The Disaster of the Rolling Release

By Ed Hurst | Posted at 5:11 AM

I’ve always enjoyed exploring. Every time I’ve moved from one residence to another, I’ve always wandered around my new neighborhood, simply to see what was there. It’s the same with computer technology. I love poking around operating systems. Lately, one aspect of this has gotten tiring in every Open Source operating system: the rolling release. The phrase refers to the sometimes feverish effort to add new features, long before the old ones even work properly. Thus, every day sees sometimes radical changes in various projects, new features and new bloat.

This is the ancient schizophrenia between creativity and usefulness. Not everything in life has to be useful to be necessary. Decoration for no other purpose is fine, and differing tastes are seldom a simple matter of refinement and morality. The original meaning of “character” is the quality of being distinguished from others in the same category, applied primarily to inscribed letters of any given alphabet. Since God knows each of us individually and more intimately than we could know ourselves, we do best to discover what He has made in us, not conforming ourselves to the artificial constraints of someone else’s character. However, individuality does not justify destroying or harming others. So, a gal can pierce all her facial features if she likes, and I’ll support her choice no matter how repulsive I find it. But she can’t expect me to gaze upon her “beauty,” much less want to kiss her. If she also wants to interact with me, she’ll have to make some sacrifices. That’s what civilization is all about: Having a set of cultural expectations which make it possible for humans to live in close proximity without killing each other.

It applies to computer stuff, too. Purity in researching the limits of software on hardware has its place, because no one can predict how crazy ideas might serve a useful purpose later. At the same time, there has to be some useful product, something which meets the drab requirements of folks who use computers only as a tool. We can’t all worship at the shrine of the electronic data gods. Some of us have other things in life we must do to fulfill our perception of our purpose. We don’t do things to serve the computer, but use the computer to serve our needs. For those of us to whom our computer is our number one tool, we still want to get work done. Most of the time, that work is done best when we have pretty much the same tools for an extended period of time. Minor improvements are fine, but constant changes in almost every feature don’t work out well. Too much work time is absorbed in exploring all the new and wondrous changes.

It seems most coders are buried up to their eyebrows in one or two projects. They know this one project really well, and most do a really fine job of designing and adding new features to just the one project. But they are all doing this on just about every project. It’s been long admitted squashing bugs is not glamorous, not half so interesting as creating new code. It requires discipline, sometimes externally imposed, sort of like the necessity of laws to back up the customs of civilization. You can have too much law and too much enforcement, but you can also have too little; the same goes with management and government. The people toiling away on their favored projects don’t seem to notice that’s what is happening on all the projects. A Linux distribution (”distro”) is usually an effort to build a fairly single-source collection of packages which blend together as a whole. The idea is to guide the selection and tweaking of various packages according to some guiding philosophy. Get enough of those packages together, and you could easily have a completely mismatched set of moving targets. It can be quite challenging to coordinate all those packages so they tend to work well together at any given point when a new user installs what you have on hand in your distro repository.

So big is the task, many distros do it poorly, and quite a few seem to make no effort at all. That many individual packages tie themselves to a particular toolkit, or a particular desktop project, can reduce the chaos. Each toolkit/desktop tends to follow some set of rules which make it possible to build just about any package in a fairly broad window within the on-going development of the underlying toolkit. Too often, they don’t bother. The package developers may try to justify that there simply must be this or that feature based on some new addition to the underlying framework. That’s the standard myopia; that it might break every other package attached to the framework is not their concern. Some toolkits/desktops are worse than others in permitting this sort of chaotic nonsense. Equally bad is simply dropping some feature because it’s a hassle to design it into the next major reworking of the project. Users get jerked around, and all too many projects pay no attention at all to that factor; some even cop an attitude about it.

The Linux evangelists are stuck with this big albatross around their necks. There is already a pretty strong division between coders and users, and precious few people who take an interest in mediating between the two. The field is wide open, currently dominated by the business-suit guys and gals because they must have it to make any money selling Linux. The larger community keeps them at arms length, to some degree, because they lack the purity of purpose. We end up with a few highly commercialized Linux distros because these folks simply can’t trust some elements in the wider community to pay much attention to their needs. We end up with an apparent conflict, and in the middle of all this the Linux advocates bemoan one or the other side, and there haven’t been too many intelligent answers why we can’t get the rest of the world to adopt Linux.

