Slashdot Mirror
Ubuntu May Move To Rolling Releases
formfeed writes "The register claims that 'Ubuntu is moving away from its established six-month-cycle and potentially to a future where software updates land on a daily basis.'
While this sounds like a sudden change, it is apparently more of a long-term thought. The Register quotes Shuttleworth:
'"Today we have a six-month release cycle," Shuttleworth said. "In an internet-oriented world, we need to be able to release something every day. That's an area we will put a lot of work into in the next five years. The small steps we are putting in to the Software Center today, they will go further and faster than people might have envisioned in the past."' But given that many of Shuttleworth's thoughts became decisions later on, it might be interesting to see, where this one leads. Interestingly enough, five years is about the time when Ubuntu will run out of letters."
Lately, there has been a gradual shift in Linux hardware support where distros are limiting support for older hardware. I understand why they are doing it, but by doing what Ubuntu is [thinking about] doing, it could literally result in a situation where one day your computer is supported and the next day, it's not. That's not a good thing.
Typically those changes are mostly bug/security fixes. New features/APIs tend to only be released every 6 months.
Monstar L
You go from one release cycle style to another. Periodic releases to constant releases. And then back.
Each style has its advantages, but in the end you just end up changing for change's sake and no real benefit is gleaned one way or the other. It's a lot like reorganizing resources in a company. You move some people here, you transfer some people there, you change from a horizontal hierarchy to a more vertical one. Then in 18 months you change it back.
In the end, the guys on the ground doing all the nitty gritty work do the same job they've always done and the company keeps chugging along.
That being said, it's usually a case of management losing touch with the guys on the ground that causes this kind of shakeup. I wouldn't be surprised if Shuttleworth is a bit disappointed in how the business is going and is looking to change the sales story for Ubuntu. From the "stable and great" OS it is now to "cutting edge and always up to date" OS it could be with constant drops.
While i understand that you want the foundation to be fairly stable that in itself creates a slew of problems. Foremost that stuff like Firefox, OpenOffice and other userend apps wont get upgraded to newer versions until the next rollover.
My personal dream would be a distribution where the user end is getting upgraded often and fast while stuff under the hood gets overhauled less often.
A suggestion would be major overhauls once every two years of the backend stuff while user applications is kept on newest stable versions. That way developers of backend stuff gets ample time to iron bugs out while users wont have to upgrade the whole desktop just to get a new version of an app.
HTTP/1.1 400
Linux Mint is already doing a 'rolling release' with a distro based on Debian testing.
If you're talking about needing to "be able to release something every day" and you're talking about Ubuntu then the first few days are simple to sort, and I'm sure you could continue a pattern that would keep 90% of Ubuntu users happy:
Day 1: Lighten purple in default background ...
Day 2: Darken orange in default background
Day 3: Darken purple (but not enough to be back to original shade of purple)
Day 4: Put orange back to what it was
Day 5: Make all window buttons red instead of red and grey
Day 6: Put all window buttons back on the right side of the window
Day 7: Add a clock widget that uses bold
Day 8: Bundle a load of random pictures, slap Ubuntu logos on them and call them "The Ubuntu Desktop Pack" (see Gnome-Look.org for examples)
Day 9: Change the cursor theme so that it turns into an Ubuntu logo when hovering over a title bar - that's a feature, right?
Profit may be in there somewhere as well.
They won't run out of letters, they're using Base64.
Since *nix tends to be case-sensitive, they can re-use the first 26 names without collisions, and it will still be in version comparison order. Then I expect to see "0-day 0liphant" and so forth. By the time we get to the plus, minus, and equals, Canonical will have sponsored the naming of 3 newly discovered species such that they can finish the cycle. At 2 per year, that gets them to (04 + 32) = 2036. That's enough time for John Titor to come back from the future to fix the 2038 bug once and for all, along with the Ubuntu naming conventions hopefully.
In other words, Don't panic.
Debian CUT == Constantly Usable Testing.
A recently started project in Debian with a similar goal of a rolling release (along with an idea of installable snapshots).
http://cut.debian.net/
http://lwn.net/Articles/406301/
"It is better to die on one's feet than to live on one's knees." - Albert Camus
The problem is that if you wait six months between upgrades then that means you spend 12 hours downloading and installing hundreds of megabytes of changes and then it crashes part-way through and your system is hosed.
It sounds an awful lot like you're installing new versions as in-place "upgrades". I've never had that work successfully, starting from RH 6.something or so around 1999. Your much better bet is to download the ISO, then install the new version in a fresh partition. Mount all your data like normal (you do have your data on a separate partition, no?), then give the new version a spin. If it hoses something, you've still got your old version on its own partition, and switching back is as easy as rebooting.
Keeping things in separate partitions and mounting as appropriate is one of the key advantages (for me, anyway) of Unix-style filesystems. An example partition list:
Create and use more or fewer OS partitions as you find useful. I have Windows XP on one (not used on the bare metal since shortly after buying the computer), Ubuntu 9.10 in the next (thinking about wiping this and replacing with 10.10), 10.04 in the third, and I keep the fourth around to play -- check out Fedora, Arch, Mandriva (when they were still viable), etc. In each OS, I just mount my data partition as appropriate -- generally just as /data, and then symlinked from the appropriate /home/[username]/data locations. (You could just keep all /home/[username] directories in your data partition, but I tend to find that this causes config file conflicts, so I just keep the equivalent of "My Documents" in the data partition.)
This way, "upgrading" is as simple as a full install in a fresh partition. This completely avoids the problem you (and I and many others) have run into: wasting time downloading and installing hundreds of megabytes of changes and then it crashes part-way through and your system is hosed. Install after a clean wipe -- avoid that "not quite fresh" feeling!
Cheers,
"What in the name of Fats Waller is that?"
"A four-foot prune."
Here's a better idea - go for more stability, not less. If Linux is maturing as a desktop OS then there shouldn't be a need for 6 monthly, let alone daily, updates.
Here's a better idea:
1. Drop the 6 month release cycle and make LTS the default option. Then people can install an OS with a sensible lifetime.
2. Don't push any updates unless they are critical security vulnerabilities.
3. Offer optional upgrades to the major application packages, drivers etc. as they become available and where possible, and keep interdependencies to a minimum - i.e. compile them against the original distro + any vital security patches, not the latest everything (statically link them if you have to - RAM is cheap now).
The problem with the current system comes for the less technical users who want to (or are sensibly advised to) stick with the packages in the official repositories. Currently, you may find that the only "official" way to get the latest office software is to upgrade your whole fricking operating system. Its like having to take the back axle off your car in order to replace the radio.
Remember this is Linux - if we /.ers want to compile our own kernel, install the latest Firefox beta from a source tarball, reformat the drive as ext6 or scour the interwebs for a suitable .deb of the very latest LibreOffice then there's nothing stopping us. Or, we can switch to a more bleeding edge distro. However, that might work for us, but it won't work for others - and even I don't want to install a new kernel just to run the latest word processor unless it really, really needs it.
The problem is particularly bad with Ubuntu: it can't be "the Linux for the rest of us" and bleeding edge, because "the rest of us" don't want to be obliged to upgrade our whole OS every 6 months just to get the latest OpenOffice.
...its understandable with commercial software where the company depends on brining in the upgrade fees, but why should Free Software care?
In a survey of 100 programmers, 111111 thought that duck-typing was a good idea.