From FedoraProject

Jump to: navigation, search

Fedora Release Engineering Meeting :: Monday 2007-11-19

Topics Discussed

  • Releasing Fedora 9 on CD--up to the Fedora Board to decide what they want
  • Enabling Jigdo
  • create and host templates outside of Fedora for now
  • if in time there is enough demand, possibly add to compose process

IRC Transcript

--- Log opened Mon Nov 19 12:58:47 2007
-!- Irssi: #fedora-meeting: Total of 82 nicks [1 ops, 0 halfops, 0 voices, 81 normal] <a href="#t12:58" class="time">12:58</a>
-!- Irssi: Join to #fedora-meeting was synced in 1 secs<a href="#t12:58" class="time">12:58</a>
* kanarip is here<a href="#t13:01" class="time">13:01</a>
* poelcat here<a href="#t13:03" class="time">13:03</a>
* jwb is here<a href="#t13:04" class="time">13:04</a>
f13oh yeah.<a href="#t13:05" class="time">13:05</a>
f13it's that time.<a href="#t13:05" class="time">13:05</a>
-!- f13 changed the topic of #fedora-meeting to: Fedora Release Engineering Meeting<a href="#t13:05" class="time">13:05</a>
f13jeremy notting warren wwoods spot ping<a href="#t13:05" class="time">13:05</a>
* spot is around<a href="#t13:05" class="time">13:05</a>
* jeremy is around<a href="#t13:05" class="time">13:05</a>
* warren here<a href="#t13:06" class="time">13:06</a>
* EvilBob is round...<a href="#t13:06" class="time">13:06</a>
f13kanarip: what was the url to your wiki page regarding jigdo?<a href="#t13:07" class="time">13:07</a>
* notting is here<a href="#t13:07" class="time">13:07</a>
kanarip<a href=""></a><a href="#t13:07" class="time">13:07</a>
f13thanks, adding it to the agenda page, I forgot to do that earlier.<a href="#t13:08" class="time">13:08</a>
-!- f13 changed the topic of #fedora-meeting to: Fedora Release Engineering - <a href=""></a><a href="#t13:10" class="time">13:10</a>
f13lets all take a moment to review this<a href="#t13:10" class="time">13:10</a>
* warren tries to load page<a href="#t13:10" class="time">13:10</a>
* warren still waiting<a href="#t13:11" class="time">13:11</a>
jwbso that is mostly a benefit for the mirror servers?<a href="#t13:13" class="time">13:13</a>
f13jwb: it can be a benefit to users too<a href="#t13:13" class="time">13:13</a>
poelcatjwb and users too<a href="#t13:13" class="time">13:13</a>
jwbhow?<a href="#t13:13" class="time">13:13</a>
f13jwb: if they have content downloaded, jigdo can use that downlaoded content to construct the iso, instead of having to download the entire iso<a href="#t13:14" class="time">13:14</a>
jwbi rolled that into mirroring<a href="#t13:14" class="time">13:14</a>
poelcatmy own example is that I sync a local rawhide tree and usually when a new iso comes out I have 75% of the content already... why download it all again via bittorrent?<a href="#t13:14" class="time">13:14</a>
spotmaybe i don't understand jigdo, but it sounds like it would make getting isos massively more complicated than it is now.<a href="#t13:14" class="time">13:14</a>
kanaripjwb, the footprint on a mirror decreases by not having to host the .iso file<a href="#t13:14" class="time">13:14</a>
nottingf13: right, but that implies they *do* have the content first<a href="#t13:15" class="time">13:15</a>
jwbpoelcat, perhaps my definition of "user" is too limited.  you (and i since i do the same) would not be considered "users" in my terms<a href="#t13:15" class="time">13:15</a>
jeremyspot: it does<a href="#t13:15" class="time">13:15</a>
jwbkanarip, right that's the major benefit i see<a href="#t13:15" class="time">13:15</a>
kanaripspot, however Jigdo isn't mutually exclusive with torrent, or with having the .iso files on the mirrors as well<a href="#t13:15" class="time">13:15</a>
nottingone thing about the notes listed - we actually put repoview on the isos?<a href="#t13:15" class="time">13:15</a>
