From Mageia wiki
Jump to: navigation, search


Distribution release cycle discussion

We have seen quite a long discussion on mageia-discuss list about how the distribution release and update processes should be (rolling, not rolling, what is rolling, etc.).


So here's a proposal to go forward. Each of you write down an article (on a separate page, on a blog, anywhere on the Internet, and link it below) stating:


  • who you are,
  • how you use your system,
  • how you would define you (user, developer, packager, other) relative to the system you use,
  • how you would like the releases/updates for the system and software applications to happen for you and why
  • how you imagine this could be done and what resource it would involve
  • how you believe others may not agree with this and why
  • what is a rolling distribution for you?
  • what is not a rolling distribution for you?

It does not matter if you cannot answer all these questions, at least, point out what you can, and just say when you can't.


In other, more specific words, please describe:


  • the release process (i.e., what you put on disc ISO)
  • based on what criteria
    • time based?
    • regular snapshot?
    • features based?
  • how to decide what packages are upgraded day to day
  • what criteria for what kind of upgrade (version upgrade, patch upgrade)
  • what procedure for upgrade (test stage, direct upgrade?)
  • for how long
  • how are updates built (i.e., against what tree)
  • what packages are not upgraded, unless there is a release, if there is one
  • what criteria warrant the stability of this component
  • how do we decide it will be upgraded

What matters is that you give a structured, easy to read and understand, document explaining what you know (and what you don't know) about a distribution release process and what a rolling distribution is for you (and how to accomplish it)


Point of views

Please add a link to your article, be it hosted on this wiki or on any outside server:


Other, shorter POVs

  • phgphd has this opinion. Why is it that when some party starts anew, they have to rehash the same old same old questions as if things have to be changed. You forked from Mandriva, so just keep as much continuity as possible. It was already an excellent distribution, just keep up the excellent work.
  • I concur with phgphd - — J.P. Pasnak 2010/10/09 04:04

goom's point of view

First of all, the choice of fixed or rolling release is just a simple technical choice since it will affect the way Mageia will be presented, advertised, how Mageia will be among other distributions, how Mageia will attract developers and users, how to develop business on Mageia.


Who's doing rolling distros ? According to wikipedia http://en.wikipedia.org/wiki/Rolling_release Gentoo Linux, Arch Linux, Sabayon Linux, PCLinuxOS, aptosid, Foresight Linux and Linux Mint Debian Edition (LMDE) are examples of true rolling releases, few among hundreds distributions. From what i have read from users (and former Mandriva users, some of them leaving mdv after the 2009 release and no longer support of KDE 3.5) PCLinuxOS seems quite good even if there is not so much developers. Some wishes in Debian to have a rolling distro based on testing branch ( http://carlchenet.wordpress.com/2010/09/11/et-si-debian-testing-devenait-constamment-utilisable-le-projet-cut/ ; http://fgallaire.flext.net/debian-apprendre-archlinux/ , http://kitenet.net/~joey/code/debian/cut/ . I particularly appreciate the introduction of Joey Hess « It's time to discard the old boolean model where a version of Debian was either unreleased, and thus constantly changing and potentially unfit for use at any given time, or released, and thus on the long slide to obselescense. So far, we've been working around and increasingly invalidating this model in bits and peices » since it could be used also for Mandriva. I would be nice to use Mageia to have new foundations, even if they are different from what Mandriva was doing. I foresee a risk on keeping on the same way of Mandriva. Times have changed, and still change, the target audience change, Linux is more and more opened for users without any special knowledge in computing (AUFML bank manager at the Banque Postale has get his computer installed with Ubuntu, he is happy with it and do not have any special knowledge, just an anecdote that tell the truth)


For me, the philosophy of a rolling distro is « Installe once, update many ». The first point I want to emphasis if that installing an OS or a distribution is not itself a goal (except for some geeks) since the real goal is to use software to do (whatever you want). So we must do our possible to avoid the installation of an OS and for a Linux distribution, just reduce it to one installation. This will be a positive point we could market during install-parties “we will install Mageia, then you will just have to update”


« Install once, update many » allows to deal with obsolescence by not dealing with it. With the actual model of Mandriva, 18 months of support it means 3 releases to maintain at the same time plus cooker. 18 months is to short to do business. Many companies are still using Windows XP and start to look to change to Windows 7 which will be tested for several months. With a 18 months cycle release, there is no possibility to enter the market. A 3 years life time start to be interesting (don't forget that big structures take 2 to 3 years to change an OS, if this is the life time of the OS, the market is closed). A life time of 3 years with a release every 6 months it means 6 releases to maintain ! Life time is a marketing technique to sell, very use in sofware world. Mandriva uses a 6 months cycle to create a keen interest every 6 months, which was important for sales. With Mageia being community-driven distribution there is no need to have a 6 month rythm of release. So of course we will not communicate every 6 months about release. I would just answer “so what ?” we will communicate in a different way, more, it will be a feature of Mageia to communicate differently. There are many opportunities to communicate, there are so many free software projects, we could communicate for the latest release of KDE, the new Gnome, XFCE, kernel, Xorg, … Do not deal with obsolescence would also allow Mageia to have third part software that just work on Mageia, no just work on Mageia 1 but not on Mageia 1.5 or need Mageia 2.2, we can just “sell” a “Mageia compatible”, good for end-users, good for third part software company. And no need to have must stress on biannual release, talks about including or not the very last version of a software 4.6.32 that brings many things while freeze is done


