Fedora 11 Features (2009-02-08 classroom)

From FedoraProject

Jump to: navigation, search
---nirik has changed the topic to: Fedora IRC Classroom Fedora 11 Features with your teacher Jon Stanley ( jds2001 ) - See https://fedoraproject.org/wiki/Classroom for tonights class schedule. Feb 07 04:16
-->XanThaOs (n=sacredde@71-90-228-054.dhcp.gnvl.sc.charter.com) has joined #fedora-classroom Feb 07 04:16
-->jzhou (n=jzhou@pcp061561pcs.unl.edu) has joined #fedora-classroom Feb 07 04:16
-->Dr_willis (n=willis@c-67-163-55-207.hsd1.il.comcast.net) has joined #fedora-classroom Feb 07 04:16
jds2001 alright, sorry about that delay Feb 07 04:16
brunowolff I think yum info will tell you that as long as the packname is unique across repos. Feb 07 04:16
-->tw2113 (n=tw2113@fedora/tw2113) has joined #fedora-classroom Feb 07 04:17
nirik fenris02: thats a tough one. gpg key that it's signed by? ;) Feb 07 04:17
nirik anyhow, jds2001: take it away! Feb 07 04:17
-->joropo (n=joropo@ has joined #fedora-classroom Feb 07 04:17
jds2001 so this came from a requested class, but I'm not sure what folks wanted. Feb 07 04:17
nirik rtnpro_: http://www.redhat.com/archives/fedora-devel-list/2008-August/msg00041.html Feb 07 04:18
-->mwilson (i=foobar@ip68-101-143-117.sd.sd.cox.net) has joined #fedora-classroom Feb 07 04:18
rtnpro_ nirik, thank you Feb 07 04:18
jds2001 I'm Jon Stanley, current chair of the Fedora Engineering Steering Committee. We handle the "Feature Process", which is the process by which new features get accepted into Fedora. Feb 07 04:18
fenris02 nirik, crumb. that was my answer too. Thanks :) Feb 07 04:18
-->openpercept (n=openperc@unaffiliated/openpercept) has joined #fedora-classroom Feb 07 04:19
-->Guest456 (i=KVIrc@c80-216-141-34.bredband.comhem.se) has joined #Fedora-Classroom Feb 07 04:19
jds2001 The current list of features is at https://fedoraproject.org/wiki/Releases/11/FeatureList Feb 07 04:19
jds2001 there are a few that I'd like to hit on that are going to be major ones for F11. Feb 07 04:19
jds2001 first, on new installs, our default filesystem is likely to be ext4 Feb 07 04:20
jds2001 we're currently on the lookout for data corruption/stability bugs and are going to revisit this in about a month. Feb 07 04:20
jds2001 however I think it's somewhat safe to say it's in. Feb 07 04:21
jds2001 It is no longer experimental upstream, either. Feb 07 04:21
jds2001 Another one is presto Feb 07 04:22
aTypical Sorry, do we question as you go or wait until the end? Feb 07 04:22
jds2001 jump right in, i have no clue what folks want :) Feb 07 04:22
jds2001 very interacrive is my style :) Feb 07 04:22
aTypical How did ext4 get the nod? Is it ready? Feb 07 04:23
jds2001 we believe so. This testing period is critical for determining that. Feb 07 04:23
che aTypical, there was a session in fedora-qa recently Feb 07 04:23
jds2001 yeah, there was an actual test day on it, I guess i'll bring those up too. Feb 07 04:24
jds2001 https://fedoraproject.org/wiki/QA/Test_Days/F11 is the schedule for "test days" Feb 07 04:24
brunowolff How are kernel updates going to work for 586/686? Feb 07 04:24
mwilson EXT4 has a working fsck? Feb 07 04:25
brunowolff Is something going to switch us automatically? Feb 07 04:25
jds2001 these are days of very focused QA Feb 07 04:25
jds2001 brunowolff: nope, fresh installs only Feb 07 04:25
jds2001 mwilson: yes, e4fsprogs exists. Feb 07 04:25
nirik mwilson: yes Feb 07 04:25
<--openpercept (n=openperc@unaffiliated/openpercept) has left #fedora-classroom Feb 07 04:25
brunowolff The latest rawhide kernel doesn't even have a 686 arch version. Feb 07 04:25
jds2001 brunowolff: really? Feb 07 04:26
-->cuga (i=cuga@CPE00080232ec3b-CM0012c92137d6.cpe.net.cable.rogers.com) has joined #fedora-classroom Feb 07 04:26
jds2001 must be some build failure. Feb 07 04:26
