From mats.d.wichmann at intel.com Fri May 27 14:50:46 2005 From: mats.d.wichmann at intel.com (Wichmann, Mats D) Date: Thu Jul 12 13:07:53 2007 Subject: [Lsb-sc] Action Required: lsbinstall Message-ID: We've discussed the lsbinstall issue backwards and forwards and it seems it's not any more comfortable than it was when we started. Here are some key points: * the LSB badly needs an lsbinstall to allow installations to be complete (as things stand now, certain applications will require an administrator to take additional steps to make it ready for use. This may cause some developers to decide it's not worth pursuing an LSB port, or they may find it acceptable to document the steps. * it's not existing practice (in the normal case, these things are done through the package manager, but several operations are not allowed, or locations not specified, so LSB applications don't have the same luxury) * the spec and implementation are not cooked yet: even the last few days' efforts have caused more ideas to surface, and the spec has changed. finishing it will take time, and that has a cost * as noted in a recent addition to bug 795, this is substantial new functionality with respect to the ISO ballot, and it could be enough for a participant to demand a new ballot as a result (this is speculation, but they do have that right) * after lsbinstall sample/tests are finished, sufficient time will need to be allowed for implementations to take it up and adapt to their environment So after chewing on this for a while, I'm going forward a proposal that was made to me earlier today as a request for a steering committee vote. With a 3-day weekend coming up (for the US), I wanted to send this out now to give people some time to think about it; we need to close on this very quickly - by Tuesday, I'd say. 1. Continue on the current implementation plan: add lsbinstall, finish the sample, write the tests 2. Continue on the current implementation plan, as in (1), but scale back the target to only three functions: profile, service, and inet (*) 3. Drop lsbinstall from LSB 3. This does mean there can't be a (core) lsbinstall until 4.0. (*) More veriage on choice 2: These are the three for which there seem no reasonable workarounds. This would drop init (which has existing install_initd), man (which mostly has needs that can be covered by profile), and cron (which has a sufficient specification in the sysinit chapter now that packages can install directly to LSB-specified places). Having the command in some form allows developing modules post-3.0 to extend lsbinstall, it's harder to extend a core command in an optional module if the command isn't in core. Choice 2 reduces the magnitude of remaining work on the lsbinstall suite but certainly leaves plenty to do. From mats.d.wichmann at intel.com Fri May 27 15:02:28 2005 From: mats.d.wichmann at intel.com (Wichmann, Mats D) Date: Thu Jul 12 13:07:53 2007 Subject: [Lsb-sc] Action Required: lsbinstall Message-ID: Rereading this bit, it came out very confusing, sorry. >* it's not existing practice (in the normal > case, these things are done through the > package manager, but several operations are > not allowed, or locations not specified, so LSB > applications don't have the same luxury) It's not existing practice because (a) it doesn't exist at all and (b) at an individual distro level, the problem is solved by using the package manager so there hasn't been incentive to develop a cross-distro tool. But (b) doesn't work for the LSB space, unless each of the items in question is fully specified in the LSB. From mats.d.wichmann at intel.com Sat May 28 21:37:31 2005 From: mats.d.wichmann at intel.com (Wichmann, Mats D) Date: Thu Jul 12 13:07:53 2007 Subject: [Lsb-sc] Update on lsbinstall vote: Message-ID: Nick: option 3 (drop lsbinstall) Stuart: no vote Chris: no vote Marvin: no vote Matt: no vote Andrew: no vote Mats: no vote From heffler at us.ibm.com Sun May 29 12:35:16 2005 From: heffler at us.ibm.com (Marvin Heffler) Date: Thu Jul 12 13:07:53 2007 Subject: [Lsb-sc] Action Required: lsbinstall In-Reply-To: Message-ID: My vote would be for option 3 where we leave lsbinstall out of LSB 3. It's way too risky to be adding something to the spec that doesn't even exist yet and where there are still lots of open questions. We need to do whatever we can to make LSB 3.0 certification available on July 1. Adding lsbinstall not only delays completion of the written spec, but it will most likely delay starting certification. Additionally, it will