The Duality of Patches

Looking back on it, I have to express some disappointment with Hardy. Right now it feels great, the world is all sunshine and roses. But that's not enough for an LTS release, and so we must ask ourselves: how did we do out of the gate, did we start strong, was there a reason for ISVs to have their act together to support the LTS on day-one?

So how did things look day one for me on Hardy? Not so good. Accessing Samba shares was unreliable, credentials kept getting forgotten. During the boot sequence, my screen was black until the login screen, I feel very bad for new users who thought their computer would not boot because they had a black screen while a scheduled file-system-check (which is automatically scheduled and they likely don't even know exists) runs in the background for twenty minutes with the progress indicator completely hid behind a veil of blackness. Working with windows network shares was very, very difficult. Nautilus frequently locked up, often forgot the username and password I had explicitly told it to remember, and could no longer open the locations I had add to my side-bar. Virt-manager couldn't run my existing qemu virtual machines, Compiz still resulted in daily hard-locks and routine visual corruption, and had to be disabled. It was the worst of times.

But now here we are at the best of times, things work by the magic of updates. Lots of updates. And I'm glad, because my life is much better with things working than not. Windows shares work better than they ever have on any Linux distro, and now are on track to be as-good-as or better than MacOS by the next LTS. The screen is no longer black while booting or checking the disk. Virt-manager now launches my virtual machines and they run as they are supposed to. And Compiz? It's enabled, I'll let you know next week how it goes.

So hurray for updates, I've got a working system, if I gave it to my mother she could do what she needed to do. But, alas, it took a while. And good things take time, lots of time, also good things come to those who wait, and other numerous cliches, but perhaps we shouldn't slap the LTS sticker on until we have let it chill out there for a while.

But Rob, it was used by tens-of-thousands (hundreds-of-thousands? anyone got figures), we worked on it during the allotted time, what more can we do? Should we give up our time-based release schedules? Hire an army of engineers with funds from sugar-daddy Mark?

No. No no no no no no no no no no no no no no no, no! But, let's have a thought experiment: Imagine if a release was not an LTS release until after the first service-pack. What would life be like? Would overall user-experience for corporate users, grandmothers, and other non-geek audiences users be improved? Would we just end up with a more buggy LTS release to start with and end up with the same quality at a later date, with everyone differing their bugs until the patch? I hope not. "Surprise, this patch is an LTS!" is about the worst release methodology I could think of. What might be better would be to plan for the next LTS to be an LTS release as soon as the first service-pack is out. The aim would be issue a release that had an equal amount of bug-fixing effort put into it as a non-LTS release. This should result in a release that is pre-LTS but still significantly less buggy than a standard release simply because the blueprints for the release would be more conservative, with about the same level of entropy as those introduced for Hardy. This slightly more stable build would then be honed and polished as Hardy was, and as of the first patch Ubuntu X.YZ would be Ubuntu X.YZ.1 LTS.

Google has taught us two things and funded the down-payment on our house by two simple principles : 1. there is still a lot of money to be made with a good algorithm and 2. Stick BETA on everything for the first five-years it is out. I think they have taken it a bit far with the BETA thing, actually ludicrously too far, but there is something to the principle. There is no substitute for a few million users doing important work with your software. I'm not convinced our LTS release are any less likely to "eat someone's data" than any other release, but I am convinced that LTS would be more meaningful if it meant that modem-users could farely reasonably keep up with the updates, OEMs and ISVs could be confident that they could get in on the PR bandwagon for the LTS release without having major headaches coming their way, and that mere-mortal developers would feel better about developing against a release that was to a great extent concretized.