spotfor example, if i wanted to download fedora to a windows box...<a href="#t13:15" class="time">13:15</a>
poelcatjwb: agreed--today :) but why not provide the opportunity to expand it<a href="#t13:15" class="time">13:15</a>
kanaripnotting, yes Fedora 8 had repoview/ on the ISOs<a href="#t13:16" class="time">13:16</a>
notting... why?<a href="#t13:16" class="time">13:16</a>
* jeremy is very against not having isos on mirrors. if the end result iso can be verified (sha1sum) vs the published sha1sum, I guess I'm okay with publishing jigdo files<a href="#t13:16" class="time">13:16</a>
f13notting: I forgot to exclude it.<a href="#t13:16" class="time">13:16</a>
nottingjeremy: technically, spins aren't on the mirrors<a href="#t13:16" class="time">13:16</a>
spotjeremy: agreed. i'm not against jigdo + isos, but only jigdo? Nah.<a href="#t13:16" class="time">13:16</a>
jwbjeremy, yeah.  i see it as an add-on, not a replacement<a href="#t13:16" class="time">13:16</a>
kanaripnotting, however one can exclude those files from becoming 'slices' the client wants to obtain from the mirrors and have it included in the .template files<a href="#t13:16" class="time">13:16</a>
f13kanarip: actually we need to just keep repoview off the iso period.<a href="#t13:17" class="time">13:17</a>
f13I need to make that change in pungi<a href="#t13:17" class="time">13:17</a>
f13jeremy: I'm with you.<a href="#t13:17" class="time">13:17</a>
kanaripjeremy, yes sha1sums can be verified on the .iso composed with jigdo<a href="#t13:17" class="time">13:17</a>
jeremynotting: given that all of the spins are live images, jigdo is irrelevant for them, so ... :)<a href="#t13:17" class="time">13:17</a>
f13The big question I have then is, for normal users (IE ones without a pile of mirrored packages), what is the point?<a href="#t13:17" class="time">13:17</a>
kanaripin fact the .jigdo also has an md5sum for the .template<a href="#t13:17" class="time">13:17</a>
* poelcat would vote for removing the part of about not hosting isos on mirrors as the benefit of this feature<a href="#t13:18" class="time">13:18</a>
f13other than yet another (confusing) method to download the release?<a href="#t13:18" class="time">13:18</a>
jwbi like the CD set thing<a href="#t13:18" class="time">13:18</a>
jeremyf13: I think that's perhaps just a matter of having get-fedora even clearer than it is now (it's a lot better than previous releases, but could still be better :)<a href="#t13:19" class="time">13:19</a>
kanaripf13, for releases there may not be much of a point for users that do not have piles of files already, although jigdo can be made to thread and/or use http pipelining<a href="#t13:19" class="time">13:19</a>
f13so... if there is not much of a point....<a href="#t13:19" class="time">13:19</a>
kanaripcompared to torrent however that'll not be much of an advantage<a href="#t13:20" class="time">13:20</a>
jwbf13, jigdo seems like a good compromise for providing templates for isos we don't want to carry<a href="#t13:20" class="time">13:20</a>
jwbe.g. CD sets<a href="#t13:20" class="time">13:20</a>
f13jwb: what isos, besides live, are we not going to carry?<a href="#t13:20" class="time">13:20</a>
nottingas a way to do cd sets, or everything<a href="#t13:20" class="time">13:20</a>
jwbright<a href="#t13:20" class="time">13:20</a>
jwbCD sets, everything, etc<a href="#t13:20" class="time">13:20</a>
f13jwb: I'm pretty sure the board is going to ask us to do CD sets for 9<a href="#t13:20" class="time">13:20</a>
f13so that just leaves Everything.<a href="#t13:20" class="time">13:20</a>
kanaripjwb, and for people that do have piles of files already, having a .template for the DVD is another ~24 MB but saves a lot, lot of time for the client<a href="#t13:21" class="time">13:21</a>
