Not logged in
Log in now
Create an account
Subscribe to LWN
LWN.net Weekly Edition for May 23, 2013
An "enum" for Python 3
An unexpected perf feature
LWN.net Weekly Edition for May 16, 2013
A look at the PyPy 2.0 release
Fedora and long term support
Posted Oct 17, 2008 21:02 UTC (Fri) by kh (subscriber, #19413)
I really wish Redhat well, so please don't take any of this as anything but my attempt at constructive feedback.
Posted Oct 17, 2008 21:07 UTC (Fri) by ceplm (guest, #41334)
Posted Oct 17, 2008 21:19 UTC (Fri) by kh (subscriber, #19413)
Posted Oct 17, 2008 21:31 UTC (Fri) by jspaleta (subscriber, #50639)
Posted Oct 17, 2008 23:56 UTC (Fri) by smurf (subscriber, #17840)
It had what looked like a LTS version because of its rather long release cycle, but it seems that the Debian maintainers are fixing that.
Posted Oct 18, 2008 13:35 UTC (Sat) by kh (subscriber, #19413)
I put Ubuntu on my son's computer, then my notebook, then transitioned a couple servers. It was a sometimes painful experience. (e.g. I realize many people would call me odd, and for a number of reasons, ;-) but I really prefer sendmail, and I ran into Ubuntu specific bugs in sendmail, and then others in postfix after I gave up on fighting the sendmail battle during my first server trials with early versions of Ubuntu.)
With all that said, I would not consider Debian now because I want a commercial company to do well in the Linux business. Actually, I want many to do well, and Redhat is one of my favorite companies - I really do not wish to criticize Redhat, it's employees, or Fedora. Their products just do not quite fit my needs right now. I still recommend RHEL, and have helped other companies deploy it. I do not think Fedora is appropriate for production servers though, unless the admins have a lot more time and energy than I do.
Posted Oct 18, 2008 15:25 UTC (Sat) by dilinger (subscriber, #2867)
By all means, if you're looking for commercial support, go with Ubuntu (or RHEL).
Posted Oct 18, 2008 3:14 UTC (Sat) by jengelh (subscriber, #33263)
Ubuntu DOES create something
Posted Oct 18, 2008 6:49 UTC (Sat) by khim (subscriber, #9252)
Sorry, but Ubuntu DOES create somethig. Yes, they pull a lot of stuff
from Debian repositories, but they DON'T send you to Debian mantainers to
fix common bugs and they DON'T close bug as WONTFIX with message "this is
as in RHEL so we can not do anything about it". If you reject any
distributions which reuse work of others then you should reject them all
and go with pristine sources...
Re: Ubuntu DOES create something
Posted Oct 18, 2008 19:19 UTC (Sat) by jengelh (subscriber, #33263)
Yes they create something, here is what...: an unnecessary indirection. 1. Replicating bugs from Debian (as unfortunate as Debian bugs are), 2. adding more bugs not present in Debian, 3. bad bug management and 4. bad user impression.
We all probably agree that distros are a good idea, because you, as a user, possibly do not want to do everything by hand. (Even Gentoo is automated.) There is however a difference between doing it and doing it wrong.
1. Now all Debian bugs are REPLICATED. I will ignore the OpenSSL debacle, but consider for example the long-standing lack of UTF8-by-default in earlier Debian versions. This propagated, and now I am stuck for years with that issue in enterprises having made the idiotic choice for Ubuntu6LTS. (Is it because of the LTS sticker? Failure to see that Debian has a stable that is at least as good? Or is it blindly following the Ubuntu hype.)
2. This indirection introduces ADDITIONAL bugs, example here is the drop of privileges in the login process after the user was authenticated. (Ubuntu6; it eventually got fixed in later ones.) This horribly broke the PAM session closing code because some modules relied on having the superuser privileges that all other major distros retained.
I agree with you that they should not send the user to Debian maintainers after all, it might be an Ubuntu-specific bug.