The driving force behind adoption in terms of numbers is business and government. Bill Gates didn’t get to be a billionaire by chasing hobbyists first. He offered something which could be sold as fairly standardized, but in some way better than what somebody else offered. Once it became a must-have for business and government, it could be sold to the general public. The free-wheeling coding community does not appeal much to the suits. Rolling release to them is anathema. Incremental improvement is fine, but bug-fixing and security come first, all within a long release and upgrade cycle. There is a cost to keeping obsolete computer stuff around, but there is also a cost in terms of time and effort to adopt something new. If what’s new offers a competitive edge, it will be adopted. If it makes the artifacts of information incompatible with partners, it will not be adopted until there’s a momentum for all of them to change at once. Poking fun at all the examples of organizations violating sound principles, hiring brainless IT people, etc., does nothing to improve matters.

I enjoy writing on technology mostly because I’m trying to encourage folks who work and play on computers to learn more. For me, the people come first; they are the reason we bother with computers at all. The path to making the world a better place to live is through helping people. If our work does not help them, we are just pursuing a hobby. If we want more people to enjoy our work, it has to be useful to them. Rolling release reflects the purist hobby mindset. It won’t sell to the wider user base. All the more so when your work offers ill-considered new features, bloat, and never bothers to fix old bugs. Windows XP came out in 2001, buggy as anything MS has produced, and it took some five years before third-party efforts made it worth using, as defined by mass consumer adoption (I’m convinced Windows 2000 is still better). If what we offer is no less buggy, why should consumers bother to switch? Both KDE and GNOME projects continue to outrun usefulness on the consumer level because developers keep introducing new bugs without nearly enough quality control in fixing old ones. If we never bother to offer something really good, even if it’s not all new and shiny, industry and government won’t buy it, and it will never become the standard.

When advocates approach me about trying their favorite Linux distro or other Open Source OS, my first question these days is: Does it adhere to rolling release? If it does, I don’t have time for it any more. We have huge problems getting web standards adopted, or finding work-arounds for proprietary offerings on websites. That’s a symptom of the underlying, damnable idiocy of pushing out the next new release before bothering to fix the old. This lack of stability means there is no standard, de facto or conceptual. It also means we don’t have anything to offer the average consumer, because we have nothing to offer the institutional users. By the time we put it in our hands, it has changed three times before it gets to theirs. Only a small slice of any given population of living humans can adapt to that. Enthusiasts should stop whining about the lack of adoption and about the corruption of corporate-grade Linux distros if they can’t offer anything they recognize as better.

Ed Hurst is Associate Editor of Open for Business. He writes on a variety of subjects concerning technology and ministry.

Join the Conversation

20 comments posted so far.

Re: The Disaster of the Rolling Release

You made a lot of accurate and wellspoken points in your article. But it appears you don’t really understand what a rolling release distro is. A rolling release disto is a distro that is simply updated constantly through the package manager with the intent of breaking that horribly annoying process of a new installation every six months. And most rolling release distros don’t update to the unstable cutting edge as you seem to think. You mentioned windows xp and how it took 5 years to get straightened out. Now look at that from your position against rolling release. Xp would have been released and as it was released it would have stayed other than some minor security updates. Microsoft would have had to come out with a replacement every 6 months just like non rolling release linux distros. An advantage to rolling relase is also that any bugs or glitches appear after a distros release you can be certain they will be fixed promply, usually immediately. Unlike a set release distro where you only get security fixes no matter how much the users complain. Just btrowse the ubuntu forums after a new release if you want to see what I mean. No developers are perfect and all will occasionally slip up but at least with a rolling release distro you don’t have to wait 6 months and do a complete reinstall to maybe get a fix.

Posted by Greg - Dec 31, 2008 | 11:08 AM

Re: The Disaster of the Rolling Release

I agree with Gregs comment above. If anyone wants to try a rolling release distro then I would heartily recommend PCLinuxOS.

I have the Minime2008 version installed on my Acer laptop, which is fully updated to the latest version. Minime is a version that contains only base packages leaving you to install extra ones as required rather than getting everything as decided by the distributions developers. There is also a full cd available, PCLinuxOS2007 which contains all these packages

Basically, as Greg says, you install the released version from cd, however long ago that was, and then you open the package manager and do a complete update. This brings all your installed packages up to the latest versions in the repositories. Then every few days you repeat this process. That way your system is always up to date and if a new cd/dvd is released it will match your setup, except for maybe a newer kernel but even those are available via the normal update route.

