Bryan J. Smith <b.j.smith@ieee.org>
2005-May-29 18:10 UTC
[CentOS] Re: CentOS and SL, together? -- the _real_ history of Red Hat Linux support
From: Les Mikesell <lesmikesell at gmail.com>> Here's a classic example: somewhere in the updates of RH 7.2, > the apache DSO module for mod_perl was finally compiled with > the correct options and became usable as shipped so people > running web sites with it no longer had to recompile apache > with mod_perl included statically. This was also supplied > in the 7.3 release. Unfortunately it broke again when RH 8.0 > moved to apache 2.0 but that's beside the point - or perhaps > it was the beginning of the new era.Yes, it's a ".0". We've been through this. At this point, I could care less what you think of the 2-2-2, 6-6-6 model. I _never_ said it's the "best." I just said that many of us believe it's the best we know of, and several vendors agree.> What, in fedora, is ever going to be equivalent of that > RH 7.2 -> RH 7.3 transition where features weren't exactly > frozen but there was a focus on getting existing things > right without introducing new problems.First off, I've been very critical of Red Hat's decision to drop revisions as of RHL8.0->9. But, Fedora Core 1 was pretty much the ".2" from Red Hat Linux 9, the ".1" from Red Hat Linux 8.0, the ".0" There were _some_ "sanity checks" in Fedora Core 1, but it was pretty much a "drop in" for Red Hat Linux 9. That's why I call them in my configuration management system: CL3.0: Red Hat Linux 8.0 CL3.1: Red Hat Linux 9 CL3.2: Fedora Core 1 Then Red Hat changed things with CL4.0, Fedora Core 2. But then they matured in CL4.1, Fedora Core 3. Now I wish Red Hat would have used a revision to warn us that Fedora Core 2 was a ".0" -- but it was well known that they were switching the kernel, among other things common to ".0" (like SELinux, udev -- although not totally until FC3, etc...). The reality here is that Red Hat _must_ continue to keep something aligned with the 2-2-2, 6-6-6 model in Fedora Core's development. So far, they have pretty much done this. Fedora Development-Test-Core replaces Red Hat Rawhide-Beta-Linux. Fedora Core _continues_ to hit a 6 month release schedule, just like Red Hat Linux before it (with exception of the "switchover" from Red Hat Linux 10 to Fedora Core 1 -- which added 2 months). There was basically a "maturity" in Fedora Core 1 beyond Red Hat Linux 9, which several things fixed in Fedora Core 1 also being applied in RHEL 3 updates. Fedora Core 4 hasn't been released yet, but it looks to be a CL5.0, which several changes (e.g., GCC 4.0). I wish Red Hat would stick with GCC 3.4 for one more release, and making it a CL4.2, and then change GCC in Fedora Core 5 (which would be CL5.0). But the quality of Fedora Core is reduced, it _will_ affect RHEL. Red Hat knows this, although I think the lack of revisions starting with Red Hat Linux 9 is starting to cause some concern. I'm one of the people that has been vocal on this. So don't you assume for a moment that I am not critical of Red Hat. In fact, I'm using Novell 9-10 / SuSE 9-10.x as an example that there is _no_reason_ Red Hat shouldn't be using revisions on Fedora Core.> No one here is interested in SLA's, or we wouldn't be having this > discussion on the mail list of a distribution that doesn't offer them.But the discussion is _why_ CentOS doesn't have this or that. This is what I mean by "ignorance" -- people don't stop to realize that it's based on RHEL, several maintainers have stated it will continue to be based on RHEL, and RHEL is SLA-focused, completely. Because Red Hat _does_ take owernship of anything it ships with it.> We just want a product that mostly works and isn't too far behind > the developers.Then if you're talking anything Red Hat developed, then you want Fedora Core. What I designate as CL4.1, Fedora Core 3, is _very_ _stable_ IMHO. Of course CL4.0, Fedora Core 2, was not, and I avoided it. And yes, I sure wish Red Hat would re-introduce versioning in the Fedora Core releases, after that stopped with Red Hat Linux 9.