IRC Logs for #crux-devel Tuesday, 2008-07-08

*** pitillo has quit IRC01:44
*** sepen has joined #crux-devel01:50
*** DarkNekros has joined #crux-devel02:08
*** mike_k has joined #crux-devel02:21
*** Rotwang has joined #crux-devel02:30
*** Man0l0 has joined #crux-devel02:54
nipuLmust resist urge to burn danny on ml03:06
cptnnipuL: :-)03:11
cptnmaybe a esr/smart question link might be appropriate03:13
cptnbut I don't think it's worth the nerves03:13
*** Man0l0 has quit IRC03:42
*** pitillo has joined #crux-devel03:59
*** fetid has quit IRC04:46
*** Rotwang has quit IRC04:47
*** DarkNekros_ has joined #crux-devel06:05
*** DarkNekros has quit IRC06:06
nipuLswitch to cgit already, you can't quickly switch branches in gitweb06:19
*** Man0l0 has joined #crux-devel06:20
*** DarkNekros_ has quit IRC06:24
*** DarkNekros has joined #crux-devel07:07
*** fetid has joined #crux-devel07:08
*** Man0l0 has quit IRC07:11
*** Rotwang has joined #crux-devel07:27
*** Rotwang1 has joined #crux-devel08:34
*** Rotwang has quit IRC08:50
*** rantolo1 has joined #crux-devel09:32
*** fetid has quit IRC09:49
*** fetid has joined #crux-devel09:53
*** mike_k has quit IRC09:55
*** sepen has quit IRC10:30
tilmanopt/screen needs -j 1 for the install target for me10:42
*** rantolo1 has quit IRC10:52
*** tilman has quit IRC10:52
*** tilman has joined #crux-devel10:59
*** tilman_ has joined #crux-devel11:03
cptnI hope it's okay to start only at 20:00 CEST11:03
*** tilman has quit IRC11:04
*** tilman_ is now known as tilman11:04
*** Rotwang1 has quit IRC11:16
*** Rotwang has joined #crux-devel11:16
*** fetid has quit IRC12:11
aoncount me in, got home early enough after all :)12:21
cptnnice :-)12:21
tilmansomeone should add screen to opt-x86_6412:23
tilman(someone not me preferably :D)12:23
cptnjust send a mail to the crux64 list, stating "Hi this really needs to be fixed"12:54
aon :D12:55
aoni got a pretty nice monitor12:56
aon21" crt :)12:56
cptnsounds heavy :-)12:57
tilmancrt? :p12:57
aonyeah, i don't have the courage to invest in new technology12:57
cptnI'd like to get a 22" LCD12:58
cptnthey're so cheap nowadays...12:59
aonalso, large lcd's are widescreen12:59
aoni hate 16:9 quite passionately :)12:59
cptnI did so too12:59
tilmani'm happy with my two 17" LCDs12:59
tilmanif only the radeon driver knew how to cope with a large framebuffer better :|12:59
aoni've been hating it since my grandma's tv blew up and she got a 16:9 to replace it13:00
aonand it's been some years since that13:01
aonsince it's a finlux safety++ and finlux went bankrupt in 200513:02
*** thrice` has joined #crux-devel13:02
aon50% of my crt's come from the same factory now :)13:02
aonanyway, i'll go teach my mum to drive the lada13:03
aonshe really has too much energy13:03
aonbe back right soon13:03
Rotwanglada great car13:03
Rotwangalmost as good as wartburg13:03
tilmanmmh, bbiab too13:07
*** jue has joined #crux-devel13:16
cptnhey jue13:16
cptnI guess we're soon ready13:18
* tilman returns13:23
tilmanlet's start13:23
tilmanafter all it's called teh CASUAL meeting :p13:23
cptnheh, yeah13:25
tilman"Tools and methods for ports tree synchronisation"13:25
cptnnipuL added that one13:25
cptnhe wrote a merge tool IIRC13:26
tilmanoh, to sync trees for different architectures eg?13:26
juejust added one point to MeetingAgenda, this is related to "Maintainer line replacement script"13:27
aon hi again13:27
aonthat was scary :)13:28
cptnso I think we can wait with all 64-bit things again13:30
cptnsince nipuL will probably come later13:30
cptnfirst of all, is the new time fine for everyone?13:31
aonfor me it is13:31
jueyeah, bit I'll leave around 22:0013:31
cptnjue: sure13:32
tilmandid we discuss cptn's magic 'unmaintained' script yet, btw?13:33
cptni think there aren't any new bug reports we have to look at13:34
cptnwe didn't but can13:34
tilmanunfortunately i still didn't merge the extract-only thing13:34
tilmani'm working on pkgadd in pkgutils6 now, btw13:34