add more work for distros trying to get certified. At this point we still need to lean toward the distros for spec and certification issues due to lack of ISVs seeking complete LSB certification. It seems to me most ISVs will meet as much of the spec as possible, but they will stop short of actual certification. Adding lsbinstall is still a good idea, but it is not good enough to delay releasing LSB 3.0. Regards, Marvin Heffler Linux Standard Base IBM Linux Technology Center 11501 Burnet Road, Zip 905-7A017 Austin, TX 78758 (512) 838-0953 T/L 678-0953 lsb-sc-bounces@base3.freestandards.org wrote on 05/27/2005 04:50:46 PM: > > We've discussed the lsbinstall issue backwards > and forwards and it seems it's not any more > comfortable than it was when we started. > > Here are some key points: > > * the LSB badly needs an lsbinstall to allow > installations to be complete (as things stand > now, certain applications will require an > administrator to take additional steps to > make it ready for use. This may cause some > developers to decide it's not worth pursuing > an LSB port, or they may find it acceptable > to document the steps. > > * it's not existing practice (in the normal > case, these things are done through the > package manager, but several operations are > not allowed, or locations not specified, so LSB > applications don't have the same luxury) > > * the spec and implementation are not cooked > yet: even the last few days' efforts have > caused more ideas to surface, and the spec > has changed. finishing it will take time, > and that has a cost > > * as noted in a recent addition to bug 795, > this is substantial new functionality with > respect to the ISO ballot, and it could be > enough for a participant to demand a new > ballot as a result (this is speculation, > but they do have that right) > > * after lsbinstall sample/tests are finished, > sufficient time will need to be allowed for > implementations to take it up and adapt to > their environment > > > So after chewing on this for a while, I'm going > forward a proposal that was made to me earlier > today as a request for a steering committee vote. > With a 3-day weekend coming up (for the US), I > wanted to send this out now to give people some > time to think about it; we need to close on this > very quickly - by Tuesday, I'd say. > > > 1. Continue on the current implementation plan: > add lsbinstall, finish the sample, write the tests > > 2. Continue on the current implementation plan, > as in (1), but scale back the target to only three > functions: profile, service, and inet (*) > > 3. Drop lsbinstall from LSB 3. This does mean > there can't be a (core) lsbinstall until 4.0. > > > > (*) More veriage on choice 2: > These are the three for which there seem no reasonable > workarounds. This would drop init (which has existing > install_initd), man (which mostly has needs that can > be covered by profile), and cron (which has a sufficient > specification in the sysinit chapter now that packages > can install directly to LSB-specified places). Having > the command in some form allows developing modules > post-3.0 to extend lsbinstall, it's harder to extend a > core command in an optional module if the command isn't > in core. Choice 2 reduces the magnitude of remaining > work on the lsbinstall suite but certainly leaves > plenty to do. > > _______________________________________________ > Lsb-sc mailing list > Lsb-sc@mail.freestandards.org > http://mail.freestandards.org/mailman/listinfo/lsb-sc -------------- next part -------------- An HTML attachment was scrubbed... URL: http://lists.linux-foundation.org/pipermail/lsb-sc/attachments/20050529/fa18fee5/attachment.htm From anderson at netsweng.com Sun May 29 12:57:58 2005 From: anderson at netsweng.com (Stuart Anderson) Date: Thu Jul 12 13:07:53 2007 Subject: [Lsb-sc] Action Required: lsbinstall In-Reply-To: References: Message-ID: On Sun, 29 May 2005, Marvin Heffler wrote: > My vote would be for option 3 where we leave lsbinstall out of LSB 3. It's > way too risky to be adding something to the spec that doesn't even exist > yet and where there are still lots of open questions. We need to do > whatever we can to make LSB 3.0 certification available on July 1. So who is goig to be ready to certify on July 1? >From what we know, none of the distros will have their release cycles in the right phase to be able to certify at that time. Stuart Stuart R. Anderson anderson@netsweng.com Network & Software Engineering http://www.netsweng.com/ 