| |||||||||||||
Fedora and long term supportFedora and long term supportPosted Oct 17, 2008 21:41 UTC (Fri) by Tet (subscriber, #5433)Parent article: Fedora and long term support Jon Stanley says: Fedora's stated goal is to advance the state of free software. Maybe it is now, but it wasn't always that way. When the project started out, the primary goal was to produce a community supported general purpose OS (see the original Fedora web site for details. Yes, being on the leading edge of the open source world was mentioned, but the main goal was to produce something usable as a general purpose OS. Those goals have sadly long since been removed from the site, as the focus has changed. As for a Fedora LTS, I think the problem might lie elsewhere. RHEL/CentOS is a much more appropriate distribution for that, and I'd love to run one of them on my servers, but quite frankly, they're obsolete. I don't need the bleeding edge that is Fedora, but I do need something a bit more up to date than RHEL. Take python, for example. RHEL is still using 2.4, which is positively archaic. 2.5 was released over two years ago, and 2.6 is available now. I don't know if there is a published timeline for RHEL releases, but I've seen speculation on the net that RHEL6 could be as far away as Q2 2010. So potentially, I'll need to wait up to 4 years between python 2.5 features becoming available and me being able to use them in apps I want to deploy on RHEL. That's simply not good enough. The gulf between Fedora and RHEL is simply too large, and there needs to be something to fill it. Either Fedora needs LTS, or RHEL needs more frequent releases. Yes, I know just how much extra maintenance overhead that would involve for Red Hat, but I suspect it's probably a better solution to them problem than Fedora LTS. But either way, something needs to happen, because at the moment, the gap between the two appears to be being filled by Ubuntu...
(Log in to post comments)
Fedora and long term support Posted Oct 17, 2008 21:51 UTC (Fri) by jspaleta (subscriber, #50639) [Link]
Are you saying that the gaps between Ubuntu LTS 6.06 and Ubuntu LTS 8.04 are smaller than the gaps between RHEL 4 and RHEL 5? They both have about a 2 year gap between releases.
Since the RHEL and Ubuntu LTS release calendars don't sync up, they are essentially staggered about a year apart, there is no direct technology comparison. RHEL 4 is a year older than Ubuntu LTS 6.06 which is a year older than RHEL 5 which is a year older than Ubuntu LTS 8.04. All of them will become stagnant overtime.
-jef
Fedora and long term support Posted Oct 17, 2008 22:15 UTC (Fri) by sbergman27 (guest, #10767) [Link]
"""
Are you saying that the gaps between Ubuntu LTS 6.06 and Ubuntu LTS 8.04 are smaller than the gaps between RHEL 4 and RHEL 5? They both have about a 2 year gap between releases. """
Well, that's 22 months for Ubuntu LTS. RHEL4->RHEL5 was 25 months. It's looking as though RHEL5->RHEL6 is going to be significantly more than that. (Whatever happened to "predictable release schedule" as a selling point for RHEL?) In my experience, I can trust Ubuntu non-LTS releases more than I can trust a new Fedora "bleeding edge" release. So, if the need arises, I can feel reasonably comfortable upgrading a client from an Ubuntu LTS release to a non-LTS release, with the intent of upgrading to the next LTS release when it becomes available, all without being uncertain about their future upgrade path or security patch coverage.
Fedora and long term support Posted Oct 17, 2008 22:26 UTC (Fri) by jspaleta (subscriber, #50639) [Link]
For the record.. I was hoping the person I replied to would actually answer the question I asked.
Your comments again point out that there is a desire to see an upgrade path from fedora into the enterprise relevant offering. Will that desire manifest itself in the form of community leadership to drive effort towards that goal?
-jef
Fedora and long term support Posted Oct 17, 2008 23:44 UTC (Fri) by sbergman27 (guest, #10767) [Link]
"""
For the record.. I was hoping the person I replied to would actually answer the question I asked. """
For the record, this is an open forum. The person you asked may freely reply at any time.
"""
Ubuntu has managed to fulfill that desire for many users and, increasingly, for admins. It is now incumbent upon Fedora devs to either respond to that desire or not, at their discretion.
Fedora and long term support Posted Oct 18, 2008 0:09 UTC (Sat) by jspaleta (subscriber, #50639) [Link]
Whether or not Canonical has been able to do it sustainably is still open for debate. I look forward to Shuttleworth being able to announce that Canonical is breaking in their service business around LTS.