One disadvantage of this process is after an initial install there can be quite a few packages that need to be updated. For me that amounts to 205 packages. However with a 2mb Bband connection the complete update takes less than 30 mins

Regards Dave Le Huray Guernsey

Posted by Dave Le Huray - Dec 31, 2008 | 12:48 PM

Re: The Disaster of the Rolling Release

If it were only bug fixes and security fixes, “rolling release” would be innocuous. It is a false dichotomy to say we must have either 6-month releases versus rolling release. We could just have fixes alone to a good current release for some years. From the mainstream user’s point of view, constant significant revision in increments less than 5 or 6 years means it’s just a toy, not a real tool. They never get to use the system, as they perceive it, because it’s always changing, always demanding too much attention to deal with it. We may well have the best development system there could ever be, but we will never gain wider adoption until we cater to that average consumer mindset.

Posted by Ed Hurst - Dec 31, 2008 | 12:55 PM

Re: The Disaster of the Rolling Release

This article is wordy and all over the place to the point that it kind of hard to know what you are talking about ..

first of all the person who came before me said ..a term “rolling release” is used to mean a linux distribution that you install once and keep on updating ..others distros forces users to reinstall on each new release. If microsoft windows OS was a rolling release in a linux distro sense ..then you could start with win2000, upgrade it to windows xp, vista and windows 7 when it comes out ..thats what a rolling linux distro mean reinstalling btw releases ..just updates..

what kind of an update path a distro will have will depend on how packages are packaged in that distro ..this has nothing to do with coders of respective projects and it has everything to do with packagers ..

“There is already a pretty strong division between coders and users, and precious few people who take an interest in mediating between the two” In the linux world packagers are the people who are btw coders and users and they package and present their packagers in a complete package called a distribution ..distributors are not coders ..the best they do is tweak packagers to make them fit well with other packagers they want to have and to their particular preferences ..since packagers arent coders, to expect “rolling releases

”{..} The larger community keeps them at arms length, to some degree, because they lack the purity of purpose.”

the larger community lacks purity of purpose? you need to relearn the fundamentals of FOSS world ..FOSS world does not exist to cater to people in suits exist to fulfill a purpose ..that purpose might not work well in a commercial environment but that does not mean their is no purpose or is impure ..

what exactly do you mean when you say “rolling release”?

Posted by mtz - Dec 31, 2008 | 1:48 PM

Re: The Disaster of the Rolling Release

My interpretation of a “rolling release” is a distribution that does issue updates, but there is a relatively long time between versions. But the distinction is becoming less clear as more version releases are issued unfinished, requiring a number of bug-fix downloads out-of-the-box.

Depending on ones perspective, one can credit or blame broadband for enabling this trend. But, as a bandwidth-deprived Linux user, I’m also finding maintenance increasingly time-consuming.

Posted by dialup - Dec 31, 2008 | 1:58 PM

Re: The Disaster of the Rolling Release

To mtz: Let me try that again. The hobby community keeps the suits at arm’s length because the suits are alleged to have an impure purpose. That reflects the many wordy attacks on Novell and RedHat.

The term “rolling release” means just what Arch Linux and FreeBSD both say: Keep upgrading in place without re-installing. The tragedy is the assumptions tied to that concept: You will upgrade or we will not help you. Having a 6-month release point only delays the problem, and makes all the changes come all at once. The term “rolling release” encapsulates the whole idea your good stuff from yesterday no longer supported.

The problem is the change — too much too often, if you ask the lowly average user, and certainly if you ask the suits. The last time I really enjoyed the way KDE worked was around 3.1.5. I’ve tested it since before 1.0, and have tried every point release up to 4.1. Some things never got fixed, though I was told it was. Every bug report met with, “fixed in the next release.” The next release broke my settings at least half the time, and I couldn’t migrate without starting from scratch. An awful lot of people tell me the same story.

My Windows friends won’t try Linux until they know what it is. It keeps changing, and the changes in their eyes are HUGE. Their confusion comes directly from their perception nothing good is ever allowed to stand still long enough for them to evaluate it, use it and become accustomed to it.