<--yunustj_afk has quit (Read error: 110 (Connection timed out)) Feb 07 04:26
nirik no, thats right... you want the PAE one. Feb 07 04:26
jds2001 so while we're on that topic, we're also changing what architectures we support and default compiler flags. Feb 07 04:26
nirik or the i586 one. Feb 07 04:26
mwilson This isn't like reiser always supposedly had a working fsck, except all it did was eat the data, I hope. No one else I've heard has committed to distributing on EXT4. Feb 07 04:26
jds2001 oh, OK. Feb 07 04:26
brunowolff I don't think my home machine supports PAE, so I probably want the 586 one? Feb 07 04:27
nirik brunowolff: what is it? look up it's specs. Likely it does support PAE. :) Feb 07 04:27
jds2001 if it's recent, I'm sure it does. Feb 07 04:28
jds2001 what processor have you got? Feb 07 04:28
brunowolff Its a 2 cpu athlon tbird. Feb 07 04:28
Falstius jds2001: what are they changing to? Feb 07 04:28
jds2001 Falstius: the specific compiler flags, im not sure Feb 07 04:28
brunowolff I'll be able to test it after I get home, as I have the rawhide updates with me. Feb 07 04:28
jds2001 but we're switching the target arch from i386 to i586 Feb 07 04:29
Falstius ok Feb 07 04:29
jds2001 so we pick up some new instructions that leads to more effeciencies. Feb 07 04:29
<--jzhou has quit ("Leaving") Feb 07 04:29
jds2001 along those same lines, we'll be installing an x86_64 kernel on 32-bit installs if the processor supports it. Feb 07 04:29
Falstius what about support for atom? (I know gcc isn't really optimized yet) Feb 07 04:30
mwilson What about user-space? Feb 07 04:30
jds2001 mwilson: userspace would still be 32-bit Feb 07 04:30
brunowolff Does /proc/cpuinfo show that info? Maybe in address size? Feb 07 04:30
jds2001 but you win from the kernel being x86_64 even. Feb 07 04:30
Falstius interesting Feb 07 04:31
jds2001 the biggest win there is the elimination of the highmem/lowmem split. Feb 07 04:31
mwilson Is that going to be a Fedora-ism, or is running a 64-bit kernel and a 32-bit userspace a normal thing? Feb 07 04:31
rtnpro_ jds2001, ! Feb 07 04:32
jds2001 I'm not actually sure, I've not looked at what other distros are doing. Feb 07 04:32
jds2001 rtnpro_: go ahead, no need for that with me :) Feb 07 04:32
fenris02 jds2001, after years of sticking to i386 set, why switch now? Feb 07 04:33
jds2001 mwilson: but even if it's a Fedora-ism, it still makes sense. Feb 07 04:33
rtnpro_ jds2001, Does that mean that with the x86_64 kernel, the kernel processes will be faster and again retain the compatibility with 32-bit applications? Feb 07 04:33
jds2001 rtnpro_: not really faster, but compatibility is retained. Feb 07 04:34
mwilson I meant in the sense of is Fedora implementing some sort of thunking layer? I always thought the kernel and user space had to match, other than compatability bits. Feb 07 04:34
rtnpro_ rtnpro, ok Feb 07 04:34
nirik fenris02: newer gcc, some more gains. Also, less really really old hardware around. Feb 07 04:35
jds2001 mwilson: not necessarily - we're not implementing anything new here other than installing an x86_64 kernel. Feb 07 04:35
fenris02 mwilson, 64-bit kernel, 32-bit userspace means that every app will be thunk'd (sp?) Feb 07 04:35
mwilson Fedora has never been about supporting old hardware. People using old hardware should be discouraged from running Fedora, their experience will not be optimal. Feb 07 04:35
jds2001 apps can exist at different spaces within the memory the kernel can address. Feb 07 04:35
jds2001 so even though a 32 bit app can only see 4GB, that could be from 3-7GB physical addressing. Feb 07 04:36
jds2001 oversimplifying, of course :D Feb 07 04:36
mwilson Can you talk about DeviceKit, and how Fedora plans to avoid another fiasco like PulseAudio? Feb 07 04:36