jwbkanarip, sure that too.  smaller benefit to most people, but still valid<a href="#t13:21" class="time">13:21</a>
poelcati think it is worth doing, but would not advertise as part of get-fedora, at least not for F9<a href="#t13:22" class="time">13:22</a>
f13jwb: if we're not hosting the Everything cd sets, where would the jigdo template for them come from?<a href="#t13:22" class="time">13:22</a>
poelcatfor the people that can take advantage of it, is the burden on rel-eng too high to justify doing?<a href="#t13:22" class="time">13:22</a>
jwbf13, people could create them and provide the templates?<a href="#t13:23" class="time">13:23</a>
f13poelcat: I'm still trying to understand what the burden is.<a href="#t13:23" class="time">13:23</a>
f13jwb: where would we put them?  Changes to the release tree are costly and should be avoided if at all possible<a href="#t13:23" class="time">13:23</a>
poelcatf13: fwict creating the template<a href="#t13:23" class="time">13:23</a>
jwbf13, i'm not following<a href="#t13:24" class="time">13:24</a>
f13poelcat: but if we're not creating teh /content/ that the template would be useful for...<a href="#t13:24" class="time">13:24</a>
kanarippoelcat, maybe i can answer that... partly<a href="#t13:24" class="time">13:24</a>
f13jwb: other people make content and provide templates, where do we /put/ those templates?<a href="#t13:24" class="time">13:24</a>
jwbreleases/N/jigo?<a href="#t13:24" class="time">13:24</a>
jwbi dunno<a href="#t13:24" class="time">13:24</a>
jwbthat seems like a fairly small detail<a href="#t13:24" class="time">13:24</a>
kanarippoelcat, creating the files for an Everything spin including CD, DVD and DVD Dual Layer media in one run takes about 6 hours<a href="#t13:25" class="time">13:25</a>
jwbmaybe i'm oversimplifying things<a href="#t13:25" class="time">13:25</a>
kanariphow about in the $arch/ directory next to iso/ and os/, jigdo/<a href="#t13:25" class="time">13:25</a>
poelcatf13: i was thinking about doing them myself and putting them on fedorapeople page, but it would make more sense to have them in a central place<a href="#t13:25" class="time">13:25</a>
kanaripthese templates are still arch specific really<a href="#t13:25" class="time">13:25</a>
jwbkanarip, that's actually what i meant, yes<a href="#t13:25" class="time">13:25</a>
f13kanarip: are you going to have the templates ready when we release it to mirrors?<a href="#t13:26" class="time">13:26</a>
f13kanarip: often times mirrors will sync release one, and never again.  Getting content changed in the release tree is not trivial.<a href="#t13:26" class="time">13:26</a>
kanaripf13, i can<a href="#t13:26" class="time">13:26</a>
kanaripf13, i understand<a href="#t13:26" class="time">13:26</a>
kanaripf13, for Fedora 9 (and t1-3) however including the jigdo file and templates would be relatively easy though, right?<a href="#t13:27" class="time">13:27</a>
f13again, it has to be ready at release time<a href="#t13:28" class="time">13:28</a>
kanaripas long as the releases/ tree doesn't change during a release, right?<a href="#t13:28" class="time">13:28</a>
kanaripright, i understand<a href="#t13:28" class="time">13:28</a>
f13will you have Everything templates ready at release time?<a href="#t13:28" class="time">13:28</a>
f13or rather, ready the 7~ days before release when we hand content off to be synced?<a href="#t13:29" class="time">13:29</a>
kanaripif i can get some pre-release access to the released tree somehow, yes<a href="#t13:29" class="time">13:29</a>
EvilBobonce the ISOs are ready we can do the jigdo dance on them<a href="#t13:29" class="time">13:29</a>