Posted by Ed Hurst - Dec 31, 2008 | 2:19 PM

Re: The Disaster of the Rolling Release

“The hobby community keeps the suits at arm’s length because the suits are alleged to have an impure purpose”

The above is true ..there is a distinction btw free(as in freedom of speech) software and open source software .. both share the same software development and distribution model but their politics are different .. how much correct/wrong the above quoted statement is will depend on where you stand on the free software/open source source software person ..

do you know the distinction btw the two? which one of the two do you subscribe to? seem to be an open source person ..

There is really not that much a distro like Arch Linux can do about what KDE developers do with their project or any other project for that matter ..

what a distro like Arch linux can do is picking a package that work best with them .. if you were a distro package maintainer, you will release your distro with KDE 3.1.5 because you seem to think that version works best ..all those who think like you will choose your distro and packagers ..

as a packager and a distro maintainer, how your users will update packages is largely what will detarmine if your distro is a rolling release or not .. complaining about rolling releases in linux context is complaining about how packagers and distro handle updates ..something that is completely independent of how KDE or any other project release circles ..

a perfect example of a rolling release in windows xp world is having a user who start with winxp service park 0 and update to service park3, another user start with sp1 cd and update to sp3, another start with sp2 and update to sp3 and another user who start with sp3 and do not update ..they all start at different levels and they can all update to the same level ..that is what a rolling release it ..

a rolling release means people can start with installation cds at different update levels and all can update to the latest update level ..this property of a distro is entirely up to the distro maintainers and it largely dependent on how they package their software ..

basically, a complain about rolling releases is a complain about what packagers do with their packagers ..not about proprietary software, not about how much KDE stuff break btw releases and not about why people in suits dont use linux

”{…} You will upgrade or we will not help you. Having a 6-month release point only delays the problem, and makes all the changes come all at once. The term “rolling release” encapsulates the whole idea your good stuff from yesterday no longer supported.”

The above has nothing to do with rolling releases ..if a distro first shipped with kde 3 and move to kde 4 ..then they at some point will stop supporting kde has nothing to do with how they moved from kde3 to kde4 but people just cant continue to support old projects has a policy of supporting only previous major release ..i think microsoft has stopped supporting windows2000 or will stop soon ..

Posted by mtz - Dec 31, 2008 | 8:09 PM

Re: The Disaster of the Rolling Release

mtz: “people just cant continue to support old projects” Yes they can. There may be good reasons for not doing so, but they can. My argument is, the Open Source community as a whole refuses to consider it, even to the point of contemptuous dismissal. A distro can’t change GNOME and KDE, true, but by the same token, they see no problem supporting the rapid pace of innovation by making no complaint about it. Indeed, most distros seem to stand on the sidelines cheering madly. The atmosphere of the entire thing is run, run, run.

You cite how MS is giving up on Win2K, but how long did it take? It’s now almost nine years later. All along, bug fixes and security updates, but features and system speed changed very little. Further, stuff written on DOS 5.0 still runs perfectly well, with rare exceptions; never mind Win3.1 thru Win98 stuff, and quite a few XP apps. I can’t run my WordPerfect 8 for Linux on any distro since about 2001, so far as I have found. Never mind desktops, the X server is so radically changed, that alone would keep it from working. WP 8, or even 6, for Windows still runs just fine on XP. Only with Vista have folks complained it won’t work.

I don’t take issue with innovation; it’s the majority of computer users who won’t step away from Windows who don’t like the brand of innovation they see. Even MS is being clobbered on this. We cannot help some of the reasons people won’t migrate. I’ve tried to make the case for Linux, one-on-one, with hundreds of people by myself, but the one thing I can’t overcome is the headlong race for innovation and zero consideration for “backward compatibility” — and I blame the cultural inability of the greater FOSS community for refusing to even think about it.

Everything else we do right won’t matter if we don’t meet those users where they are. In their minds, the lack of a stable desktop distro, which can be used pretty much unchanged for at least 5 years with full developer support, means FOSS people refuse to support their own work. This is not my personal ax to grind, this is just raising awareness of an issue I’ve never seen addressed when we start talking about Linux advocacy.

There are a handful of server distros with 5 or 7 year support plans, but not one oriented to the home desktop user.