cptntilman: you said you improved the "unmaintained" script?13:35
tilmancptn: i was referring to bug #27113:36
cptnso did anyone try the "unmaintained" script, other than Simon?13:37
tilmanno, but the idea sounds good to me13:37
tilmanit should be made to work with utf8 input, and we should switch pkgfiles to use utf8 :p13:39
cptnforgive me my ignorance, but what's needed to switch to utf8?13:39
cptnI mean13:39
cptnother than recoding the existing files13:40
tilman:e Pkgfile  :set fileencoding=utf8  :wq13:40
juehmm, I'd prefer to deumlautify them13:40
tilmani'm fine with that, too13:41
cptnI have no strong preference myself13:41
*** sepen has joined #crux-devel13:41
cptnhi sepen13:42
cptntilman: so do I have to configure all editors to follow that, or can I do it via an environment variable?13:42
cptn(if you know)13:42
tilmandunno, i only use vim ;)13:43
jueIIRC it's only Simon and me who had to change something, s/ü/ue/ and s/ß/ss/13:43
tilmanand aon, but he got rid of the umlaut in his name years ago i think13:43
tilmanoh, and richard/namenlos13:44
aonalthough i was ignorant enough to remove only a bit over half of them13:44
aongot rid of the rest of them a while ago13:44
aonanyway, you can set the LC_* variables to say utf813:44
tilmanthat won't tell $EDITOR to use utf8 to write new files though13:44
aonit seems that it does13:46
aonor at least emacs switches to utf-8 if i export LC_CTYPE=en_US.UTF-813:46
tilmanemacs schmemacs ;)13:46
aonwell i don't know what your broken editors do ;)13:46
tilmanwe should probably ask the other umlaut-victims for objections13:46
aonvim too13:47
aonor at least it seems like it13:47
aonbut yes, i tend to prefer us-ascii over utf-8 for pkgfiles, too13:47
aonsince we probably don't have to write klingon into them13:48
cptnshould we then make this the rule for now13:48
cptnsince it's the de-facto standard anyway13:48
sepenand what about utf8 plains for future?13:49
jueI'd say they are still valid?13:50
cptnwhat's "utf8 plains"?13:51
tilmanplans, i think13:52
aonwell, it's a flat area consisting of utf-813:52
cptnaon: :-)13:52
cptnnext item; any preference?13:53
cptn# having "arch maintainer: " and "maintainer:" at the same time might be confusion to the user13:53
cptn# move "Main" pages to Wiki13:53
cptn# handling of outdated ports13:53
tilmanhandling of outdated ports is interesting :p13:55
tilmanfor 3.), i'd like to have some timeout after which people get active13:56
cptnare we talking days here?13:57
cptnor what would be a reasonable time?13:57
tilmansay if the maintainer is away/unreachable for 3 days -> go ahead and fix things yourself13:57
tilmanor maybe just 24 hours? ;)13:57
jue3 days13:58
aoni don't think you need to wait really long for changing source urls13:58
aonwithout changing the md5sum13:58
aonbut more significant changes should have some kind of peer review13:59
tilmanjue: 3 days can be a long time if some popular port is affected :D13:59
tilmanalso, 3.) doesn't occur that often, does it?14:00
cptnwell, it happened recently with imagemagick14:01
jueno, I dosn't, most time it's a temporary fault14:01
cptnbut it's rather the exception14:01
aonah right, the mail14:02
aonsry :)14:02
juebroken url's and secutityfixes should be taken serious, but at all we should not follow that update-paranoia some users have14:02
cptnI think we should just have a policy to deal with them14:03
aoni liked the idea of having some way for users to report outdated ports and have them automatically assigned to the right maintainer14:04
aonsince now it gets done rather seldom14:04
juethe system arch uses looks nice14:05
aonthey have something like that?14:05
jueyes, you can flag a port outdated14:06
jue.oO cptn found that14:06
cptnsorry guys, I'm quickly on the phone :-)14:07
tilmanwould we couple something like that with flyspray?14:08
tilmanor would we write a minimal cgi script that stores <form> data in a mysql db and sends email? :p14:09
aonhow about a minimal cgi script that files bugs in flyspray? :)14:09
