On Thu, 2008-06-05 at 15:19 +0100, thveillon.debian wrote: > Damon L. Chesser a écrit : SNIP > > > I'm just an average Testing user, have been for a while, and around me > almost every Debian users I know are using Testing, mostly because it's > the Debian's flavour which can compare with other distros in term of > being usable on a reasonably new computer, with up-to-date softwares. > Stable is for servers only, Testing for end users / workstations, that > is what's in many people's minds. > I thought Sid was the toy-breaking one, at least it says "unstable" on > the box, it's explicit. Unstable means packages, versions change, often daily. Stable means the both the packages and the versions will not change (security updates excepted and some backports). Unstable is NOT a reference to usability. Again I point you to http://www.us.debian.org/releases/ to read the definition of unstable. "the edge" is often called "the bleeding edge" because, contrary to many (*young?*) users beliefs, newer does not mean better. New means new, unknown, never before found bugs. > Testing has became so popular that it can barely > be considered a developer-only version, and according to my experience > (i use it for work, along with Ubuntu stations... scary no ?) it has > always been at least as reliable as Ubuntu. Testing has always been popular to run. The reason is you have three distributions to run, Stable (sounds good!) Testing (it is one better then Stable!) and Unstable (ooooooo, scary! I don't want an unstable box!!!!). If Stable is GOOD, then Testing is one better and must be better still (it has NEWER packages!), but Unstable must not work well, the applications keep locking up. This is fundamentally wrong. The assumption that the Debian distribution is linear from Stable -> Testing->Unstable is faulty. It is in fact just the reverse: Unstable-->Testing-->replaces at some point Stable. Ubuntu is not a snapshot of Testing, it is a snapshot of Unstable (AFAIK, I will not stake my life on this) with "fixes" applied. Debian Unstable with out those "fixes" works just fine for me (I do run ubuntu from time to time, I get tired of configing my system to run the way I want and ubuntu works the way I like). If you want the bleeding edge, like FC9, Mandriva, OpenSuse and all the others out there, you want Sid. You will also get the pleasure of finding all the bugs, finding upgrades that you should NOT perform and will have to wait a week to get the latest mix of packages that will work. But unlike FC9 and the ilk, you only have to wait a few days, upgrade, dist-upgrade and that nasty application is fixed, nasty bug is gone. In the rpm-based distro, you will need to re-install to some point release, or wait for the point release and then run yum/pup/whatever. This does not mean that Testing is bug free. Far from it. Many bugs are in Testing and are found after packages are moved into Testing. If they are bad enough, the packages are "yanked" out of Testing, resulting in the OP message to this list. When that happens, "nobody cares". Testing is working as it was designed to do. Weed out all the bugs so that all the packages work to form a new nucleus that will (once it is hammered out) be issued as a new Stable. If you jump on Testing just after a Stable release, you may be in for a rough ride as sudden and frequent changes are dumped in from Unstable. The reason being: There is a freeze before a release and nothing is added to Testing and all the bugs are worked out. When the freeze is lifted, all the packages from Unstable that qualify are moved into testing. If you jump on Testing in mid to lates stream then the ride will be mostly smooth. Unstable stays pretty much the same level of fun and excitement through out. If Testing works for you, Great! You will however find times when it will break or packages are not to be found (hmmmm, sounds just like Unstable!). > I do agree with apt-listbugs, I use it and think it would make sens to > include it as a default in Testing and all the more in Sid, the only > issue with it is the non architecture specific massages : output being > cluttered with warnings because a package doesn't work on arm isn't very > relevant for a x86-64 user. > > Testing is the right choice for me, and if it no longer is I would move > to another distro immediately. Etch doesn't even install on many recent > machines. > > Is there statistics about number of hit-per-version on debian.org > servers ? Would be interesting to know the balance between Stable vs > Testing usage. > > Tom > > -- Damon L. Chesser damon@damtek.com http://www.linkedin.com/in/dchesser
Attachment:
signature.asc
Description: This is a digitally signed message part