<--rtnpro has quit (Read error: 113 (No route to host)) Feb 07 04:37
jds2001 sure, DeviceKit is a replacement for a moribund hal. Feb 07 04:37
fenris02 mwilson, esd was along the same lines, but worse. PA is an improvement. Feb 07 04:37
mwilson OTOT, the bleeding edge is what it is. Feb 07 04:37
jds2001 it's not going to replace hal immediately Feb 07 04:37
mwilson Except PA was nowhere near ready, but it was made the default anyway. Feb 07 04:38
*jds2001 isn't really a desktop guy himself, so I'm a little out of my domain of expertise with DeviceKit. Feb 07 04:39
<--XanThaOs has quit (".::The Reborn Of Evil Spirits::. €ZoMBiE€ Sc®ïþt 2.2 .ßý. xom[b] .::[1]::.") Feb 07 04:39
mwilson Something like HAL is a little more essential to the proper working of the machine, these days. Feb 07 04:39
jds2001 but I believe it's aim is to *eventually* replace hal. Rome wasn't built in a day, however. Feb 07 04:39
fenris02 jds2001, any plans to revamp /etc/init.d/* items to use upstart scripts directly? Feb 07 04:40
jds2001 right now, aiui, it's handling disk devices only Feb 07 04:40
jds2001 fenris02: i think that's blocking on essentially a rewrite of upstart. Feb 07 04:41
<--tw2113 (n=tw2113@fedora/tw2113) has left #fedora-classroom ("There are two major products that come out of Berkeley: LSD and UNIX. We don't believe this to be a coincidence.") Feb 07 04:41
jds2001 nirik: am i remembering wrong here? Feb 07 04:41
mwilson Ubuntu isn't still trying to emulate SysV, are they? Feb 07 04:41
nirik yeah, I don't think devicekit is going to replace hal yet... perhaps next cycle if it's ready. Feb 07 04:41
nirik mwilson: they do as far as I know. Feb 07 04:41
nirik there is work on a upstart re-write going on. No idea where it is tho. Feb 07 04:42
mwilson This 20 second startup thing sounds like there's upstart work being done in Fedora. Feb 07 04:42
fenris02 will readahead be tuned such that it pulls more than 4k by default? :) Feb 07 04:42
jds2001 actually I thought that too. There's no real upstart work involved there. Feb 07 04:43
jds2001 fenris02: no idea :) Feb 07 04:43
nirik see http://screwyouenterpriseedition.blogspot.com/ for info on the upstart re-write. I don't think there will be any move to switch sysvinit scripts until thats done and ready. Feb 07 04:44
jds2001 one other feature i wanted to touch on is presto. Feb 07 04:45
jds2001 as nirik mentioned in his yum talk, there's been a yum plugin to support deltarpm's for a good time now. Feb 07 04:46
jds2001 but we lacked bodhi/mash/createrepo support for it. Feb 07 04:46
mwilson presto's been talked about since F7. Feb 07 04:47
jds2001 seth made it so it's a simple flag to createrepo now, so almost no bodhi changes are needed. Feb 07 04:47
jds2001 mwilson: yep, and it's well on it's way to being an official reality I think :) Feb 07 04:47
jds2001 oh, and one more thing worthy or note Feb 07 04:50
jds2001 the text UI in anaconda is being reworked. Feb 07 04:50
jds2001 NO. it's not going away as you may have heard. Feb 07 04:50
jds2001 but it will have less functionality, and become the truly minimal install experience that folks have asked for. Feb 07 04:51
fenris02 suitable for server installs? (ie: no xorg, no bluez, ...) Feb 07 04:52
jds2001 we even more strongly urge folks to use VNC or xdriver=vesa now. Feb 07 04:52
mwilson Fedora isn't suitable for servers no matter what the installer looks like. Feb 07 04:52
fenris02 mwilson, not the point. (thinking more about el6 actually) Feb 07 04:52
jds2001 fenris02: not sure, but aiui, you'll be able to get new packages, and that's about it. Feb 07 04:52
jds2001 i've actuallly not done a text install since the anaconda changes landed. Feb 07 04:53
mwilson It sounded for a while like Fedora was going to take away text, period. Who was that, Nottingham? I guess that argument was lost. Feb 07 04:54