nipuLok, i'm up. i hate you guys14:10
nipuLi'm not saying anything else unil i've had my coffee14:11
tilmanaon: yes, that's what i meant with 'couple blabla flyspray'14:13
tilmanaon: downside: user registration somehow required14:13
tilmanit would be neat if random people could just punch in their name and the outdated port's name and be done with it14:13
tilmanthough we could have a flyspray user ("outdated_port_guy") that's used for those ;)14:14
tilmanmorning nipuL :) :D14:14
cptnthe problem is that it's not trivial to add bugs14:14
cptnunless we write directly to the flyspray DB14:14
jueoh, seems like cptn is the flyspray guru14:15
aonhow about that jabber stuff?14:15
cptnaon: didn't find any docs on that14:15
cptnbut I didn't search for a long time14:16
tilmanaon:jabber as in instant messaging?14:16
aonyeah, apparently fs supports adding bugs via it14:16
tilmanoh, for flyspray you mean? :p14:16
tilmanif we put a simple html form on, random spam bots will probably trigger it, right?14:17
cptnbut we could use a cool captcha:14:17
tilmanwww sucks :p14:17
cptnthe port name :-)14:17
aonor "what does crux lack?: "14:18
cptnso if a port isn't found it's not entered14:18
tilmancptn: that should work indeed :)14:18
cptnfully back now :-)14:21
cptnif we can do it via jabber that would be best14:23
cptnI assume however that it might be a lot more work than a stand alone solution14:23
tilmandoes that mean we expect users to have a jabber client?14:23
cptnno, I'd expect we could do it from our cgi script14:24
cptndoes anyone know how our current portdb works?14:26
cptni.e. is it using a SQL DB?14:26
tilman$dsn = 'sqlite:////home/crux/public_html/local/portdb.db';14:27
tilmannot sure i'm looking at the right php script though :D14:27
* nipuL is human now14:27
cptnthen maybe we could integrate it in there14:27
cptnhey nipuL :-)14:27
nipuLit's not overly difficult to write a bot that has to login to flyspray14:27
nipuLi've done it for a number of projects14:28
cptnokay, sounds good14:28
juecptn: Simone wrote our current portdb14:28
tilmanjue: wasn't it tillb? :)14:28
tilmanfunction sanitize($str) { return $str; }14:28
cptnwhat about the "timeouts"?14:29
cptni.e. how long should we wait before we touch other's ports?14:29
nipuLi think a few days is fair14:29
nipuLaslong as the maintainer has been contacted14:30
cptnso would you agree that it's reasonable to look at them at every casual meeting?14:30
cptnI mean, we can always decide we don't want to touch it, or not yet14:30
tilmanlet's go with 3 days for source URL fixes and security stuff14:31
tilmanbut 2.) is entirely differnt imo14:31
tilmannipuL: btw, we're talking about
tilmancptn: when did we have situations like 2.)?14:31
tilman* aon being in the woods :p14:31
tilman* sip being awol for months14:31
nipuLtilman: yeah, i know <-- coffee ;)14:31
tilmanaway without leave14:32
nipuLi thought it was absent14:32
tilmanmmh, you're probably right14:33
tilmananyway, i think case 2.) should absolutely not be common14:33
cptnwell, the point is, that once a user reports it, he has some reason why he want the update14:33
cptni.e. pidgin14:33
cptnmaybe we could say two casual meetings?14:34
cptni.e. in the "worst" case, that's 8 days only14:34
aoni guess it would be good practice to make a formal notice like "i'm away for a year, please take care of my ports"14:34
tilmanyou mean situations where a port is crippled because of its not-uptodateness?14:34
cptnthat should be enough for the maintainer to mark the bugs as "I saw it and I'm working on it"14:35
tilman(eg pidgin 2.x not being able to connect to ICQ)14:35
nipuLpeople still use icq?14:35
tilmanwindows people do :|14:35
juewell, pidgin was a special case because we broke it with our nspr/nss updates14:36
nipuLwould it work if --disable-gnutls was taken out?14:36
tilmannipuL: no idea, but it's fixed now14:37
cptnso does that look right:14:37
cptn  - 3 days for security updates, critical bugs and broken urls14:37
cptn  - 2 casual meetings for the rest14:37
