On 05.01.2012 00:16, Michael Scherer wrote: > Le mercredi 04 janvier 2012 à 20:09 +0200, Anssi Hannula a écrit : >> On 04.01.2012 19:29, Michael Scherer wrote: >>> Le mercredi 04 janvier 2012 à 16:16 +0200, Anssi Hannula a écrit : >>>> On 04.01.2012 11:54, Michael Scherer wrote: >>>>> Le mercredi 04 janvier 2012 à 11:03 +0200, Thomas Backlund a écrit : >>>>>> Anssi Hannula skrev 3.1.2012 23:05: >>>>>>> On 02.01.2012 12:21, guillomovitch wrote: >>>>>>>> Name : dsniff Relocations: (not >>>>>>>> relocatable) >>>>>>>> Version : 2.4 Vendor: Mageia.Org >>>>>>>> Release : 0.b1.1.mga2 Build Date: Mon Jan 2 >>>>>>>> 11:18:17 2012 >>>>>>>> Install Date: (not installed) Build Host: ecosse >>>>>>>> Group : Monitoring Source RPM: (none) >>>>>>>> Size : 210074 License: BSD >>>>>>>> Signature : (none) >>>>>>>> Packager : guillomovitch<guillomovitch> >>>>>>>> URL : http://www.monkey.org/~dugsong/dsniff/ >>>>>>>> Summary : Network audit tools >>>>>>>> Description : >>>>>>>> Tools to audit network and to demonstrate the insecurity of cleartext >>>>>>>> network protocols. Please do not abuse this software. >>>>>>>> >>>>>>>> guillomovitch<guillomovitch> 2.4-0.b1.1.mga2: >>>>>>>> + Revision: 189630 >>>>>>>> - drop epoch, we don't care about updating from mdv anymore >>>>>>> >>>>>>> We don't? >>>>>>> >>>>>> >>>>>> Oh yes we do. Atleast from 2010.1 >>>>> >>>>> We did for 1, not for 2 or cauldron or anything else. So as long the >>>>> package is not pushed on 1, I think we agreed that people could not care >>>>> about upgrade path from Mandriva. >>>> >>>> Well, I don't like that, IMO we should not remove upgradeability so >>>> soon, even if we won't officially support it. >>> >>> Well, if we do not officially support it, then we do not support it, >>> that's all. There is no "that's unofficially supported" or stuff like >>> that. Supported mean "we will do test and fix bug if they happen", and >>> not supported mean "we reserve our right to not do anything". >>> >>> And that's exactly what happen right now. >> >> IMO there is a level between "officially supported" and "we >> intentionally break it", which means that we advise against it but do >> not hinder people from doing it. > > Yes, there is different levels of support, obviously, since people have > different but that doesn't mean we should rely on them, or try to > officially use them. Again, saying "we support that, so we do that, and > we don't support this, so people are free to do what they want" is > simpler. > > The whole scheme of having "stuff we do", "stuff we do not promise but > try to", "stuff we do not promise and we do not try to" ( or more ) make > things less clear for everybody. Having a non uniform policy will make > things harder for newer packagers ( and for olders too ). > > We have users ( in the past ) that complained about the lack of > reliability of packages on Mandriva. And this was IMHO because we had a > policy of 'we keep everything and we say they are in a section of "maybe > supported"'. The whole message "contribs is not supported but main is" > was simple and yet, too complex to grasp ( because people didn't check > contrib/main before installing anything, )
It was not too complex, just badly implemented. The users got *no* in-GUI notification at all that contrib was unsupported (most users don't read wiki pages etc, especially if we don't link them to them). > . It was also far from the > truth because some stuff in contribs were more supported than stuff in > main, and thus we were sending mixed messages. Right. > So we should really stick on what we support, and send a simple, clear > and correct message. > > And I think we need to keep things simple to solve such issues in the > long run. I'm not advocating any change on the message sent to users, just to not break upgrade intentionally (by removing near-zero-maintentance upgrade support by dropping obsoletes/epochs that are needed for upgrade from the several last distribution versions). >>>> But anyway, this affects people doing 2010.1->mga1->mga2 as well... Or >>>> are you saying that isn't supported either, and people should do new >>>> installs?? >>> >>> We do not support upgrading mdv2010.1 rpms with rpm from mga2, so if a >>> maintainer want to remove this, he can. >>> >>> Someone doing mdv2010.1->mga1 will end with a mix of mdv2010.1 and mga1 >>> if the system is not cleaned, and that's not something we should >>> support, not more than mga X + any random repository upgrade to mga X+1 >>> >>> IE, that's not mga1 -> mga2, that's mga1 + 3rd party repo that happened >>> to work by chance to mga2. >> >> I have to strongly disagree with this. If upgrading from 2010.1 to mga1 >> is officially supported (and it is), we can't say "you can't upgrade >> your mga1 system to mga2 anymore because you have some old pkgs >> installed which we never asked you to remove" (assuming no non-mdv 3rd >> party repos here). > > First, it doesn't break the whole upgrade. > In fact, if we look carefully, people who were running non supported > software ( ie a package from Mandriva ) will still run the same > unsupported software and the same binary. And upgrade will likely work > without error messages. Because nothing requires dsniff, except its own > subpackage. > > Secondly, it didn't matter much before Guillaume uploaded dsniff. > > 1 week ago, anyone who would have upgraded to mga2 with dsniff installed > from mdv would have been in the exact same situation than now, except > nobody cared at all. And the proof that nobody cared is that nobody > pushed the rpm sooner. Would it have been pushed to 1, yes, that would > have breached what we agreed to do. But it was not pushed to 1 ( and I > would say "likely on purpose" ). So the only change with this upload is > for people installing dsniff later. I care. I also still have dozens of missing mdv packages in my TODO that I intend to import to cauldron and mga1. dsniff was one of those, though I'm not sure yet if I care enough about it to request it to submitted to mga1 updates, since I don't use it on my mga1 systems (only on cauldron). > 3rd point, the whole point of saying "we do not support this" is not to > say "we don't support, but we should still support it to some extent". > It is to be able to say "we do not support, so the maintainer can clean > it if he want". You are free to support it if you wish, but Guillaume is > also free to not support it, and choose to clean instead ( because epoch > tags are ugly ). I completely agree. However, I consider this to break MGA1->MGA2 upgrade, which *is* supported. > If we wanted to support upgrading from mdv 2010.1/2 to mga2, or > upgrading people who mix distribution packages ( be it because they do > not know, or on purpose, that's the same problem from a technical PoV ), > it should have been said much sooner. I'm fine with us not supporting 2010.1->mga2. However, I'm not fine with breaking 2010.1->mga1->mga2. And saying "it didn't completely break" while user has in his mga2 installation old packages is IMO not an option. If it was intended that 2010.1->mga1 system wouldn't be completely upgradable (all packages) to mga2, we should've made it clear when we offered the mdv2010.1 -> mga1 upgrade path. But we didn't, so we should completely support 2010.1->mga1->mga2 (with which I agree with). > I do not understand, could people tell me what did they understood we > would do when we said "we will not support upgrade this after mageia > 1" ? I understood it as 2010.1->mga2 would not necessarily work, but 2010.1->mga1->mga2 would still work fine, without old packages left, except for those for which no new versions exist in the distribution. -- Anssi Hannula