<--franciscod_ has quit ("leaving") Feb 07 04:54
<--francisc1d_ has quit ("leaving") Feb 07 04:54
fenris02 "blockdev --getra /dev/mapper/*" - default is 256. changing this to something larger (like 4096) increases perf from say 140MB/s to 320. ymmv of course. (Hence the earlier question, but I had to find the details) Feb 07 04:54
-->franciscod (n=Ankur@ has joined #fedora-classroom Feb 07 04:54
mwilson Is that number real? How do you get 320 out of real disk hardware? Feb 07 04:55
fenris02 mwilson, it's real. Feb 07 04:55
fenris02 mwilson, software-raid5 Feb 07 04:55
nirik fenris02: I saw those bugs... I don't see it here on my rawhide machine, but don't know if it was fixed or something about the way I installed changed it. Feb 07 04:55
mwilson Is LVM going to still be the default, or will Fedora *finally* move away from that? Feb 07 04:56
fenris02 mwilson, is there something better around? Feb 07 04:56
mwilson There's no *reason*, most of the time. Feb 07 04:57
jds2001 really? Feb 07 04:57
fenris02 mwilson, resize and snapshots Feb 07 04:57
jds2001 i like LVM Feb 07 04:57
mwilson Let's make a LVM, just so we can make one slice inside it. Feb 07 04:57
jds2001 LVM will always be the default, im sure. Feb 07 04:57
nirik there was talk of it, but it has not gotten to fesco yet. ;) http://fedoraproject.org/wiki/Features/NoDefaultLVM Feb 07 04:57
<--che has quit ("Verlassend") Feb 07 04:57
mwilson Complexity for the sake of complexity is never good. Feb 07 04:57
*jds2001 is somewhat opposed to that. Feb 07 04:58
nirik it may go away when btrfs comes in too tho, as it has a lot of lvm functionality already built in. Feb 07 04:58
mwilson Most people don't resize and snapshots are like backups, no one does. Feb 07 04:58
jds2001 you can do stuff with root on LVM that you can't do otherwise. Feb 07 04:58
mwilson But are they things that need doing? Feb 07 04:58
fenris02 elevator/deadline/none iosched tuning would be fitting too. that should be determined by hardware type. (none for flash media for ex) Feb 07 04:58
*nirik notes that crypt is done via dm-crypt, so it kinda needs lvm Feb 07 04:59
fenris02 mwilson, snapshots for backups are absolutely *awesome*. You have a point though. However, I'd disagree with making defaults change just because users are lazy Feb 07 05:00
fenris02 nirik, just dm-crypt can be done w/o lvm, no? Feb 07 05:00
nirik not sure, it may be able to... Feb 07 05:00
jds2001 yeah, you can encrypt physical devices Feb 07 05:01
mwilson fenris02: No, there's something someone should propose. Take all these pseudo-TimeMachine-like things and actually roll them into a distribution. Make backups easy enough that it's harder NOT to do it, than to do it. Feb 07 05:01
kaos01 i thing lvm sesize is extremely usefull Feb 07 05:01
jds2001 that's the default if you select encryption actually - it encrypts your partition that the rootvg is on. Feb 07 05:01
mwilson Resize is a non-starter for most everyone. Feb 07 05:01
fenris02 mwilson, good point. have any suggestions how? Feb 07 05:01
brunowolff I run dm-crypt over md raid and no lvm on my systems. Feb 07 05:02
jds2001 oh, and one more change that there is. Feb 07 05:02
fenris02 resize is "hard" because decent gui tools are non-existant Feb 07 05:02
jds2001 the anaconda-created volume group is now named for the hostname of the machine. Feb 07 05:02
jds2001 rather than VolGroup00 Feb 07 05:02
fenris02 VG-host/LV-slice ? Feb 07 05:02
nirik jds2001: excellent. Feb 07 05:02
mwilson Yes, but making it easier to break the box isn't on, either. Wrap this shiny GUI around it, everyone thinks they can do it, until they hose it, and don't have a backup of it. Feb 07 05:03
jds2001 which is great, since moving drives from machine to machine is easier. Feb 07 05:03