> Come on - I would have guessed that you still had some RH 7.3 boxes > in production too. Or do you only work with companies that will > pay for vendor support on everything?First off, Red Hat Linux 7.3 is _still_ supported by Fedora Legacy. But that's not the point here. The point is _please_ show me any release that Red Hat supported for 4 years other than Red Hat Linux 6.2 and Red Hat Linux 7.3? Pretty much Red Hat's unofficial history has been "support the last .2 release for 2 years, drop most of the .0/.1 releases." Red Hat's unofficial policy has _never_ been to support Red Hat Linux releases a long time. When Red Hat Linux 8.0 came out, and Red Hat Linux 6.2 dropped, Red Hat was also going to drop Red Hat Linux 7, 7.1 and was considering 7.2. All of the sudden, people bitched and moaned. Why? Because a lot of companies had standardized on Red Hat Linux 6.2 nad Red Hat Linux 7.1. And that's when this whole "myth" of Red Hat supposedly supporting .0/.1 releases for 2+ years began, let alone .2 releases for seemingly forever. The reality is that people had sold companies on the fact that Red Hat was Microsoft, and 5+ years support was going to be maintained so they could get Linux in companies. That was _never_ true. So when bosses started questioning why, the buck was passed onto Red Hat. In fact, Red Hat _rarely_ did it more than 2 years for a ".2" release. E.g., the second Red Hat Linux 7 came out, Red Hat dropped 5.2 -- and they had dropped 5.0 and 5.1 well before that (by Red Hat Linux 6.1). At various points of Red Hat Linux 7.x releases, they started dropping Red Hat Linux 6.x releases. But by the time Red Hat Linux 8.0 came out, Red Hat Linux 6.2 and 7.1 were so widely adopted, people didn't want to budge. The old, unwritten rule of upgrade to the last .2/.3, and the previous version's .2/.3, was no longer being accepted by users when Red Hat Linux 8.0 came out. They expected Red Hat to support there distros longer -- a total _waste_ at bugfixes for 7 simultaneous revisions. Which is why the new 1 year support policy was announced -- over 18 months _before_ the Fedora Project was announced. Now Red Hat continued to support Red Hat Linux 7.3, because it was the "last .3," but they wanted a blanket policy of "don't expect us to support it more than one year." That's always been their "underpromise, overdeliver." They just wanted people to stop expecting them to do things they _never_ did. I still can't believe people still proliferate things they _never_did_ do before they introduced Red Hat Linux 8.0. Companies that want a 5+ year supported distro need to pay for RHEL. And those same consumers were the ones already complaining that Red Hat Linux was "too experimental." Hence the SLA ownership. Now, after all this, am I just an asshole? Or am I merely trying to explain the realities of Red Hat's _real_ history? People who believed that Red Hat would support .2/.3 releases more than 2 years, or every little .0/.1 release more than 1 year were lying to themselves. Red Hat never ran into that _until_ they tried to introduce Red Hat Linux 8.0 and drop support for at least 6.2, 7 and 7.1, as well as 7.2 now that 7.3 had been out. Not even SuSE promised to support their releases more than 2 years, and have typically fallen short. And only the Debian project seems to have a history of support on-par with Red Hat (but they also aren't as adopting as Red Hat simultaneously either). So, when is this going to end? When I've decided to live in this "When Red Hat was 'good'" fantasy that many others seem to live in? -- Bryan J. Smith mailto:b.j.smith at ieee.org
minichaz
2005-May-29 23:34 UTC
[CentOS] Re: CentOS and SL, together? -- the _real_ history of Red Hat Linux support
On Sun, 2005-05-29 at 14:10 -0400, Bryan J. Smith wrote:> From: Les Mikesell <lesmikesell at gmail.com> > > Here's a classic example: somewhere in the updates of RH 7.2, > > the apache DSO module for mod_perl was finally compiled with > > the correct options and became usable as shipped so people > > running web sites with it no longer had to recompile apache > > with mod_perl included statically. This was also supplied > > in the 7.3 release. Unfortunately it broke again when RH 8.0 > > moved to apache 2.0 but that's beside the point - or perhaps > > it was the beginning of the new era. > > Yes, it's a ".0". We've been through this. > > At this point, I could care less what you think of the 2-2-2, 6-6-6 > model. I _never_ said it's the "best." I just said that many of us > believe it's the best we know of, and several vendors agree.Guys... As exciting as it is watching you argue this out over and over again, can you take it off-list now please? Thanks, Charlie
Reasonably Related Threads
- Re: Demonizing generic Linux issues as Fedora Core-only issues -- WAS: Hi, Bryan
- CentOS and SL, together?
- Re: CentOS and SL, together? -- attributing statements to me (that I didn't make)
- Re: pronunciation? -- loving CentOS doesn't mean you have to bash Red Hat
- Apache 2.2 EOL - what is Red Hat's story for RHEL6?