cptn  - if a bug is marked as being worked on, it should not be touched14:37
tilmancptn: users will also complain about trivial port updates...14:37
j^2$compllain && $complain && $complain14:38
cptnthey will ask for an update on the day of release14:38
nipuL i think it would be good to add a status indicating that on FS14:38
cptngood idea14:38
nipuLbut a page describing the process to uses would also be needed14:39
cptnmaybe the non-critical updates should even only be done if they are trivial14:39
tilmancptn: sounds okay to me. i don't get your math though -- 2 casual meetings == 8 days? :p14:39
cptn8 - 1414:39
nipuLie, 1. contaict maintainer, 2. add FS task14:39
cptnor 8-1314:39
tilmancptn: ah, i think you missed the -x part earlier :D14:39
cptnprobably :-)14:40
tilmanhaving a status for BEINGWORKEDON sounds good14:40
cptnokay, I'll add a note on improving the status14:40
tilmannipuL: yeah14:40
nipuLwork in progress14:40
tilmancptn: i'll add that14:41
cptnwhat about reviewing?14:41
cptnI think that URL updates do not need a review if (and only if) the md5sum is the same14:41
nipuLwhat ever feels right for you :)14:42
tilmancptn: what you described in that mail sounds good to me14:42
nipuLaslong as it's a broken url and not just a tempory outage14:42
tilmanie one or more ACKs on other updates14:42
cptnthat's good, I'll summarize this in the meeting notes14:43
tilmani added a 'work in progress' status for all flyspray projects14:43
cptnand we should probably write a doc on that for the wiki14:43
tilmanoh, we need to make sure not to mess up the footprint of other people's ports when updating them14:44
cptnwe'll se ehow it goes14:44
cptnmaybe we'll have to adjust the process later on14:44
tilmani'd be annoyed if someone did that to me :p14:44
cptnbut I think it's good to have a process to begin with :-)14:44
cptnwiki cleanup (hand-on) or x64?14:45
juewiki cleanup14:45
nipuL64 :)14:45
tilmansince nipul made it today, x64?14:45
tilmanmoving stuff from Main/Documentation to Wiki/Somewhere doesn't need much discussion ;)14:46
jueok, I'll start doing some work here the next days14:47
nipuLyeah, just someone to do the work14:47
tilmanjue: i'll try to do the same14:47
cptnI'd actually like to do this coordinated14:48
cptnoh well :-)14:48
nipuLso lets coordinate jue to do all the work14:48
cptnso nipuL, can you update us on the ports tree sync work you've done?14:49
nipuLall i have at the moment is a handfull of scripts14:49
nipuLi can upload them14:49
cptnregarding "# having "arch maintainer: " and "maintainer:" at the same time might be confusion to the user14:52
tilmancan't we just document it and be done with it?14:52
cptnI liked the idea of only having "maintainer" there14:52
tilmanwith what value? :p14:52
tilmangeneral maintainer or arch maintainer?14:52
cptnsince that's the one who should be contacted14:53
cptnif there's a need for it14:53
nipuLi agree, the user doesn'treally  need to know who the general maintainer is14:54
nipuLi think i suggested it in the ML, but seeing .arch is being ACK'd couldn't it be put there?14:55
cptnsounds good to me14:55
cptnit's only relevant to developers and tools14:56
nipuLwe don't want users complaining to the upstream maintainer about an architecture they don't use14:56
juemmh, in most cases I've been contacted because the user wants another option or something like that14:56
jueso that's a maintainer task14:56
cptnmmmh, good point14:57
jueI don't see a problem to list both of them14:58
cptnI'm just not sure if our users will understand14:58
tilmani liked how .arch only contained the architecture and nothing else:D14:58
nipuLsomething to add to a bug reporting wiki page14:58
juewhat is CRUX? for exp ..14:58
cptnokay okay :-)14:58
cptntilman: what are you using for now?14:59
cptn"maintainer" / "arch maintainer"?14:59
cptnor "Primary" / "arch"?14:59
nipuLi've also been using arch packager for the ports i've touched for multilib15:00
tilman"maintainer" / "arch maintainer"15:00