mwilson Some things should still be hard. Feb 07 05:03
fenris02 having two drives with VolGroup00 is "entertaining" Feb 07 05:03
nirik just like two with 'LABEL=/' :) Feb 07 05:03
fenris02 very much like that Feb 07 05:03
mwilson One of these days I'm going to look at the LVM toolchain enough to learn how to change that VoLGroup00 name to what I want it to be. Feb 07 05:04
<--franciscod has quit ("leaving") Feb 07 05:04
jds2001 mwilson: vgrename Feb 07 05:05
mwilson fenris02: In answer to your question, no, I don't know how. I do it with rsync and a script. But someone can wrap a GUI around the concept and make it easy enough to use. Ubuntu tried, but it went nowhere. Feb 07 05:06
fenris02 mwilson, mdns + rdiff-backup should work, but needs a wrapper Feb 07 05:06
mwilson Why does mdns need to be a part of it? Feb 07 05:06
fenris02 mwilson, make it work automagically Feb 07 05:07
mwilson Complexity for complexity's sake again. Feb 07 05:07
fenris02 mwilson, how else would you propose a system perform automated backups w/o user intervention? Feb 07 05:07
mwilson You presume that most people would be doing it over a network, vice to another local device. Feb 07 05:08
mwilson I presume the reverse. Feb 07 05:08
fenris02 *nod* comparing with timemachine. Feb 07 05:11
jds2001 anyhow, anything else? Feb 07 05:11
jds2001 I think I'm almost out of time here. Feb 07 05:11
<--TheFord has quit ("Leaving") Feb 07 05:12
mwilson New python, but that's just, well, new python. Feb 07 05:12
nirik crashcatcher might be nice... we will see how it turns out. Feb 07 05:12
jds2001 yeah, and no idea when py3k will land - that will be the interesting one. Feb 07 05:12
*jds2001 suspects not before F13 Feb 07 05:13
mwilson Fedora can put as many bandaids on the bug process as they like, the main problem is bugzilla itself. Feb 07 05:13
*nirik thought the main problem was not enough people to fix all bugs. Feb 07 05:14
jds2001 well that's a little out of scope here, and we're out of time, but what else do you propose? Feb 07 05:14
fenris02 nirik, barely sufficient to sort out dups anyway. let alone EOL and NEEDSINFO Feb 07 05:14
jds2001 i think that's the case with anything, really. Feb 07 05:15
nirik yeah. Feb 07 05:15
<--aTypical has quit ("Leaving") Feb 07 05:15
nirik thanks for coming jds2001 ! Feb 07 05:15
fenris02 if bugcatcher can properly file them under the correct app - that alone is a perk. Feb 07 05:15
jds2001 np, look forward to coming back again! Feb 07 05:16
mwilson Bugzilla is opaque, always has been, hard to interface with. I always look at Debian's BTS as how you *should* do bugs. Make the bar as low as possible. The idea that you have to have a *login* to file a bug... Feb 07 05:16
fenris02 s/bugcatcher/crashcatcher/ Feb 07 05:16
---nirik has changed the topic to: Fedora IRC Classroom - 15min break. Next class at 05:30UTC - see https://fedoraproject.org/wiki/Classroom for schedule of tonights classes. Feb 07 05:16
mwilson Anyway, old argument. Feb 07 05:16
mwilson Thanks for the informative meet. Feb 07 05:16
fenris02 thanks jds2001 Feb 07 05:16
herlo thanks jds2001 Feb 07 05:17
nirik yeah, and provide enough info to figure out whats going on... many bug reports lack info and reporters don't respond to fill them in. Feb 07 05:17
inode0 release lifecycle is not conducive to end user satisfaction with whatever bug processing gets done - they will always move on without getting bugs resolved Feb 07 05:19
mwilson Having more bugs filed is always preferable to no bugs getting filed. You can always close them. Feb 07 05:22
<--KishanGoyal has quit (Read error: 60 (Operation timed out)) Feb 07 05:22
<--mwilson (i=foobar@ip68-101-143-117.sd.sd.cox.net) has left #fedora-classroom ("Leaving") Feb 07 05:22

Generated by irclog2html.py 2.7 by Marius Gedminas - find it at mg.pov.lt!