Rolling distro would avoid to have on Mageia release with a software in version 3.1.2, the next release with the same software in version 3.2.4, and the following release with software in version 3.3.1 without any clear and valuable explanation. If this seems not clear, let's check OOo version in Mandriva 2009.1, 2010, 2010.1, so we have OOo 3.1.1-0.3, OOo 3.1.1-2.5 et OOo 3.2-4. During an install party i can't explain why there are those differences (especially when latest rpms are available on OOo website). Here we just can see clearly limits of a fixed release, the obsolescence after 1 year (even not the life time of a computer). Other example, digikam 1.4 that need KDE 4.5, but there is no official KDE 4.5 for Mandriva 2010.1 (nor for 2010 and 2009.1) so an user can't have digikam 1.4 that comes with improvement. Neoclust has done rpms for KDE 4.5 but there are not official so we may have digikam 1.4 if updates were provided. So is it possible to update or not ? No clear answer, it annoys me as an user


I am all the more bothered that very often the answer is the opposition between stability and novelty. For me this is a void argument. Let's take again the example of Digikam. Version 1.4 comes with bugs correction from Digikam 1.3 and also comes with new functionnalities (that can bring new bugs) so which choice do the user have ? Digikam 1.3 with bugs (and no quite new) or Digikam 1.4 with some corrected bugs but maybe new bugs due to new functionnalities ? This is almost the case with most free softwares, the new version corrects bugs from former version (nice !) and add new functionnalities (cool !) with the risk of new bugs (bbouuuhhh). Those news bugs will be reported only by users having the newest version and will be corrected only if there is a report so it would be more efficient to have people using the last version (and not a previous one that will have no or nearly no bug correction. If you are using Digikam 1.2 and you report a bug today, the first answer wil be upgrade to Digikam 1.4 (well 1.5 since monday !), that's a good reason to have a continuous evolution, a better collaboration upstream


Few exception, Scribus is one, stable version is 1.3.3.14 (stable branch is 3-4 years old), unstable version is currently 1.3.8 (but very usable), and development versin is currently 1.3.9svn which will give unstable 1.3.9. 1.3.3.x branch are bug corrections only, while unstable version are bug correction and new functions. According to the way free sofware are developped, a new version comes with bug corrections and new functions at the same time, we cannot not to propose updates arguing of stability. My KDE 4.3.5 of my Mandriva 2010 will have no more bug correction nor updates, i don't think this is good now in 2010.


Stability is one of the goal of Cauldron. This is the development version, there we work on stability. We could add another step “debian like” with testing for packages “stable” that need more testing before being pushed to stable (or go back to cauldron). Rolling distro do not mean that updates are to be done the day a new software is released. It means that the new version will go to update and be available for all when it will have met QA criteria, and then stability is ok. This is a guarantee that the end-user will have the lst Gnome, KDE, Xfce, OOo, etc. it does not tell when, it can take some times.


I have the feeling that a rolling distro would have prevent what we have seen for Mandriva 2009 when stopping support for KDE 3.5 and the further withdrawal of KDE 3.5. I guess that in a rolling, packages of KDE 3.5.10 would be still available even if there is no more correction. As Mageia aims to be a community distribution, it will not have to follow the same policy as Mandriva with officially supported packages (main) and other not supported (contrib). KDE 3.5 would continue to be available and usable for those who like this DE.


A rolling distro means another view for media since backport have not needed any longer and we can have 3 versions : cauldron, testing and stable (even just 2 by deleting testing, but i think this is not a good choice, just like in Chemistry where you never go from laboratory experiments to production, there are steps to validate the scale change since there are parameters to be taken into account, the same for software, development, testing, large range availability). Using a rolling also means starting with KDE 4.1 and updating regularly to KDE 4.5.


About backport, i have the feeling that it lets think that “hum yes we could have a rolling after all” without going to end of the road (of course, if anything that is under development goes then to backport at a moment, it would become de facto a rolling


Every project of free software is desynchronized, KDE is neither in phase with Digikam, Amarok, Koffice, Gnome, GCC, Kernel, Xorg. A fixed release, with a freeze of Cauldron will always lead to such a desynchronisation. To avoid this, i guess that rolling distro is a nice way to do things. Let's allow time to stabilize softwares and not doing it every 6 months in an hurry time (to be continued …)

Retrieved from "https://wiki.mageia.org/mw-en/index.php?title=Archive:Rolling_debate&oldid=57808"