On Wed, Mar 04, 2015 at 04:20:26PM -0700, Tod Hansmann wrote: > I hear this point a lot, and I find it extremely lacking. "Stable" and > "buggy" are ambiguous terms, and if we want "stability" of any extreme > definition (like security or conservative in change), we should be talking > about BSD or something.
While I understand what you are saying, I made a fairly distinct definition of "stable" according to Debian, and that Ubuntu doesn't match this same rigor with their LTS releases. Most of the Ubuntu bugs we've been hit with are userspace, mostly Upstart and libvirt. We have run into a couple interesting KVM bugs that are Ubuntu specific, but they've since been addressed. > Waiting for a .1 release of Ubuntu has some prudence attached to it, but only > if you have some very specific bugs to deal with (I've seen some, personally > mostly on the desktop side than the server side), and you should know what > those KINDS of bugs are. It's upset our stability enough to make it a general policy to wait for the .1 release before upgrades. I couldn't list everything that has impacted us with initial LTS releases; partly because I wasn't employed during most of that time, and partly because we just don't have a centralized list of all bugs that we've been hit with. > It's not typically kernel bugs, so maybe that's the > only thing that matters to you? Maybe not. Dismissing anything as "unstable" > or whatnot across the board is just being hand-wavy (and I'm not accusing > Aaron of this, it's just an email chain, he's usually very reasoned in his > approaches, this is just general audience stuff). Most has been userspace, indeed. Because staff run Ubuntu on their workstation, and we're trying to standardize on a release for support, we even found with the initial 14.04 LTS release, that there was an X bug that prevented Unity from executing, leaving staff without a graphical desktop with our installed graphics cards. I don't know if this has since been addressed, or not. > The bigger complaint against Ubuntu should be its change in structure (like > upstart) all the time. That doesn't mean "stability" whatever it's > supposed to mean isn't important, but it should be a pragmatic and > definitive requirement to our overall system goals, not some nebulous > ideology. I have found Fedora to be more of a testbed of testing various changes that make dramatic changes to system, such as moving X to /dev/tty1 and migrating to systemd(8) (among many other things). I haven't found Ubuntu to be that upsetting it terms of changing its core structure, other than perhaps, forking GNOME to create Unity, and the Amazon spyware in the search lens. Ubuntu seems to be more conservative in this regard, than Fedora, for example. > If "has lots of testing done by various users on lots of hardware in > different environments" is important to us, Ubuntu wins here. If "has lots of > performance testing done in the wild for security purposes by a few niche but > very focused groups" is a better definition, than Debian does. If "has > support for our enterprise" is our definition, RHEL is going to beat > Canonical (I love Canonical, but their support is not comparable to RedHat's > at that level in any review I've ever read, fan bias aside). I haven't encountered support from Canonical, but Landscape is inferior to Satellite/Spacewalk. Also, Canonical's traning program is vastly inferior to the RH training program. These are just my opinions though. > That said, you don't typically write software for Ubuntu, you write it for > some runtime or some server daemon of some kind, so it might not matter > beyond sysadmin headaches, or it might very well matter because sysadmin > headaches cost a lot (if you have thousands of servers, for instance). > > The point in all this is, choose the OS, like any other tool, that does the > job you need it to, not the dogma you want or the system you know. If I'm > an apt fan but have never use yum, well, learn yum and THEN do an analysis > without my pride on the line. Whatever's popular is irrelevant unless > you're doing this socially (and some of us are). Indeed. The OP asked about the stability of Debian in an enterprise environment, and how it might stack up to CentOS as a result. So, I gave the QA process of packages migrating from "unstable" into "testing", as well as links that further explain the Debian QA process. I only wish to answer the OP question that Debian is a robust and stable operating system for mission critical applications. Now, whether or not it has the vender support or performance that other operating systems like RHEL or HPUX have, isn't something I addressed (although I would be willing to). -- . o . o . o . . o o . . . o . . . o . o o o . o . o o . . o o o o . o . . o o o o . o o o
pgpY1jQ2MdSBU.pgp
Description: PGP signature
/* PLUG: http://plug.org, #utah on irc.freenode.net Unsubscribe: http://plug.org/mailman/options/plug Don't fear the penguin. */