In the meantime, for the next several years until that happens, I'm more than happy to talk to any group of people..volunteer or commercial entities which would like to help extend the Fedora project services in a sustainable way..by committing to contributing their time or resources to that effort.
-jef
Fedora and long term support -- copy the Ubuntu solution Posted Oct 18, 2008 3:37 UTC (Sat) by qg6te2 (guest, #52587) [Link] Perhaps the solution is simply to take Ubuntu's idea of LTS and apply it to the Fedora/Red Hat divide -- e.g. Fedora 10 (or 11) would become RHEL 6. The official paid RHEL support would be for a selected set of packages. All other packages would be unofficially (or community) supported. The attitude of "Fedora is upstream for RHEL" should be removed -- anybody worth their salt reads this as "Fedora is a beta / proving ground for RHEL" anyway. Sidenote: I recently had to install a Linux based OS on a new machine (projected lifetime: 5 years). As much as I like Fedora (used it since F1), it's simply too unstable and breaks too many things between releases. RHEL 5 / CentOS 5 are too ancient to consider seriously (the requirements for the machine were a recent gcc, python, etc). Ubuntu fits this niche perfectly, as the upgrade path is clear and a lot less error-prone. Having said that, Ubuntu's/Canonical's contribution to the community is a considerably less than Fedora/Red Hat. Rather disappointingly I had to put Ubuntu on the machine.
Fedora and long term support Posted Oct 18, 2008 21:28 UTC (Sat) by sbergman27 (guest, #10767) [Link]
"""
Whether or not Canonical has been able to do it sustainably is still open for debate. I look forward to Shuttleworth being able to announce that Canonical is breaking in their service business around LTS. """
I would like to point out that you are attempting to generate Fear, Uncertainty, and Doubt surrounding Ubuntu. This is not the first time you have done that; It seems an ongoing campaign. That is, in my opinion, beneath the dignity of the Fedora and Red Hat communities, and I recommend that you cease, lest it be thought that the Fedora community condones such shady tactics.
Fedora and long term support Posted Oct 21, 2008 9:33 UTC (Tue) by ekj (guest, #1524) [Link]
Would you please stop doing that ?
You do this repeatedly and it amounts to nothing more than FUD.
Yes, it is true that Ubuntu, or more precisely Canonical, may fold at some unknown time in the future. This is true of all other distributions too. There is no evidence that Ubuntu is about to have the plug pulled on it, and it's not as if it's the first linux-company to run in the red for a while.
Please stick with actually relevant critiques and drop the insulting FUD. There's no guarantees Redhat will be around in 3 years either. That's just how the world works.
Fedora and long term support Posted Oct 21, 2008 15:44 UTC (Tue) by kragil (subscriber, #34373) [Link]
You guys need to stop. Don't feed the Spaleta troll. He makes up FUD because he can't stand Ubuntus success and it even turned into hatred towards Mark.
I am not joking. Just look at any story on LWN about Ubuntu and you will find long posts describing his twisted view of reality.
Just ignore him.
Fedora and long term support Posted Oct 17, 2008 21:51 UTC (Fri) by JoeBuck (subscriber, #2330) [Link] I've run every Fedora since FC1, as well a the old box-set Red Hat releases going back to the late 1990s. It was clear when FC1 came out that it was, in effect, the beta for what was going into RHEL releases, so everyone expected it to be bleeding edge and a bit rough, though not as rough as Red Hat marketing people of the time were suggesting (they were actively trying to scare non-hobbyists away from Fedora by exaggerating its instability and risk).
Fedora and long term support Posted Oct 18, 2008 1:24 UTC (Sat) by SEJeff (subscriber, #51588) [Link]
Fedora Core 1 was tolerable. When they released Fedora Core 2 with SELinux
turned on by default was when it was a nightmare. I've got a Fedora 10 box beside me and it is actually quite nice. Most of my workstations are Ubuntu thought because it "just works TM".
Ubuntu doesn't belong on a server, RHEL or CentOS does. Just my 2 cents.