Posted by Ed Hurst - Dec 31, 2008 | 9:34 PM

Re: The Disaster of the Rolling Release

you have good points microsoft has over 60,000 employees and they made more than 60 Billion dollars last year ..long term support is crucial especially in a commercial environment where business dont like to upgrade their software that often but long term support costs money, a lot of it and FOSS community can not afford it

kde just release kde4, to expect them to continue to maintain kde3 5 years from today is asking a bit too much from them ..if would be nice if they but they simply can not afford it ..only Microsoft can support their OS ten years down the line much does this long term support cost? windows 98 has memory requirenment of 32MB ..win2000 had 128 recommended, winXP has 256MB minimum required, you need a gig of memory in vista much does vista cost? i am sure you also pay for legacy support much in hardware cost does a user pay to support the latest microsoft OS? ..yes, the legacy support is there in windows but it is expensive ..

there are always trade off ..FOSS drop support of old releases much quicker than microsoft ..there are costs on both fronts .. i tried a minimalistic linux distro the other day that had a minimum memory requirements of 64MB ..there is no way microsoft can compete with linux on minimum requrenments and keep legacy support at the same time ..there is no way linux can give long term support without forcing people to pay a lot of money for that support ..

“In their minds, the lack of a stable desktop distro, which can be used pretty much unchanged for at least 5 years with full developer support, means FOSS people refuse to support their own work”

you are mixing things again the FOSS world, there are three kinds of people ..there are developers, distributors(distros and packagers) and users ..if you want long term support in linux, you dont go to developers of those projects, you go to packagers and distributors want 5 year KDE support? dont look at KDE, look at companies like IBM, redhat or novel and you will have to pay them to give you the support you want ..

you seem to fail to see the role distributors and packagers play in the FOSS world, not that many people take linux kernel from ..most people take it from whatever distro they prefer and it is the responsibility of that distro to support them and how long that support will last will depend on the distro or the company behind the distro..if you want first class support, you have to pay for it and i am sure IBM or redhat will gladly give you that support ..

“There are a handful of server distros with 5 or 7 year support plans, but not one oriented to the home desktop user” redhat refused to enter linux desktop area ..conanical has stepped up and they now have long term support for both their desktop and server offerings ..if you want long term support in desktop linux today, go for LTS versions of their releases ..i think they have a 3 year support period ..again, only microsoft can affort 10+ years of support

there are also windows programs that were released in 2001 for windows 98 that fail to run on anything newer and there are a lot of windows programs that has winXP as minimum requirement ..

Posted by mtz - Dec 31, 2008 | 11:11 PM

Re: The Disaster of the Rolling Release

If it wouldn’t have been for the comment section I would still be guessing about the purpose of this article. I recommend you to update it by more clearly specify what you by a certain term mean and also more precisely formulate your point.

As already said, Linux distributions following a rolling release scheme are few, probably less than 10 among the more established ones, and even fewer if the criterion is that it releases constant upgraded packages as soon as they’re released. My main and only distribution of choice is such a rolling distribution, Arch, which has been mentioned here in comments.

If discussing Arch you should also point out that the project very clearly explains its purpose and what the user can expect and not expect from Arch. Hence the problem you get here is that your article is based on singularity operating system market, where every product is supposed to be the same as the next. That’s more of a testimony about how damaged this market is, and how innovation got a serious setback in the midst of the 80’s. As in most markets the consumer should have a choice to find something that suits his needs better, not to find a big selection of the same. Arch defines its product and a conscious consumer can evaluate its suitability.

Personally a rolling distribution as Arch meats my computing needs better than any other distributions I’ve tested in the past. There’s no such thing as perfect though, so Arch is human in having both pros and cons. The same can be said about every other distribution, just as about every single operating system in use today.

I might however agree that there’s a distinction between desktop and server use. Even Arch as a so called bleeding edge distribution has a rock solid core system and server oriented packages. If anything temporarily breaks (note “temporarily”, because I’ve never had a Arch system break in the real sense of the word) it’s on the desktop. What might be the cause? To me the most notable issues have been caused by important changes in code of languages like python, ruby or others. Usually developers of desktop applications based on those languages are able to keep up and adjust to these changes, but some projects are small and can’t keep up the pace as well as others. A complaint in this regard can only be solved by better support from users or skilled volunteers to help with the code.