3. Consider a relatively small package that nothing depends on and which has a bug. Commonly, sid is updated with the new version and users will usually pick the package from sid, silently, to get their bug fixed.
You are absolutely right: I just looked at the package I was secretly implying. Of all the bugs filed in the last 6 months, they DID NOT SEND bugs to either Debian or upstream, causing at least one bug that I would mark as confirmed to be missed. They also do not close as WONTFIX (just as you said), leaving bug reports TOTALLY UNANSWERED instead!
4. As a result, users continue to hit the bug and complain in forums. Many of these users cannot distinguish between (a) it is a newly-discovered bug, (b) bug is fixed upstream, but Ubuntu does not update it, (c) bug in the Ubuntu package (see item (2)). This puts the software into an incorrect bad light because users just claim xyz not working, software sucks.
Posted Oct 20, 2008 5:48 UTC (Mon) by ekj (guest, #1524)
Any derivative always risks introducing own bugs. Nobody is immune to this. It's not as if Debian, Redhat, or anyone else is free from introducing bugs not present in upstream. (you yourself mention the most prominent recent example, then say you'll 'ignore' it. That's fine but it doesn't make it go away)
Any derivative that DOESN'T change things duplicate the bugs in upstream. Duh !
As for the "bad user impression", did it ever occur to you that people use Ubuntu for reasons OTHER than that it gives them a bad impression ? And reasons -other- than being clueless ?
Frankly, your condensing tone doesn't in the least make me sympathethic to that part of your complaint that is legitimate. (namely that communication and integration with upstream could definitely be better)
It also misses the mark. I've run Linux since I installed the 1.2.13 kernel coming with slackware whatever-it-was from a CD-ROM that was stilled divided into "floppy sets", I've run and administered hundreds of linux-machines of 3 architectures on half a dozen or so different distributions. Today I run ubuntu on all my personal boxes, except one that runs a different flavour every month that I use for testing and playing around.
Ubuntu has, in actual fact, managed to get something right. Refusing to allow that doesn't make it go away either.
And the tone you, and many like you, are using isn't likely to actually make anyone sympathethic to your views.
Posted Oct 20, 2008 10:06 UTC (Mon) by tialaramex (subscriber, #21167)
Ubuntu is terrible at doing half of what a distribution is supposed to do (mediating between users and upstream). That's not the fault of Ubuntu users, it is Canonical's fault. And the Ubuntu fanboys respond to every criticism with "But Ubuntu is popular" as though somehow "popular" makes it all OK.
There is no reason Ubuntu can't fix their mess and stay popular. There's no reason to believe that people who don't like Ubuntu because it is /causing them problems/ will still not like it if those problems are fixed. Characterising everyone who doesn't agree with you as a "half-troll" doesn't fix anything.
Actually that's not true...
Posted Oct 20, 2008 11:01 UTC (Mon) by khim (subscriber, #9252)
Ubuntu is terrible at doing half of what a distribution is supposed to do (mediating between users and upstream)
But Ubuntu's business model practically depends on this. If you'll look at Ubuntu's site you'll find that they sat that "every computer user should have the freedom to download, run, copy, distribute, study, share, change and improve their software for any purpose, without paying licensing fees" - and they follow on this promise. Yet they don't say "every computer user should have the freedom to whine to us and we'll help him". This feature is sold separately. How many users who complain that bugreports are ignored and communication is one-sided are actually paying for support?
Ubuntu quite explicitly separated these two issues - it has nothing to do with being popular or unpopular. Since Ubuntu is popular they can offer you palliative: ask users for help - sometimes it helps, sometimes does not. But if you want to have non-zero priority for your bugs - you should pay for the privilege. Looks pretty logical to me.
There is no reason Ubuntu can't fix their mess and stay popular.
But can they fix their mess and still be profitable? That's the question.
There's no reason to believe that people who don't like Ubuntu because it is /causing them problems/ will still not like it if those problems are fixed.