Fedora and long term support Posted Oct 18, 2008 1:49 UTC (Sat) by rahulsundaram (subscriber, #21946) [Link]
Fedora Core 2 did not include SELinux by default in the GA release. You are plain incorrect about that. I assume you are referring to the development release (rawhide), when you refer to "Fedora 10" as well.
Fedora and long term support Posted Oct 19, 2008 0:22 UTC (Sun) by sbergman27 (guest, #10767) [Link]
Fedora Core 2 included it. IIRC it was set to permissive mode or disabled. I can't remember which. But I'm not sure what the significance of your "correction" is other than to prevaricate, since the SELinux nightmare did begin in earnest in Fedora Core 3. (I recognize a tactic that you have employed previously.) It was only in FC6 that users were even provided the tools to fight SELinux effectively. Does the fact of its introduction being in FC2 or FC3 really make a difference today, 4 years later?
Fedora and long term support Posted Oct 19, 2008 0:49 UTC (Sun) by rahulsundaram (subscriber, #21946) [Link]
Getting the details right is always important even if we are discussing something in the distant past. SELinux was not enabled by default in Fedora Core 2. So the claim that it was a "nightmare" for that release was obviously incorrect. If your experience is only with a test release of Fedora Core 2 then extrapolating it to the GA release or even a later release would be incorrect which I have seen many many users do.
Fedora Core 3 (only about a dozen network facing services and I doubt it presented any major obstacles) and later releases up until Fedora 8 enabled only the targeted policy by default. Strict policy was available in the repository however. In Fedora 9, strict and targeted policy was combined together.
Tools related to policy management were provided right from the beginning including ones to manage SELinux booleans, audit policies etc. I am assuming you are referring to SELinux troubleshooter which is only one among many of the tools provided within Fedora related to SELinux.
Fedora and long term support Posted Oct 19, 2008 2:10 UTC (Sun) by sbergman27 (guest, #10767) [Link]
"""
Getting the details right is always important even if we are discussing something in the distant past. """
The most important detail regarding SELinux is that there has been a great deal of pain experienced by users. Turning off SELinux is still a standard policy among those users, though actually saying so in public has become stigmatized.
It is good that Fedora devs had the presence of mind to provide tools to turn it completely off.
Fedora and long term support Posted Oct 19, 2008 3:40 UTC (Sun) by JoeBuck (subscriber, #2330) [Link] I've run my home systems with SELinux on for some time. By now, they've gotten the kinks out, and the only way to do it was to get lots of people running it and working out the bugs.
Fedora and long term support Posted Oct 19, 2008 3:58 UTC (Sun) by vonbrand (subscriber, #4458) [Link] Same here. Yes, it was a pain at the start, now it is almost unnoticeable.
Fedora and long term support Posted Oct 19, 2008 3:59 UTC (Sun) by sbergman27 (guest, #10767) [Link]
"""
I've run my home systems with SELinux on for some time. """
Good for you. Last I looked adding a printer using lpadmin with SELinux enabled created a situation where the printer worked perfectly... until the system was rebooted, and then it was as if you had never executed any of your lpadmin commands.
"Security" is supposed to do less damage to you than to the other guy. I'm not at all sure that SELinux achieves that goal except in certain select circumstances.
Fedora and long term support Posted Oct 19, 2008 10:48 UTC (Sun) by dwmw2 (subscriber, #2063) [Link] Good for you. Last I looked adding a printer using lpadmin with SELinux enabled created a situation where the printer worked perfectly... until the system was rebooted, and then it was as if you had never executed any of your lpadmin commands.Bug number please. I'm willing to bet it got fixed within days of being filed.... if you actually bothered to file it. The SELinux folks are really good at responding to bug reports like that.
Fedora and long term support Posted Oct 19, 2008 16:15 UTC (Sun) by sbergman27 (guest, #10767) [Link]
It was in RHEL/CentOS. So it made it quite a few "days" before being fixed. Months to years, actually.
Fedora and long term support Posted Oct 19, 2008 22:49 UTC (Sun) by wtogami (subscriber, #32325) [Link]
> It was in RHEL/CentOS. So it made it quite a few "days" before being
> fixed. Months to years, actually.
You claim to be impartial, yet I find many of your comments to be incredibly annoying and with half-FUD.
http://people.redhat.com/dwalsh/SELinux/RHEL5/
Where is the bug that you filed for your issue? If you speak in generalities without citing real examples then you are spouting FUD.