Microsoft doesn’t support a lot of products, but it’s idea of backward compatibility is also its curse. The idea of an stable API might help developers at a certain time of development, but as a result you have a Windows with a mix of new and deprecated APIs, which isn’t good for stability, security and performance. Linux doesn’t have a stable API and it hence can rewrite necessary pieces of code from scratch instead of piling up a historical mix of APIs. This illustrates to things: it’s very difficult to find a solution that doesn’t have an cons, but it also proves that sometimes a at a certain time more troublesome solution pays off in the long run.

That’s the core, but if we look at software on top of this base we see similar problems. Excel is so backward compatible that it eventually breaks its own documents. The code in Excel is huge, and it gets bigger for every upgrade of the office suit and for every revision of document formats. This I view as a bigger problem than if one of several applications using some file or document format temporarily is broken (OpenOffice can rescue some broken Micrsooft documents).

So my own view is that the struggle to get and secure open document formats is more important at the moment than the temporary issues we might face with software being upgraded. Likewise some of the issues with software isn’t a result of its developers in person, but because of changes to code making up the language base. Some problems with for example mobile phones and other devices have been the result of new firmware breaking things, not Linux breaking a protocol. We attach thousands of gadgets and devices to our computer nowadays, interacting in new ways, something that in itself force even desktop environments to change (performance wise it’s not a good idea to add indefinitely new features to a base not designed for the purpose).

To unify some projects might be good, but not as thumb of rule. Sometimes we’re upset because of a symptom of a problem is buggering us, but it doesn’t mean we know the cure.

Posted by KimTjik - Jan 01, 2009 | 3:13 PM

Re: The Disaster of the Rolling Release

The problem with a rolling release is that I first must download and install an outdated disk, then go through countless updates. It is wasted effort.

If I use a distro such as Ubuntu, it is never more than six months old and even then it has likely been updated if a month or two has elapsed since release. This means that I won’t be installing packages that will need to be updated.

The problem with PCLOS is that its repositories are limited to begin with and they are frequently stale. No matter how good the distro, it does not make sense to use it when their are fresher and more complete alternatives. I recommend PCLOS, but only if you don’t care about the size of the repositories and about having the latest and greatest.

Debian is the classic rolling release and we all know the glacial pace of development in that community. Yes, they get it right in th end, but at the cost of losing all sense of keeping abreast of the times. Stable is good, but it makes sense to keep up with what is happening at large or you get left behind.

Each user must find what works for them. If a distro is on either extreme (bleeding edge or ultra stable) it appeals to a small group of users. Debian is finding that distros that are managing to attract users are finding an acceptable mix of stability and usability.

There is a cost to be paid for staying with their model. That is fine as long as there is no resentment towards those who elect for a shorter and faster development cycle. However, some Debian users want it both ways. They want to stick with their model and then to look at the success of other distros with envy and blame them for Debian’s troubles.

Posted by LinuxCanuck - Jan 01, 2009 | 8:52 PM

Re: The Disaster of the Rolling Release

“The problem with a rolling release is that I first must download and install an outdated disk, then go through countless updates. It is wasted effort.”

The release cycle differs, it’s not written in stone. Regarding Arch you’ll see that there’s a new snapshot every second to forth month. This is how a typical installation procedure for me looks like:

  • installation of base packages from install media 10 minutes
  • reboot and command “pacman -Syu” which upgrades everything to the latest in about 3 minutes
  • from here no outdated packages will be installed

That’s for the installation effort. Besides that you could take a couple of minutes now and then to read the home page of Arch and see if there’s any announcements about how to handle certain upgrades (usually not but it happens a few times per year) before executing “pacman -Syu”. That’s about it. Mirrors are fast and hence such a total upgrade every second week is done usually within 2 minutes, while I’m surfing the net. So I don’t spend time on upgrades besides the rare occasions when an application includes adjustments to its configuration, something the pacman utility informs you about. There’s no real difference between this and running a regular distribution, only that my installation always is “the latest”.

Every user has the right to decide what effort to put into his computing. It’s however best to avoid the assumption that just because rolling distribution X works that way it also applies to every other rolling distribution. Arch isn’t for everybody, that’s not my point and this was strictly an example not some kind of promotion. What you “must” do and “if it’s wasted effort” is decided by personal preferences.

Posted by KimTjik - Jan 01, 2009 | 11:11 PM

Re: The Disaster of the Rolling Release

Hi people, This has all been most informative. Here are just a few comments from a user’s perspective. Today I realised I need to look for a new OS. My Kubuntu repositories have gone. This is is very annoying. I have only been using Kubuntu for 18 months and now it’s all over.

