[12:45] <pivi> nawty: it is it is
[12:45] <pivi> good night
[12:45] <Plug> pdr: if you use the PPC version, I believe so
[01:22] <Plug> hmmm
[01:22] <Plug> I restarted apache2 and now it's not interpreting php any more
[01:22] <Plug> just trying to serve it as x-httpd-php
[01:22] <Plug> libapache2-mod-php5 is installed and enabled
[01:30] <Plug> ..that was weird
[01:30] <Plug> and now its working :)
[01:33] <Plug> it seems not to like redirecting / to index.php atm
[01:45] <h3sp4wn> put DirectoryIndex index.php into a file called .htaccess
[01:45] <Plug> it 'used to work'
[04:55] <xerophyte> does anybody know where can i find the cyrus-imapd new version of debs  the drapper only has 2.1.18 but the new version is 2.3.x something
[04:56] <Plug> try apt-get.org, but if its not in dapper, you might have to massage them to get them going on ubuntu
[04:57] <xerophyte> i think the older version of cyrus-imapd has some security issue too
[04:57] <xerophyte> so just wondering why does not drapper didnt update it
[04:58] <Plug> Ubuntu takes what debian sid has at the time
[04:58] <Plug> packages.debian.org - what version is in Debian?
[05:00] <xerophyte> 2.1.18-1
[05:00] <xerophyte> hmmm
[05:00] <xerophyte> but there seems to be some security issue on the version 2.2
[05:00] <xerophyte> this is older than that
[05:00] <xerophyte> and there is 2.3
[05:00] <xerophyte> hmm
[05:01] <Plug> sure, but that doesn't mean someone has gone to the effort required to package the newer version for Debian or Ubuntu
[05:03] <xerophyte> hmmm
[05:12] <xerophyte> if i have file name /usr/bin/mail how can check which package its belong to like rpm -qf filename
[05:13] <Plug> dpkg -S
[05:13] <xerophyte> thx
[06:00] <xerophyte> why there is two version of cyrus-imapd just wondering cyrus21-admin cyrus-imapd-2.2
[06:00] <xerophyte> ??
[06:16] <Plug> because you can install either 2.1 or 2.2
[06:54] <xerophyte> Plug, have you ever setup mailserver with postfix +cyrus-imap + mysql + ssal + mysql-pam + with squirremail i need to setup
[06:54] <xerophyte> so just thinking how can i go about configuration it and get ti working :)
[07:24] <Plug> xerophyte: no
[07:24] <Plug> I prefer Exim as an MTA
[07:43] <maswan> Any plans for a next-gen servercandy spec that I could give some input to, or should i look at speccing up some neat stuff myself? (hpc-ubuntu springs to mind)
[07:48] <fabbione> maswan: hpc is in my list too and it was specced for dapper..
[07:48] <fabbione> maswan: you might want to pick up the old UbuntuClusters wiki page, clean it up and make it ready for Edgy
[07:48] <fabbione> be aware to please keep the dapper stuff untill dapper is released
[07:52] <maswan> fabbione: Yeah, it was something like that I was thinking of. Any suggestions on how to do the wiki thing? Copy to EdgyClusters?
[07:53] <fabbione> maswan: i suggest you add all at the bottom of UbuntuClusters
[07:53] <fabbione> we will clean it up when it's time
[07:53] <fabbione> like:
[07:53] <fabbione> [07:53] <fabbione> and starts from there
[07:53] <maswan> Ah, ok. I'll take a go at that then.
[07:54] <fabbione> yes
[07:54] <fabbione> there are for sure leftovers from Dapper that we want to carry to Edgy
[08:15] <maswan> Now, what were all those neat ideas I had yesterday regarding this? :)
[08:20] <maswan> Ok, a first dump of edgy crack there.
[08:20] <maswan> I'll talk to my collegues here about what kind of crack we'd appriciate. :)
[08:22] <maswan> "We will be changing from Debian to Ubuntu at Benedict, and I would like to known which releases we should keep" <- more academic HPC clusters on ubuntu coming up. :) You happen to be close to Aalborg University, fabbione? :)
[08:22] <fabbione> someone... it's about 400km
[08:22] <fabbione> somehow ^
[08:24] <fabbione> maswan: i am pretty sure i can get sponsored to be there if there is interests in discussing the topic
[08:24] <fabbione> but that assumes that it's not during the release process of Dapper
[08:25] <maswan> fabbione: Probably not, just a point of curiousity really.
[08:25] <fabbione> ok :)
[08:27] <maswan> Local scratchpad: SSI, Xen
[08:28] <maswan> openafs
[08:32] <fabbione> maswan: please don't mix Xen with cluster..
[08:32] <fabbione> it's really not the same thing
[08:33] <maswan> fabbione: Well, it has real applications here for grid clusters in sandboxing processes.
[08:33] <maswan> s/processes/jobs/
[08:33] <maswan> but sure, I can avoid it here. :)
[08:34] <fabbione> Xen has to be considered kernel/virtualization spec
[08:34] <fabbione> it's not "server" or "cluster" spec
[08:34] <fabbione> note: *spec*
[08:34] <maswan> Currently braindumping, needs to be spec:ed.