cptnokay, let's start with this15:01
cptntilman: did you write the git 'update' hook btw?15:02
cptnany idea who did? :-)15:02
tilmansimone did15:03
tilmani think i wrote post-update in core/xorg/tilman.git15:03
cptnnipuL: regarding flyspray / bug submission bot: do you have code you could reuse for that?15:03
cptntilman: okay15:03
nipuLpossibly, i'm not sure how reusable it is15:03
cptnone thing I'd really like it to close bugs via commit message15:03
tilmanbtw, maybe we should also enforce the move to the new git->rsync export thing that i wrote15:07
cptnfor contrib?15:08
tilmanit would fix the errors that romster nicely pointed out again15:08
tilmanfor all git repos15:08
cptnoh, I must of missed that one15:08
tilmancptn: see hooks/post-update in xorg.git15:08
tilmanunlike romster i don't think it's urgent15:08
cptnokay, I'll add that to the list15:08
tilmani've been the only one who has been complaining every now and then :p15:09
sepenabout contrib post hook?15:09
tilmansepen: no15:09
tilmanwhat about that one though? is it enabled already?15:09
cptnthey wanted to see it15:09
cptnthat's why I asked who wrote it15:09
sepenI mean about notify hook15:09
cptnI'll sak sip if he minds if I commit it to git15:10
sepenand some other if could be public15:10
tilmanokay, we're talking about the git -> mailing list hook now, right? :p15:10
sepenyes :)15:10
tilmanthose hooks aren't exactly rocket science15:10
tilmani don't get why it's so interesting15:10
cptnme neither15:10
tilmanand i didn't interpret mike's mail that way, anyway :p15:11
sepenok ok15:11
juebtw, my [notify] today dosn't work15:11
cptntilman: but I posted a link on update hooks in general15:11
cptnso if they want something special, I think they'll have the docs available15:11
tilmanwooo, lots of stuff in /var/log/mail15:12
cptnso maybe we can just put the notify hook in place for now?15:12
cptntilman: yeah :-/15:12
tilmancptn: if you're going to email sip, can you ask him about how to update, too?15:12
tilmancptn: i think so, yes15:12
cptnwill do15:12
cptnsepen: would that be okay?15:12
nipuLcptn: i hope you like poorly written python
tilmannipuL: one of us should write haskell bindings for libpkgutils when it's done :)15:13
cptnnipuL: heh, thanks15:13
tilman.oO(i can has comments?) :D15:13
nipuLtilman: why not just write pkgutils7 in haskell :)15:14
tilmanjue: which [notify] do you mean?15:14
juemy last commit, gimp-print15:14
tilmannipuL: ugh, no thanks :D15:14
nipuLi've already started
cptnso, back to the agenda:15:17
cptn"    * Master/Arch Maintainer politics??? "15:17
tilman??? :)15:17
cptnI think that's mainly about whether we want a "master" arch or not15:17
nipuLkind of15:17
nipuLin that if an x86_64 maintainer introduces a port, and the i686 maintainer decideds to add it15:18
tilmani think we should think of i686 as the master arch, at least for now15:18
nipuLwho is the main maintainer15:18
aoni really think we shouldn't mind the user confusion that much15:18
nipuLis it delegated to the i686 by default, or do we introduce the arch maintainer meta data to the i686 ports15:19
aonimo it should be in i686 too15:19
cptnme too15:19
tilmanthings could get sooo messy...15:20
juehmm, thought we discussed that already15:20
cptnjue: we talked about it, but I'm not sure if we came up with a plan15:21
cptnother than "as long as we don't have an official release, we shouldn't care"15:21
cptnwell, maybe that's wrongly put:15:22
cptn"once we have an offical release, we should allow primary maintenance in both trees"15:22
cptnI think that's about it?15:22
cptntilman: was that an agreeing "mmh"? :-)15:24
tilmando both of you agree with that, or do you just agree that that is the question to solve? :p15:24
cptnor a critical one15:24
tilmansorry. my mmh's are always pondering ones, not agreeing ones ;p15:24
nipuLsounds good to me15:24
nipuLas you said, it's not overly important for now15:25
jueI agree with that15:25
tilmani'm not sure i like the thought of diverging i686 and x86_64 eg15:25