But there are reason to believe that people who are happy with free support will not bother to pay.
P.S. I'm not saying Canonical created this problem on purpose. But since they have no incentive to fix this problem and every incentive not to... Why bother? IMO sensible approach for them will be to ignore bugreports for most of the cycle then early after release check if they still persist and fix them if they do for the next version of Ubuntu. Unless these are security-related problems, of course.
Posted Oct 20, 2008 12:33 UTC (Mon) by ekj (guest, #1524)
There's a large difference between harm and too-little-help. The real critique of it has been mostly the second variety; not ENOUGH is done to ensure bug-reports are efficiently communicated upstream, and bugfixes make their way upstream etc. I can agree with this. But that doesn't mean the hyperbole. I don't see anyone pulling anyones hair, frankly.
I didn't say being popular made anything ok. I said being popular may sometimes be a sign of actually getting some of the things that matter to USERS right.
I do agree it should be possible to do right onto users and at the same time ALSO do right unto upstream.
The trollish aspect is the typical open or hidden claim that anyone who uses Ubuntu is somehow clueless or harmful or doesn't know his own best. That's frankly insulting, and certainly nonconstructive.
I've run Slackware, Debian and Fedora for years. Frankly I don't think which distro I ran made much difference at all to how useful (or not!) I am to the overall Linux ecosystem. The few things I *did* trough Launchpad (some minor translation-work and reporting a single bug) are both present upstream today (in Gnu nano and nautilus-actions), so though things aren't perfect obviously some stuff is working some of the time.
In short, "ubuntu should do X" is constructive. "ubuntu-users are idiots, ubuntu is similar to the bully you knew in highschool" isn't.
Posted Oct 20, 2008 15:30 UTC (Mon) by sbergman27 (guest, #10767)
I'd love to have an answer to this question, as well. People whom we otherwise respect go ballistic to attack Ubuntu. I've already mentioned this in the current thread. But I really do think that Morrisey hit it on the head way back in 1992. We Hate It When Our Friends Become Successful:
Posted Oct 21, 2008 2:09 UTC (Tue) by joey (subscriber, #328)
Debian is a vast collection of software. For a long time, much of this software had issues with utf-8. Debian and others performed a lot of work to fix various things to work with utf-8, and it eventually reached a point where Debian decided to enable utf-8 by default, since not too many things would break. (Although the number of utf-8 issues that remain is not exactly anywhere near nonzero... I seem to run into about 1 per week.)
You make it sound as if Debian perpretrated a lack of default utf-8 on the world somehow, which is a really strange reading of the above facts.
Furthermore, if Ubuntu chose to base off of a version of Debian that did not use utf-8 by default, there's no particular reason they couldn't revisit Debian's decision. Since the part of Ubuntu distributed on CD is about 21 times smaller than all of Debian, they could look at that subset, and, if it supported utf-8 well enough, make the switch. This ability to move quicker could be construed as one of the advantages Ubuntu can bring to its users. I don't remember if they significantly leapfrogged Debian in enabling utf-8 by default or not, but they certianly *could* have.
(Good greif, did you actually coax me into *defending* Ubuntu?)
Posted Oct 21, 2008 2:32 UTC (Tue) by jengelh (subscriber, #33263)
Hell I am not even talking about the vast software, but the base system. Simple stuff like the terminal being in utf8 mode and $LC_CTYPE being sth. like "en_US.UTF-8". I think it was Suse and Fedora that were among the first that I remember had utf8.
But you can also look at it another way if you don't like me pointing at Debian. Look at Slackware. If Debian is your standard definition of when features are complete and should be activated, then Slackware is quite late in enabling utf8.
>there's no particular reason they couldn't revisit Debian's decision [...]they could look at that subset, and, if it supported utf-8 well enough, make the switch.
Well that would have been a *real merit* if they did. But they did not do squat.
