Slashdot Mirror
OpenSUSE To Offer a Rolling Release Repository
dkd903 writes "While the rumors of Ubuntu moving to a rolling release have been brought to a halt, another major Linux distribution is looking to provide a rolling release. In a message to the opensuse-project mailing list, openSUSE developer Greg Kroah-Hartman announced a new project – openSUSE Tumbleweed. OpenSUSE Tumbleweed will provide a rolling release for those openSUSE users who wishes to have a rolling release. It will essentially be a repo containing the latest stable versions of the applications."
For those who want a Debian based distro Linux Mint Debian will be a rolling release. http://blog.linuxmint.com/?p=1527
by TheSpoom (715771) Uncaring Linux user here. I have nothing to add to this but please continue. *munches popcorn*
What I like about rolling releases is you get to deal with application incompatabilities one at a time as they come up, rather than having to spend a week or few all at once when upgrading a distro.
I think it's also probably better for security, as you get the latest patches for the software. (I know the security patches get applied to downlevel releases as well by distributors, but that seems so cumbersome compared to following the application's software releases.)
I do not fail; I succeed at finding out what does not work.
Just how many distributions does he develop for? He's a openSUSE developer and a Gentoo developer, are there any others?
Seriously. I *like* to know I'm running a specific release that is fixed for a while so that I know what I'm dealing with if I run into problems, and so that if it's working fine I can *stay* on that release for a while. I also prefer not to run a release that is on the bleeding edge.
"Rolling release" looks like a sure recipe for support/dependency hell.
I'm assuming there are some positive reasons for it. What are they?
This is a great option for those who would want such a thing. And, upon further consideration, I think that I might just be such a person. Having each appliation synched to its own release schedule rather than to the operating system's sounds is very practical. Also, I like the name.
Please provide reliable market share statistics, I can't find any.
Change is certain; progress is not obligatory.
This is basically what Gentoo does. The only numbered releases they have are the annualish install discs. And if you go with the minimal install, even that doesn't draw much from the install medium, you download all the current packages.
Does anyone know if packman will have repositories that go along with tumbleweed?
People say my sig is the best thing about me.
Awesome news, too bad we're talking about Suse.
I'll use WindowsME, thanks.
Mod me down with all of your hatred and your journey towards the dark side will be complete!
...in an attempt to demonstrate that they're still relevant, now that they've been acquired by a company that nobody's heard of.
At the moment, most Linux systems (Apart from embedded) are Servers.
This is ideal for LTS versions. i.e. Ubuntu LTS, RHEL, SLES etc.
My biggest gripe is that producers of apps that are targetted for use on Servers seem to always want the latest & greatest versions of dependent packages such as PHP and not the versions that are supported on LTS Distros.
UH? Why?
A rolling release won't solve this. It will probably make it worse as the stable base provided by SLES/RHEL etc is suddenly not so stable anymore. Now do you know when a server may upgrade from one version of PHP to another? you don't.
At least with a release like RHEL, you know the version of the base platform for each release and that between releases (Eg 5.3 & 5.4) the PHP version won't change apart from bugfixes.
Rolling releases IMHO is a big thumbs down. It's going to give me more problems in the long run. I'm now less likely to go with SUSE than before.
I'd rather be riding my '63 Triumph T120.
any stats will be very rough estimates and completely depend on what B.S. methodology you wish to use. I took distrowatches distro popularity page, took the top 23 which included everything I care about, and normalized Ubuntu 10%
Mint 8.6
Fedora 8.6
openSUSE 6.7
Debian 6.4
Sabayon 5.1
PCLinuxOS 4.3
Arch 4.1
Ultimate 3.5
CentOS 3.5
Puppy 3.5
Mandriva 3.2
MEPIS 3.1
Red Hat 3.0
Unity 2.8
Slackware 2.8
Chakra 2.7
Macpup 2.6
Tiny Core 2.5
Pinguy 2.4
BackTrack 2.2
FreeBSD 2.2
Gentoo 2.1
should have mentioned that's hit for last month (not the 12 month column they also have), and kept the FreeBSD in there for relative comparison purposes, since I care about BSD though not that particular one.
Fedora has a rolling, rolling, rolling release called rawhide.
I've got a fever and the only prescription is more COBOL.
The whole model is still too distribution centric, and not making enough allowance for the difference between one user and another. We all have our own set of mission critical applications. Sometimes a critical application needs to be stable above all else, even if it's long out of date. Sometimes it must be the last major release, even if the release has many problems. Sometimes you need both, and a way to control which is the default and which is the sandbox.
If I'm working with a bleeding edge C++ compiler release and it messes up my code generation, I'm prepared to hunt that down and deal with it. I'll probably A/B against another compiler that recently worked.
There are features on my desktop that are huge benefits for my productivity. At the same time, if my X desktop goes flaky, I don't have the tip of my fingers skillset (or the interest) to bug hunt. It's not my bag. Other people can do dogfood duty before I sign up.
On another front, if my IM client baked out, I might not miss it for a month or more. I know people who wouldn't last five minutes without their buddy list.
The conflict arises when a tool you've marked for aggressive update needs to drag in something unproven you've marked for maximal stability. This needs to be resolved by the user. Tentative updates with rollback points would be nice.
In my view, this entire debate is worthless until it's centered more around the requirements of the user, and less around distribution politics.
People will say "it's too big of a nightmare to have everyone upgrading on selective fronts". The same argument was put forward about template header libraries in C++. Everyone will combine the components differently! It will be a debugging nightmare!
Actually, no. The extra design rigour involved in making the components fit together in the first place more than compensates for combinatorial deployment. There is pain, but it's not that pain.
It's ugly in practice because it takes C++ into territory where the tools struggle to produce meaningful diagnostics. Concepts would have mitigated this in part, but it was ambitious, and it risked imposing bureaucracy, so Stroustrup wisely raised a red flag late in the day. Meanwhile, new compilers such as clang are coming down the pike that flounder less badly on obvious typos.
I don't think migration to heterogeneous updates will be painless. But I think the pain involved pertains to a real problem, one well worth solving. In the long run, debugging likely gets easier, because there is more variance from one user to another in installed components to correlate against. When two people report the same weird ass bug, they won't have a Zesty Zaphod identical to ten million other users.
It just takes a different mindset to treat combinatorics as an asset rather than a liability. The ground changes under our feet in more ways than we appreciate. The recent adoption of PEG parsers is a good example. No ambiguity, linear time. Why not twenty years ago? For the packrat parser, memory consumption is a constant multiple of the input stream. Has anyone looked at the backflips in TeX to avoid memory consumption linear to input size? In the 1980s, this was a no fly zone.
Will we ever trust the dependency graph enough to give the user his due? It can't be harder than unwinding the C preprocessor in C++ refactoring tools.
Given the situation that SUSE is in right now (being spun off of Novell, with its future in question) naming a distribution after a weed often depicted in popular media as a sign of a deserted down is not the best choice.
Why not just say "OpenSUSE/SUSE users who want the latest and greatest apps can enable the 'Factory' repository?"
The Christian Right is Neither (Christian nor right). See: Matthew 23, Matthew 25, Ezekiel 16:48-50
Honestly, the method distrowatch uses is hardly accurate or representitive of marketshare:
Change is certain; progress is not obligatory.
... since "Rawhide" went off the air. Okay, let's all sing along now ...
I may be a complete fool and totally ignorant in software engineering but... since disk space is usually not an issue... why not going back to statically linked applications in order to avoid incompatibilities?
well, that "B.S." I referred to sure doesn't stand for Bona Fide Scientific. It stands for very expensive high quality cattle feed, after it's been processed by the bull.