nipuLthat was one thing i liked about my multiarch implementation (even if it was ugly)15:26
nipuLa unified ports tree15:26
juethat will happen anyway, at least if you are looking at contrib or other repos15:26
tilmanbut we should strongly encourage opt maintainers to provide ports for i68615:27
nipuLi have no problems being both15:28
jueyes, core, or better core + opt should be the same15:28
tilman*shrug*, i don't know. didn't think about this much yet :)15:28
nipuLmy core+opt+xorg ports for x86_64 are getting close to identical15:28
nipuLthey contain all the ports i have installed on my x64 box15:29
tilmanxorg.git and xorg-x86_64.git only differ in mesa3d/Pkgfile15:29
tilmanand with the next mesa release (due out ... now) that will vanish, i believe (autoconf...)15:29
nipuLi think jue meant the actual ports, now how they are built15:29
nipuLmost of them are identical though15:30
tilmanie versions?15:30
tilmanokay, i agree with that15:30
tilmanwe should probably discuss x86_64-multilib again15:31
nipuLi was just about to suggest that15:31
tilmanbut i should go to bed soon :p15:31
nipuLpfft, i got up at 5am15:32
nipuLi should still be in bed15:32
tilmansorry :|15:33
tilmannipuL: anyway, we'll need a foo-compat32 port for every libfoo, right?15:33
nipuLnot every one15:33
nipuLjust the ones that are needed to get a binary running15:33
tilmanevery lib that's used by wine and $binary_app_of_your_choice?15:34
nipuLi have my compat32 repos there15:34
tilmandoesn't look too bad15:35
nipuLi haven't tackled wine yet15:35
juethat's a lot15:35
nipuLit could be worse15:36
nipuLthe alternative is a 32bit chroot15:36
nipuLwhat's better re pure/multilib15:38
nipuL2 entirely seperate repos15:38
nipuLor one minimal repo to overlay the other?15:39
tilmandepends on how we solve the glibc/gcc problem that you discussed on the mailing list15:39
tilmani don't remember whether it was solved yet15:39
nipuLi don't think it can be solved15:39
tilmani think i'd prefer an additional repository, but i'm not sure yet15:39
nipuLenabling multilib in gcc required a compat32 glibc15:39
tilmanseriously, i gotta run15:40
tilmannipuL: we should try to talk on the weekend, at some more aussie-friendly hours15:41
cptnbye tilman15:41
nipuLmaybe official meetings can be TZ friendly15:41
juebye tilman15:41
RotwangnipuL: btw. rar conflicts with unrar15:42
nipuLprt-get info unrar15:42
nipuLPackage 'unrar' not found15:42
nipuLok, hangon15:44
Rotwang# prt-get info unrar15:44
RotwangName:         unrar15:44
RotwangPath:         /usr/ports/opt15:44
cptnany other things to discuss?15:45
Rotwangits a minor thing, but any thoughts about crux cloaks?15:45
cptnI think jaeger contacted the freenode stuff a long time ago15:45
cptnnever heard from them15:45
nipuLoh right15:45
nipuLi don't want one15:46
Rotwangbut i want :D15:46
Rotwangwell if no one else wants it, then its not needed15:46
*** thrice` has left #crux-devel15:47
Rotwangbut i think it would be nicer than simple unafilited/nick15:47
cptnnipuL: maybe we could do the meetings in shifts16:03
cptnsuch that we do the first part in CEST16:03
cptnand then a second part the next day (CEST time), when it's evening in .au16:04
cptnalso, we might even shift it an hour back16:05
cptnso that if you join, there will be some bigger overlap16:05
nipuLi can handle one eraly morning a week16:05
nipuLit's good for me16:06
nipuLplus the wife likes it when i'm up to make everyone breakfast16:06
cptnso in the best interest of your wife, we'll keep the time as is :-)16:07
cptnanyway, I'm offline as well16:07
cptnI'll send a summary to crux-devel, hopefully tomorrow16:08
cptnmight become thursday though16:08
cptngood night16:08
*** jue has quit IRC16:30
*** Rotwang has quit IRC17:06
*** Rotwang has joined #crux-devel17:07
*** DarkNekros has quit IRC18:01
*** Rotwang has quit IRC19:18
*** sepen has quit IRC19:26
*** fetid has joined #crux-devel19:58

Generated by 2.11.0 by Marius Gedminas - find it at!