Fedora and long term support Posted Oct 20, 2008 2:09 UTC (Mon) by sbergman27 (guest, #10767) [Link]
"""
You claim to be impartial, yet I find many of your comments to be incredibly annoying and with half-FUD. """
Warren,
I don't claim to be impartial. I claim to be a RHEL/Fedora fan with some complaints. I'm sorry that you find my comments to be annoying, and take issue with your claim that they are "half-FUD". I certainly respect your contributions.
Please do not hide behind the "where's your bug report" facade. I may have reported it. But I think I worked around that irritation and moved on. IIRC, I had 6 retail stores to open that weekend.
Bugs in RHEL/CentOS are rare enough. I don't want to give the impression that they are not. But the bug I reference did make it through the process, bug number or no. And it's Fedora that's really buggy.
Fedora and long term support Posted Oct 20, 2008 10:55 UTC (Mon) by rahulsundaram (subscriber, #21946) [Link]
Any actual bug reports related to SELinux are usually fixed within a very short time. As David Woodhouse indicates, SELinux developers are among the most responsive and fix any real issues very quickly. Despite unsubstantiated claims to the contrary, actual stats (smolt, RHN etc) indicate that majority of users leave SELinux enabled on their systems and that number tend to go up over time.
There are many real world security exploits getting mitigated or prevented by SELinux. It is also getting adopted by Ubuntu and even OpenSUSE. Feel free to draw your own conclusions from all that but it seems obvious to me.
Fedora and long term support Posted Oct 20, 2008 14:04 UTC (Mon) by sbergman27 (guest, #10767) [Link]
You don't have to take my word for it. Install RHEL/Centos 4.3. Create a printer with lpadmin. Test it out. Reboot the machine... and you will find that it does not even appear in printers.conf. This may be true in the current 4.7 release, as well. It made it through Fedora and at least 4 releases of RHEL. Why not 4 more?
Ubuntu provides the Selinux libraries in Intrepid. But the far more sane and less problematic AppArmor is used by default. I have not kept up with what the MS-Linux devs are doing. But last I looked, AppArmor was the default there, as well.
Fedora and long term support Posted Oct 20, 2008 15:34 UTC (Mon) by rahulsundaram (subscriber, #21946) [Link]
Filing a bug report is easier for you and certainly much easier than getting me to install another operating system to verify any bug. I am not questioning your claim. Merely saying that reported bugs have a much better chance of actually getting fixed and that SELinux developers are a pretty responsive bunch in my experience.
Ubuntu doesn't seem to have included the latest policies yet but they likely will considering that Tresys is working on it.
I am not sure, Apparmor has a bright future considering Novell's action's.
http://www.news.com/8301-13580_3-9796140-39.html
It is unlikely, Ubuntu has developers working on it either. This is a still a solution that hasn't gotten upstream yet though it might change at some point
http://james-morris.livejournal.com/35287.html
Let's see
Fedora and long term support Posted Oct 20, 2008 19:02 UTC (Mon) by sbergman27 (guest, #10767) [Link]
"""
Filing a bug report is easier for you and certainly much easier than getting me to install another operating system to verify any bug. """
That's pretty typical of my interaction with Fedora officials: "You are the user. We expect you to do the work".
No wonder Fedora has lost so much ground with Linux users over the last few years. I won't mention which distro has picked up all the ground that Fedora has lost.
Fedora and long term support Posted Oct 20, 2008 19:38 UTC (Mon) by rahulsundaram (subscriber, #21946) [Link]
Steve, you are now bordering on trolling. You are talking about a CentOS bug and then referring to losing Fedora users. Get your story straight. No project can magically fix issues without it getting reported with the specific details.
If popularity is the only argument for all the tired conversations, Windows must be fixing all their bugs to be so popular! Yes, Fedora does rely on its users to share some of the burden and I believe so does all Linux distributions. In this case, it is simple: You as a user report the SELinux bug you claim to run across and developers will fix it pretty quickly usually completely for free. Seems a fair deal to me. I am not going to install CentOS 4.7 to verify the bug you claim to exist.
A) Because I am not a CentOS user. I run Fedora on pretty much all my systems 24/7 and my primary system at the moment runs rawhide in part because I want to help fix bugs before it hits most users. That will help you, the CentOS user as well in the long run but not immediately.