[08:35] <maswan> Hmm, there is an openafs spec around too.
[08:41] <maswan> Hmm.. "Sandboxing" in the more general sense might make sense to include though. In that if you have a batch system setup, you want to be sure you've cleaned up after the previous job before starting the next.
[08:47] <maswan> Does that make sense in a spec way?
[08:51] <fabbione> + Profiling: A vmlinux binary should probably be shipped somewhere for distribution kernels, in a well-documented or well-known location, for use with oprofile and so.
[08:51] <fabbione> ^^this is kernel spec
[08:52] <fabbione> + Commercial support advocacy: A collaborative way of indicating that you want support for weird hardware without free drivers (quite common for narrow high-performance devices) or commercial software. This might be neat if it can be used to lobby some manufacturers into supporting more than just redhat and suse. This is possibly out of scope for this spec, but this is one place where the need can be seen.
[08:52] <fabbione> there is already a spec for that
[08:52] <fabbione> 3rdy part vendors something support
[08:52] <maswan> 'k
[08:53] <maswan> only thing I find is an UDU BOF with no real content
[08:54] <maswan> and only regarding software
[08:54] <fabbione> maswan: i remember seeing one
[08:54] <fabbione> but for that kind of thing you better talk to Malcom Yates directly
[08:54] <fabbione> mdy@canonical.com
[08:55] <fabbione> because it's something that usually starts from $company and not really $distro
[08:56] <fabbione> + Tuning knowledgebase or good defaults for various high-performance loads (databases, fileservers, number crunching, networking, etc).
[08:56] <fabbione> i am not sure what you mean here
[08:56] <fabbione> i think you are mixing a few concepts together
[08:56] <maswan> The thing I was aiming towards was a way to tell $company that there is a large bunch of users that want it. Otherwise $company usually requires $money.
[08:56] <maswan> Probably
[08:56] <fabbione> when you write a spec you need to isolate what you want
[08:56] <fabbione> you want HPC, right?
[08:57] <fabbione> what is HPC..
[08:57] <fabbione> HPC is.. bla bla bla foo bar
[08:57] <fabbione> what are the HPC solutions out there?
[08:57] <fabbione> a) b) c) d)
[08:57] <fabbione> which one make more sense to bring in and why?
[08:57] <fabbione> a) c)
[08:57] <fabbione> how do we push them in Ubuntu
[08:57] <fabbione> we do: foo) bar) baz) biz) ping) pong)
[08:57] <fabbione> stop
[08:58] <fabbione> what you are writing in the specs are just general ideas
[08:58] <fabbione> and most of them not even relevant for Clustering
[08:58] <fabbione>  high-performance loads might need to be achieved indipendently of a cluster environment
[09:00] <maswan> Ok, I'll go over that and remove most of the things and leave the relevant.
[09:00] <maswan> The thing is, on another level, clusters are never a goal in themselves. You want either performance or reliability.
[09:00] <maswan> So yes, I'm confusing goal with method on several levels.
[09:01] <maswan> That'll probably take an hour or so, but I will get back to it.
[09:06] <fabbione> take your time
[09:06] <fabbione> the goal of a spec is to get from 0 to be able to install a working foo
[09:07] <fabbione> try to think at distcc
[09:07] <fabbione> the most "stupid HPC cluster"
[09:07] <fabbione> - get distcc packaged
[09:07] <fabbione> - move it to main
[09:07] <fabbione> - auto install distcc if option foo is selected
[09:08] <fabbione> - set default sane config (if possible)
[09:08] <fabbione> - auto start distcc at boot
[09:08] <fabbione> this is how a spec would look like
[09:08] <fabbione> now if you think it in HPC terms
[09:08] <fabbione> let say that a lot of people use these MPI implementation:
[09:08] <fabbione> - what is the best MPI implementation out there: foo
[09:09] <fabbione> ok it's foo...
[09:09] <fabbione> than just s/distcc/foo/
[09:09] <fabbione> and you get your spec
[09:09] <fabbione> more or less
[09:09] <fabbione> the final app written on top of MPI is not our goal
[09:09] <fabbione> it can't be
[09:09] <fabbione> our goal is to provide the underground for it
[09:12] <maswan> Yeah. That's why stuff like profiling ended up there though, because I'm thinking "cluster as usable computational resource"
[09:15] <maswan> This is a large and complex system though, and the base OS is just part of it. Trying to figure out what bits are needed, and what parts are resonable to call Clusters
[09:15] <fabbione> maswan: i know.. it's not easy
[09:16] <fabbione> that's why you usually sit around a table with N other people that have experience in that field
[09:16] <maswan> Specs are much simpler if you have just one feature. :)
[09:17] <neuralis> fabbione: yeah, i imagine we'll want at least one clustering and virtualization bof at the next dev conference
[09:17] <fabbione> the point is that you can split the spec into multiple smaller specs
[09:18] <fabbione> neuralis: yes, too bad i won't be there. so you will have to work it out for me
[09:18] <neuralis> fabbione: do we have dates nailed down for that yet? i'm not positive if i'll be able to come yet, either.