Posted Oct 21, 2008 8:22 UTC (Tue) by jamesh (guest, #1159)
Ubuntu has defaulted to UTF-8 locales since 5.04 – the second release ever made.
If you have a box that has old non-unicode locales set up, it'd be interesting to know how it got into that state. If you want to fix it you could try the recommendations in the 5.04 upgrade guide.
For all the Ubuntu boxes I've set up in recent memory, unicode locales have been selected by default.
Posted Oct 17, 2008 21:34 UTC (Fri) by sbergman27 (guest, #10767)
Bleeding edge. Less well tested. (e.g. in F8, gdm defaulting to not allowing more that 16 users and then ignoring its config file, for several months before a fix was in place.) Have to upgrade every 7-13 months.
By the time you get the software it's already old. Then you get to live with it for, nominally, 18-24 months. Longer in actuality.
It is (unofficially) possible to "side-grade" between them. But to go from Fedora to RHEL/CentOS you can only do it by hanging back (as much as you can without sliding out of the security update window) on your Fedora upgrades until a window opens in which the new RHEL/CentOS packages are newer than your current Fedora packages and you can do it.
Ubuntu gives you a different choice:
Reasonably well tested release. Current, but not bleeding edge. At any time, you can choose to upgrade or not. If you upgrade to a non-LTS release, you can upgrade or skip the next 2 releases (12 months), but must upgrade by at least the 3rd. (18 months).
If you choose only LTS releases, you can go 3 years between upgrades for desktops, and 5 years for servers. You can, of course, upgrade to a non LTS release at any time.
Posted Oct 17, 2008 21:57 UTC (Fri) by jspaleta (subscriber, #50639)
And that right there is the challenge to rise to. Can this be handled better? Is there the community interest to make room for a better way to do that transition? Can things be more structured to create better understood transition windows? Quite frankly is a totally unexplored area of discussion.
Posted Oct 17, 2008 22:54 UTC (Fri) by sbergman27 (guest, #10767)
And I'm quite glad you are exploring it. But I'm not sure that simply providing a more serviceable corridor between Fedora and RHEL will get you the same thing. With the Ubuntu upgrade path you have a continuous thread you can climb, in a release by release fashion, or in a skip one/take one/skip three/take one fashion. But each release is intended to be production ready. The support schedule is what varies. With Fedora/RHEL you have a somewhat larger gap. Fedora (as a technology showcase) can say "Hey, Look! We've just dumped cool new Feature X into the distro!" (crash). Meanwhile, RHEL/CentOS are chugging along with very serviceable and reliable, yet somewhat behind the the curve, offerings.
Some of us would like choices that a little more in between.
Posted Oct 17, 2008 23:03 UTC (Fri) by jspaleta (subscriber, #50639)
There's what people want, and there there is what is sustainable...I'm only interested in exploring ideas which have a potentially sustainable future. I am not interested in figuring out how to make sure everyone gets the pony they really want. If your version of a purple pony happens to be something in-between, but ultimately unsustainable, then I guess you'll be as disappointed as I am every morning when I don't find my purple pony standing in my front yard.
Posted Oct 17, 2008 23:12 UTC (Fri) by kragil (subscriber, #34373)
Fedora cannot be deployed in production from my experience. When it is released it is way too bleeding edge ( buggy ) and after 3 to 5 months when most bugs have been squashed it has only 10 to 8 months support.
That is way too fast for almost everybody!
Ubuntu on the other hand really accels in this scenario.
That was the point.
Posted Oct 18, 2008 0:01 UTC (Sat) by sbergman27 (guest, #10767)
Indeed, he did. That's what people do when they've promised purple ponies in the past... and then some other distro actually delivers them.
Posted Oct 18, 2008 18:22 UTC (Sat) by smoogen (subscriber, #97)
Unless you are paying Canonical, and convincing other Ubuntu users to pay for LTS.. you will not see it last. The reason LTS exists is because Shuttleworth is paying for people to work on fixing old crap. When Shuttleworth runs out of money (which depending on how much he had invested in the Stock Market/Icelandic banks could be pretty quick).. the free ride is over and people will have to find someone else to fix the stuff because most developers don't like trying to figure out if some 5 year old program can be fixed against some new style of attack.