B) The particular issue with printers is unlikely to be something I can verify easily considering that I don't have access to a printer at the moment.
If that makes me a bad guy, so be it. Good luck.
Fedora and long term support Posted Oct 21, 2008 3:13 UTC (Tue) by sbergman27 (guest, #10767) [Link]
"""
Steve, you are now bordering on trolling. """
I don't think so. Microsoft certainly must be doing a few things right to retain their popularity with the general public. Though, as we both must know, they do a few things wrong, as well. Conversely, Fedora does a few things right, and a lot of things wrong, as well.
Fedora and long term support Posted Oct 21, 2008 6:42 UTC (Tue) by rahulsundaram (subscriber, #21946) [Link]
I am sure Fedora does some things wrong just like any other project but not fixing unreported CentOS bugs isn't in that list. I am happy to tell you that.
Fedora and long term support Posted Oct 20, 2008 23:27 UTC (Mon) by mmcgrath (subscriber, #44906) [Link]
"No wonder Fedora has lost so much ground with Linux users over the last few years. I won't mention which distro has picked up all the ground that Fedora has lost."
That's so strange - all of our metrics, which are publicly available btw, show a continued growth in use of Fedora. As to which distro seems to have picked up all the ground we've "lost" which metrics / numbers are you referring to?
Fedora and long term support Posted Oct 21, 2008 3:01 UTC (Tue) by sbergman27 (guest, #10767) [Link]
As to which distro seems to have picked up all the ground we've "lost" which metrics / numbers are you referring to?
Do I even need to mention it? No doubt the absolute numbers show an increase. Go Linux! Go Home Computing, and Computing in general!
But Fedora's percentage of the market has fallen dramatically in the last few years.
Fedora and long term support Posted Oct 21, 2008 3:11 UTC (Tue) by mmcgrath (subscriber, #44906) [Link]
Sorry, that's an opinion. Not a fact. Show me metrics from any distro or show Fedora how it's own metrics show lost share. Any actual facts released by a distro will do, otherwise you're treating opinion as a fact.
Fedora and long term support Posted Oct 18, 2008 8:09 UTC (Sat) by pcampe (guest, #28223) [Link]
>As for a Fedora LTS, I think the problem might lie elsewhere. RHEL/CentOS
>is a much more appropriate distribution for that, and I'd love to run one >of them on my servers, but quite frankly, they're obsolete. I don't need >the bleeding edge that is Fedora, but I do need something a bit more up to >date than RHEL. Take python, for example. RHEL is still using 2.4, which >is positively archaic. 2.5 was released over two years ago, and 2.6 is >available now. I don't know if there is a published timeline for RHEL >releases, but I've seen speculation on the net that RHEL6 could be as far >away as Q2 2010. So potentially, I'll need to wait up to 4 years between >python 2.5 features becoming available and me being able to use them in >apps I want to deploy on RHEL. That's simply not good enough.
I think that the solution for this is having a RHEL release once per year (now is about every 18 months) with a shorter lifecycle (now it's 7 years).
But you cannot *desire* to have the latest python release on a production server, as you don't want to have the latest gcc release (and C language has been standardized some *decades* ago). This is because all the certified software on RHEL, which is a major asset for Red Hat when they sell RHEL to large customers, will need to be re-tested, and it's simply a too big effort for a marginal improvement.
If you really need python 2.6, you could: a) install it by yourself, this is not forbidden by Red Hat, you just have no support for it; b) create a virtual machine that contains it, so you could further separate between production and stable systems and advanced systems.
Fedora and long term support Posted Oct 23, 2008 3:35 UTC (Thu) by mmarlowe (subscriber, #11374) [Link]
I somewhat agree with this, but with the following comments:
A new RHEL release each year probably isn't going to happen (RedHat wouldn't want to release that frequently). I think they release every 2 years now and I've heard of movement towards the cycle lengthening rather than shortening (RHEL having a 8YR lifetime, and new releases coming out every 3-4 years - "the half way point in each releases lifetime").
RHEL does an excellent job of what is designed to do -- I can't believe Dell/HP/IBM are selling new servers that are fully supported under RHEL3, yet that is apparently the case as I see frequent discussions of the matter on vendor mailing lists. Larger or very production focused companies need to use not only a single distribution but a single release across multiple generations of hardware and over a prolonged period.