[09:19] <fabbione> neuralis: i think it's going to be one or two weeks after release
[09:19] <fabbione> no later than that..
[09:19] <fabbione> how long? no clue
[09:21] <maswan> Yeah, it'd be neat to be able to get input as such a bof, but being not being a developer it's hard to motivate going to a dev confereance
[09:23] <fabbione> uh why?
[09:23] <fabbione> you as a "user" have quite a lot of voice
[09:24] <fabbione> it's not like  you are asked to sit in a corner and listen
[09:24] <fabbione> ask people that have been there
[09:24] <fabbione> neuralis for example has been almost crossburned alive
[09:24] <maswan> Ehm, motivating to your side, but motivating to myself and to my employer
[09:25] <neuralis> maswan: well, your employer is another story. but you should have plenty of motivation to come.
[09:25] <maswan> Of course, if it happens to be resonably close, it might happen. But probably not travelling accross the world.
[09:25] <fabbione> should be EU this time
[09:26] <maswan> That'd be considerably cheaper, yes. It also depends on the general setup, if it is only 2 hours of server/hpc and the rest is silly desktopping... It's a priority question, really, between time and money and stuff on the other side.
[09:27] <neuralis> maswan: there will be several server bofs, but probably not much more than 2-3 hours total.
[09:28] <maswan> Anyway, I'm interested in going yes, so far I haven't been able to talk my wallet or my employer into it though. :)
[09:28] <fabbione> maswan: you know that we do sponsor people?
[09:28] <fabbione> not everybody
[09:28] <fabbione> but there is such option
[09:29] <fabbione> also
[09:29] <fabbione> bofs are really a matter of who is there
[09:29] <fabbione> some bofs can take hours and hours
[09:29] <fabbione> others are like 10 minutes
[09:29] <fabbione> it seriously depends who is sitting around the table
[09:29] <fabbione> neuralis and I have been fighting for about 3 hours on a solution
[09:29] <fabbione> YES
[09:29] <maswan> Hmm.. Also, we might look at an internal mirroring bof, that'd be most interesting if we can get the right people to it. That's a pretty internal discussion, sure, but it is interesting and important.
[09:29] <fabbione> NO
[09:29] <fabbione> YES
[09:29] <fabbione> NO
[09:29] <fabbione> etc.
[09:30] <fabbione> maswan: suggest it
[09:30] <fabbione> register the spec in launchpad and open a page on the wiki
[09:32] <neuralis> maswan: what do you mean when you say internal mirroring?
[09:33] <maswan> neuralis: mirroring of ubuntu
[09:33] <maswan> bbiab
[09:34] <neuralis> fabbione: i think it's not clear to most people what bofs are in the ubuntu dev conference sense, because they're not just a place for people with similar interests to get together and hang out, but to actually work on some aspect of improving ubuntu and then write it up
[09:35] <neuralis> fabbione: (unless i'm not getting what maswan is talking about, he wants a bof in the traditional sense)
[09:37] <fabbione> yeah i think so too
[10:06] <maswan> This is true, and yes, that's the primary bit I was thinking of. Overloading a term always has that problem though.
[10:12] <maswan> Because what'd be most useful for the mirroring situation is probably a traditional bof. It can probably be disguised by some generic "people download isos fast"-spec, I guess. But I dont' know.
[10:12] <neuralis> i believe the plan is to offer a community/users' conference separately from the development conference
[10:12] <neuralis> or at least mark said he was thinking about it back in montreal
[10:13] <neuralis> so while users will be welcome at the devconf, it will be a *dev* conf, meaning informational bofs (like this one with mirroring) will be off the table.
[10:14] <maswan> Not so much informational as unstructured. There will be a goal, in making the mirroing infrastructure better.
[10:15] <maswan> It's just that the hows etc are unknown.
[10:31] <maswan> Heh. I give up, now it's clear that that stuff is comments, not an actual spec. I can't wrap my head around UbuntuClusters as being a feature that we need to get into eft. I can just think of improvements to it, most of which are more generic.
[10:46] <maswan> Btw, regarding the metadiscussion, perhaps a development process transparacy project or howto might be good?
[10:53] <maswan> Something I probably can write at least half a spec for is LustreInUbuntu, I don't know what to do with the implementation though.
[11:39] <allee> maswan: where is the openafs spec? I only found OpenAFSSupport
[12:14] <neuralis> maswan: write the (half) spec
[12:14] <neuralis> maswan: we'll see what we can do with it.
[12:34] <maswan> allee: That's what I saw.
[12:37] <maswan> neuralis: I'll let it simmer in my head a bit to get a proper spec out of it, expect it on monday
[07:38] <pdr> Plug: thanks, will give it a go in the near future
[07:39] <pdr> (re ppc install on ibm servers)
[10:49] <xerophyte> Could somebody compare the Hula vs (postfix+cyrus-imap+squirellmail) ??? brief should good good