f13is there a reason the templates have to be on the mirrors or in the release tree?  Couldn't they be offered up elsewhere?<a href="#t13:30" class="time">13:30</a>
f13since we're not going to advertise it on get-fedora...<a href="#t13:30" class="time">13:30</a>
jwbthey could<a href="#t13:30" class="time">13:30</a>
kanaripf13, is that the target? complete Everything/ Spins?<a href="#t13:30" class="time">13:30</a>
kanaripf13, no reason to have the files in the release tree<a href="#t13:30" class="time">13:30</a>
kanaripf13, is just as good<a href="#t13:30" class="time">13:30</a>
kanaripanywhere actually<a href="#t13:31" class="time">13:31</a>
f13kanarip: well, if the board comes back and says they want CD versions of the Fedora spin, that just leaves Everything to worry about.<a href="#t13:31" class="time">13:31</a>
jwbwhy would the board do that?<a href="#t13:32" class="time">13:32</a>
kanaripbecause there is a number of users that wants the CD's, as the numbers of downloads from Fedora Unity spins have shown<a href="#t13:32" class="time">13:32</a>
f13jwb: because they feel there is still enough want/desire for that media, and if the project is going to produce tha tmedia and make it easily available, it should go through the same compose/QA processes that the other bits of the release goes though.<a href="#t13:33" class="time">13:33</a>
Southern_Gentlemjwb because from our respins 20% of the downloads have been cd<a href="#t13:33" class="time">13:33</a>
jwbso rel-eng was wrong<a href="#t13:33" class="time">13:33</a>
jwbthis is good.  means we learned something<a href="#t13:33" class="time">13:33</a>
f13jwb: releng + the board.<a href="#t13:34" class="time">13:34</a>
f13jwb: not doing CD media wasn't our decision alone<a href="#t13:34" class="time">13:34</a>
jwbsure<a href="#t13:34" class="time">13:34</a>
jwbfedoraproject was wrong<a href="#t13:34" class="time">13:34</a>
jwb:)<a href="#t13:34" class="time">13:34</a>
kanarip+, now we have a way to potentially decrease the overhead these extra images take on mirror disc space<a href="#t13:35" class="time">13:35</a>
Southern_Gentlemjwb no sense beating a dead horse<a href="#t13:35" class="time">13:35</a>
jwbSouthern_Gentlem, wasn't<a href="#t13:35" class="time">13:35</a>
kanaripf13, concerning the footprint on the mirrors btw... i was wondering why most of them still carry FC1-FC5, is that ever going to expire?<a href="#t13:35" class="time">13:35</a>
f13kanarip: except that I don't feel comfortable /at/ /all/ not having the actual content on mirrors.<a href="#t13:35" class="time">13:35</a>
f13kanarip: it'll likely expire when the master expires it.<a href="#t13:36" class="time">13:36</a>
jwbf13, some of the big mirrors actually run with those older versions<a href="#t13:36" class="time">13:36</a>
kanaripf13, i understand, as i said none of this is mutually exclusive with current distribution methods<a href="#t13:36" class="time">13:36</a>
kanaripi wouldn't want that either<a href="#t13:36" class="time">13:36</a>
kanaripso what are we planning to do? release F9t1 with jigdo, after it's released create the jigdo files and publish them somewhere?<a href="#t13:39" class="time">13:39</a>
f13it seems to me that if we're hosting the jigdo files outside of the tree, they could be produced at any time right?<a href="#t13:39" class="time">13:39</a>
f13by any body?<a href="#t13:39" class="time">13:39</a>
kanaripat any time, right<a href="#t13:40" class="time">13:40</a>
jwbyes, as long as the content they're created against doesn't change<a href="#t13:40" class="time">13:40</a>
kanaripby anybody, right<a href="#t13:40" class="time">13:40</a>
kanaripjwb, +1<a href="#t13:40" class="time">13:40</a>
jwbi liked the idea of having them hosted on<a href="#t13:40" class="time">13:40</a>