I am definitely looking for a rolling release distro this time. That’s what brought me to this page.

There are two main issues: Firstly: “They never get to use the system, as they perceive it, because it’s always changing,…” Exactly: as far as I’m concerned I haven’t really even finished installing Kubuntu yet. I mean, there are lots of things I have not yet learnt how to configure or tweak (e.g. some glitches with sound and Jack, buggy KDE).

Secondly: changing to a new release is extremely time-consuming. It’s a HUGE interruption to my work. It would be nice if I could just concentrate full-time on the upgrade for a couple of weeks - but I can’t, because I have to do my work.

I can’t overstate this ‘time suck’ issue enough. As KimTijk said, we attach thousands of devices to our computers these days - well, I don’t know about ‘thousands’ - but it’s certainly true that getting ALL our gear and software to work doesn’t happen in five minutes.

E.g. when I first installed Kubuntu I spent an entire 8-hour day getting printing to work properly. Part of the problem was with me: I didn’t know where to look for the most likely source of the problem. But that doesn’t alter the fact that this issue only came up because I was installing a new system.

Another example: I spent a lot of time setting up HylaFax (twice now). Quite frankly I NEVER want to do it again. Same goes for Samba. I suppose I wouldn’t mind setting up those things again once in seven years or so - but not just 18 months later.

I don’t mind changing OS when I need to get a new computer. But, with the Kubuntu repositories vanishing, I feel I’m being forced into an upgrade on someboby else’s timetable.

Anyway, this discussion has been most helpful. KimTijk’s comments have swung me towards Arch - but I’m still considering Debian too.

Posted by CB - Apr 10, 2009 | 2:03 PM

Re: The Disaster of the Rolling Release

This article is something that can be expected of one who attempts to be functional in the rational Engineering/Science world while clinging to irrational religious beliefs. It appears that Ed has gotten used to preaching to people who are just waiting for somebody to tell them what to think.

Ed wrote this article out of fear of the unknown and is trying to find a way to cope even though it has no trace of proof and hides behind generalized terminology while trying to sound knowledgeable. Clearly he has no idea what rolling release means and how it makes things so much easier for the end user.

Rolling release means you install and configure your system once and the applications that you add on top are always up-to-date. Rolling releases are not unstable, but ANY Linux distro’s experimental branch will be, but that risk is clearly stated. The only time you should have to perform an install is when you are distro-hopping. A rolling release distro is the model of how an operating system should be.

Just try to tell me FreeBSD is unstable. Even Microshaft uses it for their mail server. Fortunately the responses to this article are largely critical and my faith in the human race is restored after Ed’s article crushed it.

Ed, you are writing articles to an educated community we are not like the sheep followers you preach to. They are the ones that need you, not us.

Posted by Lucifer - May 23, 2009 | 4:54 AM

Re: The Disaster of the Rolling Release

nothing more needs to be said. People praise the rolling release model and groves of users are switching out there release cycle OS’s for this amazinbly robust linux OS.

Posted by jacko - Aug 14, 2009 | 8:41 PM

Re: The Disaster of the Rolling Release

FYI: Debian is not a roling release, it just has long and irregular release cycles, during these, only security patches are applied. libc6 2.7 in lenny will not be replaced by any higher version, only the patchlevel increases. but still, the new release is announced on a day (usually about 6 months overdue) and from that day, every package from the testing repository is marked as stable, no matter if it is openoffice or blender, which will always remain blender 2.4.6 in lenny while blender 2.4.8 is testing while 2.4.9 is already out.
but it is true you do not necessarily need a new install to upgrade: you could have started with woody in 2002 and upgraded to squeeze (the current testing) now, even skipping sarge, etch and lenny. but as a rule of thumb, there are always a few packages per dist-upgrade that make some minor trouble, so one must take care, though.
windows is no rolling release either. you do not buy win2k und get upgrades once a month until you a arrive at win7. you could even see the service packs as distict releases, they are definitely not rolling.

my expierience with non-rolling debian is: sooner or later, there is a package whose version in debian stable is too old for me. then i start mixing the system with testing or even install packages from source, which sometimes requires more packages to be upgraded, and then wait feverishly for the next debian release, testing becoming stable, upgrading the rest of my system and having a again “stable-only”. so rolling releases have some advantages for me, and i think i will use them in virtual machines.

