.35 | landley | I worked for IBM on OS/2 for the powerpc and Warp 4. That was my first job out of college. |
| 15:05.40 | chouimat | I just got ecomstation 2.0 running in virtualbox |
| 15:05.45 | landley | Wound up on Linux from OS/2. |
| 15:05.47 | dalias | and even then the certification was probably BS (it might be so old that the important stuff like sockets was optional back then... ;-) |
| 15:05.51 | *** join/#edev kosiini (~janne@pdpc/supporter/student/kosiini) |
| 15:06.06 | landley | Anyway, I gotta go catch my ride back ot the hotel. Might be on from there, or might fall over with The Tired. |
| 15:06.32 | chouimat | :) |
| 15:06.57 | landley | If you guys actually do have an interest in toybox restarting, email me. Maybe I'll set up a new list for it. |
| 15:07.02 | landley | meanwhile, gotta go... |
| 15:07.15 | dalias | i have an interest |
| 15:07.34 | dalias | if you have an interest in fixing things if/when it breaks with non-glibc/non-uclibc |
| 15:33.05 | *** join/#edev miltonm (~miltonm@32.97.110.53) |
| 15:37.56 | a7r | hey |
| 15:42.42 | *** join/#edev orbarron (~orb@nat/ti/x-llyakbkimbuqnolz) |
| 15:55.24 | *** join/#edev landley (~landley@93.188.120.203) |
| 15:55.24 | *** mode/#edev [+o landley] by ChanServ |
| 16:13.26 | dalias | wb |
| 16:13.44 | dalias | you missed as you were leaving: |
| 16:15.13 | dalias | i have an interest in continued toybox development, if you have an interest in responding to glibc/uclibc dependencies that creep in |
| 16:43.02 | *** join/#edev landley (~landley@93.188.120.203) |
| 16:43.02 | *** mode/#edev [+o landley] by ChanServ |
| 17:03.36 | *** join/#edev dijenerate (~dijenerat@204.212.240.167) |
| 17:35.48 | *** join/#edev miltonm (~miltonm@32.97.110.53) |
| 18:58.00 | *** join/#edev miltonm (~miltonm@32.97.110.51) |
| 19:10.50 | *** join/#edev florian (~fuchs@fsr7-1.physik.uni-siegen.de) |
| 19:10.50 | *** join/#edev florian (~fuchs@Maemo/community/contributor/florian) |
| 19:40.32 | *** join/#edev Zygo (startkeylo@startkeylogger.hungrycats.org) |
| 19:40.52 | *** join/#edev JoeLlama (~snork@unaffiliated/joellama) |
| 20:04.34 | *** join/#edev dijenerate (~dijenerat@204.212.240.167) |
| 20:26.01 | *** join/#edev kergoth (~kergoth@ip24-251-170-95.ph.ph.cox.net) |
| 20:43.31 | *** part/#edev orbarron (~orb@nat/ti/x-llyakbkimbuqnolz) |
| 21:14.28 | *** join/#edev kosiini (~janne@pdpc/supporter/student/kosiini) |
| 22:27.25 | *** join/#edev roxfan (dunno@91.180.91.43) |
| 22:37.25 | *** part/#edev TemporalBeing (~bmeyer@12.52.185.66) |
| 22:42.40 | Mirell | Hmm. |
| 22:50.00 | Mirell | Heh. |
| 22:51.26 | Mirell | landley can be really nasty when he's sleep deprived. |
| 23:02.06 | dalias | :) |
| 23:53.49 | *** join/#edev NightMonkey (debian-tor@pdpc/supporter/professional/nightmonkey) |