So my main statement is if you are for Ubuntu. PAY FOR IT if you want to PROMOTE IT so it will be there 10 years from now.
Posted Oct 19, 2008 1:29 UTC (Sun) by sbergman27 (guest, #10767)
While I certainly support the idea of investing in a distro one believes in, I don't appreciate the campaign of spreading Fear, Uncertain, and Doubt against Ubuntu. Regarding your specific concern, let me remind you that Mark's fortune was built upon knowing what to invest in and *when to get out*. He sold Thawte in December 1999. The dot-com bubble burst 3 months later.
I would like to see Canonical turn a profit. But the situation is not so dire as you make it out to be. If the worst should happen, and the paparazzi photos of Mark in line at the soup kitchen begin appearing in the supermarket tabloids (BTW, did you know that Britney is having Elvis' baby?) then the Ubuntu formula *will* be replicated. Because Ubuntu has clearly demonstrated a formula that attracts users better than the formulas used by other distros. Whether it is a formula that other distros want to follow is another matter. For example, it would be a terrible fit for Fedora. The Fedora devs want to present the bleeding edge of Linux development. They would never be happy cultivating a large user base. Some in the Fedora community seem to *think* that they would. But that is, in my opinion, a case of "be careful what you wish for".
Ubuntu attracts many users
Posted Oct 19, 2008 3:49 UTC (Sun) by vonbrand (subscriber, #4458)
So what? If the result isn't a self-sustainable business model, it is of no real importance. AFAIU, Ubuntu has yet to show a profit, even with its massive user base coupled with few full-time developers.
Posted Oct 19, 2008 4:52 UTC (Sun) by sbergman27 (guest, #10767)
And the even more massive "sour grapes" phenomenon it has generated among fans of the less popular distros.
If the result isn't a self-sustainable business model, it is of no real importance
Spoken like a true FOSS devotee.
Posted Oct 19, 2008 17:44 UTC (Sun) by vonbrand (subscriber, #4458)
If the result isn't a self-sustainable business model, it is of no real importance
Spoken like a true FOSS devotee.
Some way of paying for build servers, distribution, bug handling, ... has to be found. Last but not least somebody will have to cough up the paychecks for a few key people. If all that is just "provided by $MECENAS, until his patience runs out", it is not a sustainable model.
Posted Oct 19, 2008 20:45 UTC (Sun) by sbergman27 (guest, #10767)
I get the impression that the FOSS community (or is it the RedHat/Fedora community?) is really reaching here. Who pays those bills for Debian? I guess Debian is "unsustainable". And has been for the last 15 years. Yes, Red Hat has found a way to make money... lots of it... from FOSS. I applaud them. And I am pretty much in the Red Hat camp. Used RedHat Linux from 4.2 on, 1997, use CentOS today, yadda, yadda, yadda... But Ubuntu is an impressive new force which is causing people to sit up and take notice. I wonder why we are so resistant to it? Could it be that we hate it when our friends become successful?
Posted Oct 20, 2008 10:35 UTC (Mon) by tialaramex (subscriber, #21167)
But you can't say "I will take this distro's resources, and assert that it could be combined with this other distro's philosophy and produce this third distro's product" because there's no reason to believe that works.
The existence of Ubuntu doesn't prove that Debian style volunteers could produce a Fedora alternative but with long term support, any more than the existence of oil-rich Saudi Arabia proves that desperately poor Ethiopia could become lush green New Zealand.
Maybe that's not a great analogy, but hopefully it makes my point.
Posted Oct 20, 2008 16:20 UTC (Mon) by sbergman27 (guest, #10767)
Posted Oct 20, 2008 8:00 UTC (Mon) by michaeljt (subscriber, #39183)
Posted Oct 17, 2008 23:57 UTC (Fri) by drag (subscriber, #31333)
Well Debian gets close to that somewhat.