1024D/37A79149: 0791 D3B8 9A4C 2CDC A31F BD03 0A62 E534 37A7 9149 From cyeoh at samba.org Sun May 29 21:40:28 2005 From: cyeoh at samba.org (Christopher Yeoh) Date: Thu Jul 12 13:07:53 2007 Subject: [Lsb-sc] Action Required: lsbinstall In-Reply-To: References: Message-ID: <17050.39228.238244.656142@gargle.gargle.HOWL> At 2005/5/27 14:50-0700 Wichmann, Mats D writes: > 1. Continue on the current implementation plan: > add lsbinstall, finish the sample, write the tests > > 2. Continue on the current implementation plan, > as in (1), but scale back the target to only three > functions: profile, service, and inet (*) > > 3. Drop lsbinstall from LSB 3. This does mean > there can't be a (core) lsbinstall until 4.0. I'd go for option 3 - I don't think we should be including features into the specification until they have been implemented. Preferably they would also have been rolled out in real life on a distribution too, but would be happy for that restriction to be weakened if there was a good test suite available. Chris -- cyeoh@au.ibm.com IBM OzLabs Linux Development Group Canberra, Australia From ajosey at rdg.opengroup.org Sun May 29 22:43:22 2005 From: ajosey at rdg.opengroup.org (Andrew Josey) Date: Thu Jul 12 13:07:53 2007 Subject: [Lsb-sc] Action Required: lsbinstall In-Reply-To: Stuart Anderson's message as of May 29, 3:57pm. References: Message-ID: <1050530054322.ZM5415@skye.rdg.opengroup.org> According to FSG there are a number of folks pushing for LSB3.0 certification. A bigger issue is that this puts the ISO approval in question, since lsbinstall is not something that was in the ISO doc, it could lead to 3.0 needing a reballot regards Andrew On May 29, 3:57pm in "Re: [Lsb-sc] Action ", Stuart Anderson wrote: > On Sun, 29 May 2005, Marvin Heffler wrote: > > > My vote would be for option 3 where we leave lsbinstall out of LSB 3. It's > > way too risky to be adding something to the spec that doesn't even exist > > yet and where there are still lots of open questions. We need to do > > whatever we can to make LSB 3.0 certification available on July 1. > > So who is goig to be ready to certify on July 1? > From mats.d.wichmann at intel.com Mon May 30 04:44:31 2005 From: mats.d.wichmann at intel.com (Wichmann, Mats D) Date: Thu Jul 12 13:07:53 2007 Subject: [Lsb-sc] Action Required: lsbinstall Message-ID: >At 2005/5/27 14:50-0700 Wichmann, Mats D writes: >> 1. Continue on the current implementation plan: >> add lsbinstall, finish the sample, write the tests >> >> 2. Continue on the current implementation plan, >> as in (1), but scale back the target to only three >> functions: profile, service, and inet (*) >> >> 3. Drop lsbinstall from LSB 3. This does mean >> there can't be a (core) lsbinstall until 4.0. > >I'd go for option 3 - I don't think we should be including features >into the specification until they have been implemented. Preferably >they would also have been rolled out in real life on a distribution >too, but would be happy for that restriction to be weakened if there >was a good test suite available. If proceeding with lsbinstall, sample implementation and test are absolutely release criteria. From cyeoh at samba.org Mon May 30 05:37:36 2005 From: cyeoh at samba.org (Christopher Yeoh) Date: Thu Jul 12 13:07:53 2007 Subject: [Lsb-sc] Action Required: lsbinstall In-Reply-To: References: Message-ID: <17051.2320.367214.433581@gargle.gargle.HOWL> At 2005/5/30 04:44-0700 Wichmann, Mats D writes: > > > >I'd go for option 3 - I don't think we should be including features > >into the specification until they have been implemented. Preferably > >they would also have been rolled out in real life on a distribution > >too, but would be happy for that restriction to be weakened if there > >was a good test suite available. > > If proceeding with lsbinstall, sample implementation > and test are absolutely release criteria. Would they be released before the specification is released? Even then I would be a bit nervous about standardising on something that has not yet actually been used in real life yet. Chris -- cyeoh@au.ibm.com IBM OzLabs Linux Development Group Canberra, Australia From mats.d.wichmann at intel.com Mon May 30 12:47:03 2005 From: mats.d.wichmann at intel.com (Wichmann, Mats D) Date: Thu Jul 12 13:07:53 2007 Subject: [Lsb-sc] Updating lsbinstall vote again Message-ID: 1: keep in 3.0 2: keep, reduced functionality 3: drop from 3.0 Nick: option 3 Stuart: no vote yet Chris: option 3 Marvin: option 3 Matt: no vote yet Andrew: option 3 Mats: no vote yet