f13so then really, as far as releng is concerned, it's just business as usual no?  We produce trees/isos, put them in expected places, somebody can come along and make jigdo files for them and add it to spins.fp.o<a href="#t13:41" class="time">13:41</a>
kanaripf13, yes<a href="#t13:41" class="time">13:41</a>
jwbi think so.  have to do the board approval dance?<a href="#t13:41" class="time">13:41</a>
jeremyshouldn't be any need<a href="#t13:42" class="time">13:42</a>
poelcatin time though, if demand is high enough, mabye tag onto the end of the compose process by releng and then everything stays in sync and is automated?<a href="#t13:42" class="time">13:42</a>
jwbjeremy, oh true<a href="#t13:42" class="time">13:42</a>
jeremypoelcat: yeah.  this lets us see what the real demand with a lower-weight process first<a href="#t13:42" class="time">13:42</a>
poelcatwow, that makes too much sense :)<a href="#t13:43" class="time">13:43</a>
jeremyI know.  crazy, isn't it?<a href="#t13:43" class="time">13:43</a>
f13Proposal:  Jigdo feature owner works on getting up spins.fp.o to be able to host jigdo files.  Files are created out of band initially and hosted there.<a href="#t13:44" class="time">13:44</a>
f13+1<a href="#t13:44" class="time">13:44</a>
jwb+1<a href="#t13:44" class="time">13:44</a>
jeremy+1<a href="#t13:45" class="time">13:45</a>
spot+1<a href="#t13:45" class="time">13:45</a>
kanarip+1<a href="#t13:45" class="time">13:45</a>
notting+1<a href="#t13:45" class="time">13:45</a>
jwbpoelcat, ?<a href="#t13:46" class="time">13:46</a>
poelcat+1<a href="#t13:46" class="time">13:46</a>
Southern_Gentlem+1<a href="#t13:46" class="time">13:46</a>
f13Southern_Gentlem: hey now...<a href="#t13:47" class="time">13:47</a>
EvilBoblol<a href="#t13:47" class="time">13:47</a>
f13that's the majority of us, looks like it passes.<a href="#t13:47" class="time">13:47</a>
Southern_Gentlemfyi <a href=""></a> you can see the demand for FU f8 cd jigdo<a href="#t13:48" class="time">13:48</a>
-!- f13 changed the topic of #fedora-meeting to: Fedora Release Engineering - Goals for F9 release<a href="#t13:48" class="time">13:48</a>
f13Last week I had asked ya'll to think up some goals for the release cycle, so that we can file milestones in Trac to track progress and deadlines.<a href="#t13:48" class="time">13:48</a>
poelcatDoes gcc4.3 fit in here?<a href="#t13:48" class="time">13:48</a>
f13Did anybody come up with anything?  I know I failed to file my signing server stuff.<a href="#t13:48" class="time">13:48</a>
jeremypoelcat: it's a release feature, not anything rel-eng specific<a href="#t13:49" class="time">13:49</a>
jwbi have a few things that are sort of rel-eng related<a href="#t13:49" class="time">13:49</a>
poelcatjeremy: mass rebuild?<a href="#t13:49" class="time">13:49</a>
jwbbut it depends on the outcome of other discussions<a href="#t13:49" class="time">13:49</a>
f13poelcat: I'm thinking more regarding infrastructure changes or policy changes during the release.  LIke introducing a signing server<a href="#t13:50" class="time">13:50</a>
jwbf13, question for you... can the arch teams use trac as well?<a href="#t13:50" class="time">13:50</a>
* poelcat withdraws his suggestion and holds for later<a href="#t13:50" class="time">13:50</a>
f13jwb: our rel-eng trac instance, or their own instance on fedora hosted projects?<a href="#t13:51" class="time">13:51</a>
jwbeither.  though rel-eng would obviously be simpler overall<a href="#t13:51" class="time">13:51</a>
f13I don't see a reason to deny that.<a href="#t13:51" class="time">13:51</a>
poelcatdoes "secondary arch" project already have a trac instance?<a href="#t13:52" class="time">13:52</a>
f13poelcat: I don't think so<a href="#t13:52" class="time">13:52</a>