Of course, within the first half of the enterprise lifetime, RedHat is putting out new service releases every 6-12 months with slightly enhanced/newer functionality that meets the needs of most production users.
The needs that are not meant, however, and which have caused so many users to switch to fedora are:
Are these needs being met in fedora? Yes, somewhat, but at the cost of essentially creating a 2nd distribution with its own set of problems rather than simply extending RHEL. The split between fedora and RHEL is really a high maintenance solution(of which I think is eventually doomed), and its natural that RedHat will want to minimize the costs associated.
I have yet to see a clear explanation of why RedHat shouldn't solve this situation by:
Some readers are probably thinking this is nothing more than a reinvention of "RedHat Linux", I don't think so because:
Maybe I'm missing something, but the whole Fedora project has seemed like a waste of effort to me from the beginning.
Fedora and long term support Posted Oct 23, 2008 8:55 UTC (Thu) by nim-nim (subscriber, #34454) [Link]
It is totally unrealistic to expect a system with only 24 months cycles to work. Fedora has a 6 months release schedule because you *need* the release reality check this often to avoid drifting into vista-like land.
And even with a six month schedule people still think Fedora drifts too far into experimental land.
Musings on python by a fairly long-term user. Posted Oct 18, 2008 17:14 UTC (Sat) by maney (subscriber, #12630) [Link] You need to understand that I love Python - it's been my language of first choice for most things (viz., applications-level work) for years now. I mean, heck, Python saved my from PHP! <wink>But the way the language changes - no, it's not so much that it changes that's a problem. At the language and base library level they really do a pretty good job of minimizing the necessity of writing backwards incompatible code. It's the way nearly everyone else doing Python library, framework, and application development seems unable to resist using every new feature the moment it becomes available, with little or no apparent consideration as to whether the benefit, as opposed to the shinyness, of doing so really should outweigh consideration for backwards compatibility. The Django web framework is a blessed exception to this tendency, but alas it's an all too rare one. The typical Python project relies (at least in its actively developed trunk) on features available only in the very latest release of the langauge, even when they're merely syntactic sugar for something that's been easy enough to do for years; the typical Python project also lasts only a few years, if that, before becoming abandonware. Of course a lot of opensource projects go that way no matter the implementation language... So anyway, before I wander further off track, I don't think a two year old version of a langauge is obsolete by any sane criteria. The language developers seem to agree - not only 2.4 but even 2.3 has had a minor version release this year. I think this churn is a real issue - it's not an original thought, but consider how beneficial the relative stagnation of the Perl language seems to have been for its adoption. A language really shouldn't be a source of frequent new shinyness - it's role is to be the platform on which you build stuff, and stability in that base platform is a real help to application developers, at least ones who need to deploy to more than a handful of boxes they keep in a froth of constant updates. Consider again Django - it's changed amazingly in the three years between the initial public release and its long overdue 1.0 release, but I believe it still works with Python 2.3, as was originally targeted as a goal. I know for a fact it runs under 2.4 just fine, though I can't speak to the vast majority of the third-party apps out there...
Musings on python by a fairly long-term user. Posted Oct 21, 2008 0:19 UTC (Tue) by jamesh (guest, #1159) [Link] Why do you think it is bad for a language to improve over time? Often language improvements let developers do things they couldn't do before or let them express things more concisely (which usually results in fewer bugs). It has managed to do this while maintaining good source level backward compatibility. Python certainly isn't the only language to have seen improvements in recent years (generics in Java and C# come to mind). What you seem to be complaining about is libraries and applications that have made use of these features without thought for compatibility with older versions of Python. There are a number of causes for this:
I can understand you being annoyed at the second two cases, but just as often it is one of the first two. When I was working on version 2.0 of PyGTK, I moved the minimum Python version requirement up from 1.5 to 2.2 because I needed the new class system introduced in that release. It caused some short term pain, but was the right thing to do. Similarly, Django uses the decorator syntax introduced in 2.4 to good effect, which cuts off support for 2.3. At some point, the benefits of the new syntax just outweigh the downsides of requiring the newer version (where this point occurs depends heavily on what platforms you're targeting of course).
|
|||||||||||||