Posted by me - Dec 06, 2009 | 5:56 PM

Re: The Disaster of the Rolling Release

http: //§ion=all
a package in etch was dropped in lenny (i guess due to dependency problems) and a higher version appears in squeeze.
Posted by me - Dec 06, 2009 | 7:06 PM

Re: The Disaster of the Rolling Release

A rolling release OS (Arch, Debian testing), like a fixed release OS (Windows, Ubuntu), is the product of the package developers and the package maintainers. Technically neither OS type are more frequently updated or contain newer packages than the other, though practically a rolling release OS is more frequently updated and contains newer packages.

Which packages are included and the specific versions available determine the stability and usability of the OS as a whole and are the sole domain of the package maintainers. The stability and functionality (seemingly opposing forces) of the packages available are determined by the developers, allowing the package maintainers only to choose what they feel will best work together and to the aim of their OS.

What I find crazy in my open source adventure is that more popular distributions (Ubuntu specifically) aren’t rolling releases. With a six-month release cycle this becomes even more appealing, as their Draconian rules about which kinds of updates packages can receive serve only to punish their users and force otherwise unnecessary full upgrades, frequently breaking systems. An ideal solution, though Canonical probably doesn’t have the funding to pursue it, would be to establish the Backports repository as its option to upgrade to a rolling release. The Backports repository could be a smooth and stable path to attaining the next Ubuntu release without waiting six months. As it stands users can only wait six months or change their sources.list to the in-progress development release (commercial for Bad Idea Jeans).

If anyone’s reading this who hasn’t read Lucifer’s response to the article, please hit Ctrl-F and type “Lucifer” if you want a good laugh. He/she must be my long-lost evil twin.

Posted by Kevin Fishburne - Dec 20, 2009 | 4:38 AM

Re: The Disaster of the Rolling Release

Ah, there’s nothing like stirring the pot!

Let’s see: FreeBSD is stable indeed, which is why I wrote so much about it the past few years. But desktop packages running it on can crash like nobody’s business, usually GNOME and KDE related.

Intelligent community? I’m counting on it. But I don’t think you addressed anything I said, Lucifer, so I’m left wondering if you are part of it.

This was not about distros, nor exactly about packages, but an underlying theme of Open Source development in general. That theme is inherent to the process. That theme is also a guarantee Linux will never cease being a hobby for most end users of it. Yes, in the real world there are fuzzy, overlapping categories. Some people will tend to stick with Linux, like me, despite a lack of interest in making it a hobby.

But the next time someone throws out a strong sales pitch, advocating the masses move to Linux, they had better be prepared for a tepid, very limited response from the buying public. Unlike many, I don’t blame the buying public for being “stupid” and not seeing the glories of Linux. I merely state the reality of why they don’t see it. Some distros are making the effort to understand why. Some distros by accident reach that same buying public, but the one common element of success in the OS market place, I contend, is what appears to that buyer as stability. I defined what I thought that meant.

I know the definition of “rolling release,” and I also know it, as with many other terms, is symbolic of something bigger. If you choose to ignore that larger symbolism, that’s fine. But then, you aren’t part of the conversation; you’re just talking past me.

Posted by Ed Hurst - Dec 20, 2009 | 8:57 PM

Re: The Disaster of the Rolling Release

Comparing a rolling Linux release to Windows is comparing apples to oranges: In the Windows world, users tend to get their applications from third parties. Even Microsoft applications tend to be treated more or less as third party applications. Each application has its own installer; each application brings with it its own libraries, if needed. Each app creates its own resources, registry entries, and the like.

In Linux land, users tend to get their applications from the repositories of their distro of choice. Yes, adventurous users can certainly venture outside the supported and community repos to get the latest wine, mozilla, whatever; but the majority of users are effectively held hostage by the fixed release system. They get updates for bug and security fixes only. If they want the latest version of the app for new features, they have two choices: upgrade their distro (and all that may entail - new boot manager, new service manager, new sound system), or go outside their repos and risk breaking their system to get the latest version of their favorite app working.

I would be in favor of a partial rolling release - leave the core of the OS alone (kernel, grub, init, hal, X, sound daemons), but allow users to upgrade apps from a supported repo (emulators, browsers, office apps, games, etc.).

Posted by Duane Hawkins - Mar 09, 2010 | 2:58 PM