jwbpoelcat, what secondary arch project?<a href="#t13:52" class="time">13:52</a>
jwb:)<a href="#t13:52" class="time">13:52</a>
poelcatlol<a href="#t13:52" class="time">13:52</a>
poelcatjwb: i keep hearing things are happening in different places<a href="#t13:52" class="time">13:52</a>
jwbf13, reason i ask is that i would think the 2ndary arch teams might want to "clone" some of the milestones you've already created<a href="#t13:52" class="time">13:52</a>
f13jwb: yeah, I don't know of a good way to do that in Trac.<a href="#t13:53" class="time">13:53</a>
jwbok, well new ones are fine too<a href="#t13:53" class="time">13:53</a>
f13I'd have to get a list of names to add the trac admins for this instance, but should be OK<a href="#t13:54" class="time">13:54</a>
jwbok<a href="#t13:55" class="time">13:55</a>
jwbwe'll work that out later<a href="#t13:55" class="time">13:55</a>
f13well... in the interest of actually getting something done today, is there anything else anybody would like to talk about?<a href="#t13:55" class="time">13:55</a>
-!- f13 changed the topic of #fedora-meeting to: Fedora Release Engineering - Open Discussion<a href="#t13:55" class="time">13:55</a>
jwbdo we need to run the jidgo thing by fesco?<a href="#t13:56" class="time">13:56</a>
f13jwb: it's a proposed Feature<a href="#t13:57" class="time">13:57</a>
f13jwb: therefor it'll go through FESCo<a href="#t13:57" class="time">13:57</a>
jwboh, right<a href="#t13:57" class="time">13:57</a>
jwbsorry<a href="#t13:57" class="time">13:57</a>
poelcatf13: do we need a mass rebuild for gcc4.3 ?<a href="#t13:57" class="time">13:57</a>
f13poelcat: I haven't looked closely at the feature page, but we'll deal with that when it comes time to review all features<a href="#t13:57" class="time">13:57</a>
poelcatf13: I'm not sure if their is a feature page<a href="#t13:58" class="time">13:58</a>
* poelcat can't spell<a href="#t13:58" class="time">13:58</a>
f13poelcat: until there is, I'm not going to plan for a mass rebuild (:<a href="#t13:59" class="time">13:59</a>
poelcatf13: :)  this should get interesting <a href="#t13:59" class="time">13:59</a>
poelcatwhen i asked Uli he said "it is just a complier" :)<a href="#t13:59" class="time">13:59</a>
f13poelcat: point him at the part of the feature planning where "would involve coordinating with large sets of other packages"<a href="#t14:00" class="time">14:00</a>
f13or that concept<a href="#t14:00" class="time">14:00</a>
* poelcat completely agrees<a href="#t14:00" class="time">14:00</a>
poelcati'll work on this a little more then<a href="#t14:00" class="time">14:00</a>
jwbpoelcat, why did you ask Uli?<a href="#t14:01" class="time">14:01</a>
poelcatjwb: i'm having internal meetings with each of the groups to see what they are thiking about or planning for F9<a href="#t14:02" class="time">14:02</a>
jwbisn't jakub the gcc maintainer?<a href="#t14:02" class="time">14:02</a>
poelcatjwb: correct<a href="#t14:02" class="time">14:02</a>
jwbdid you ask jakub?<a href="#t14:02" class="time">14:02</a>
poelcatjwb: i'm going to talk to him<a href="#t14:03" class="time">14:03</a>
jwbpoelcat, ok.  there are certain things that could benefit from a rebuild with 4.3<a href="#t14:03" class="time">14:03</a>
jwbbut i'll refrain from commenting on that since it wouldn't make sense overall<a href="#t14:03" class="time">14:03</a>
f13if there's nothing else...<a href="#t14:05" class="time">14:05</a>
-!- f13 changed the topic of #fedora-meeting to: Channel is used by various Fedora groups and committees for their regular meetings | Note that meetings often get logged | For questions about using Fedora please ask in #fedora | See ttp:// for meeting schedule<a href="#t14:06" class="time">14:06</a>

Generated by 2.3 by Marius Gedminas - find it at!