Ubuntu considers “huge” change that would end traditional release cycle (ars technica)

I'm not totally sure I'm convinced by this myself (I'm on the Ubuntu release team and generally fairly heavily involved in this kind of stuff); the six-month cycle is something of a signature, has served us pretty well, and means we get reasonably frequent feedback on things like the installer which people tend to only take another look at around releases. My natural instinct tends to be to stick with the devil we know in the absence of thoroughly compelling arguments to the contrary.

On the other hand, there are certainly advantages to switching model. Six-monthly releases are expensive: as well as the cost of stabilisation alone (which clearly we still have to pay), there are all sorts of ancillary costs both in terms of the highly-qualified developers on the release team having to spin lots of wheels rather than doing other useful development work, and in more traditional terms such as updating marketing collateral, physical media, and the like. Supporting four or five stable releases at once isn't cheap in terms of human attention either, and we could do a better job of issuing timely stable updates if we consolidated on fewer targets. And third-party developers would have to do much less work if they only needed to build for, say, the last LTS and a current rolling release.

It's also worth noting that the current Ubuntu development cycle ("raring") is significantly different from previous ones, because we now gate uploads through a no-increase-in-uninstallability check, rather like that used by Debian testing (indeed, based on the same code, although with some different details of behaviour). Once we figure out the glue code we'll also be gating uploads on a variety of automatic tests. This has already made it a lot more feasible for reasonably technical users who just happen not to be Ubuntu core developers to run the development release all the time without having to watch apt like a hawk, and this was one of the major prerequisites for a rolling release model.

We shall see; things are not yet set in stone.