As you know there are three main branches of Debian at any one time.. Stable, Testing, and Unstable.
With Debian's package management system is fairly flexible when dealing with those sorts of things. You can optionally configure 'apt-pinning' which gives different weights for different branches.
For example you can configure apt-get to always pull packages from Testing, but if they don't exist in Testing apt-get will pull them from Unstable.
If you configure Stable + Testing + Unstable in your system you can give the greatest weight to Testing. This way you can avoid a lot of the churn that comes into play when using Unstable. That way if you want to use OpenOffice.org 3, for example, it's avialable for Unstable, but not testing. So you can pull that specific package down and give it priority.
This sort of thing is what I do to get buy with Debian. My work system uses testing and my home systems use unstable. I'll use approx (a package proxy) to cache packages so that I don't waste bandwidth on copying down multiple copies of the same packages.
The major suckage about Testing, however, is that directly after a release it's mostly useless. Unless your involved in development your far better off tracking stable for some time, after a release, before switching up.
Now if you want to track 'Stable' and backport packages it usually takes a bit more effort.
For doing that your better of just leaving your apt-get stuff configured to use stable only for binary pakcages, but track testing or unstable for source packages. Also tracking backports.org is a good idea.
This way you can recompile packages from source specifically for stable. Almost all packages should backport themselves with little to no effort. Sometimes you have to recompile some dependences, but that's not usually much.
This is because if you pull binary packages from Testing then it has the tendency pull many more dependences then it actually needs. So you end up with a hybrid testing/unstable/stable system, which defeats the whole point of running stable.
But by recompiling packages you get the ability to benefit from stable systems, but get the newer versions of software you need for your production system.
This sort of approach is largely unsuccessfull with Ubuntu, unfortunately. This is because Ubuntu is much less disciplined about packages and backwards compatability. The only reason it works as well as it does in Debian is just through brute force developer hours.
However even if you use pure Unstable system you won't be able to keep up with Fedora. Fedora's a developer's playground and has the cutting edge features before any other system.
Debian benefits from this hugely. If it wasn't for Fedora being cutting edge and thrashing out bugs and doing all that sort of work.. that work would fall to Debian which would consume massive amounts of resources and manpower. I don't think Debian, as a orginization, would be able to cope with what Fedora does.
Without Fedora (and Ubuntu) Debian would be much more of a mess to deal with.
(and visa versa.. without Debian working on multiple arches and making sure that all the diverse software options worked with one another cleanly (for example: With Debian I can equaly choose between using Sendmail, Exim, or Postfix without breaking stuff badly.. and I can choose equaly between KDE, Gnome, LXDE, or XFCE, and even others. Everything 'just works', mostly.) then Fedora and others would be worse off)
Debian mix and match, long term support
Posted Oct 19, 2008 0:02 UTC (Sun) by vonbrand (subscriber, #4458)
That is a lot of work for the user of such a mongrel... and the idea of using a distribution is precisely to avoid such labor-intensive setups. Besides, if you want rock-solid stability, you won't get any guarantee for such a mix-and-match setup.
On the distribution side, keeping stuff backward compatible (in a fashion) for who knows how long is a resource drain that I'd prefer see spent on moving forward.
Posted Oct 19, 2008 20:51 UTC (Sun) by drag (subscriber, #31333)
The way things work the packages are designed to work like this. In fact doing what I described works better then trying to be a 'pure' testing user. If you 'pin' packages so that you tell the OS to prefer 'testing' over 'unstable' but allow you to install 'unstable' packages then you avoid a lot of pitfalls that go along with package management.
And back porting newer packages from testing to stable, by recompiling packages from source packages allows you to use a stable distribution, but selectively upgrade the specific peices of software you need that may be too outdated for what you need. For example: if your tying to run a PHP website you downloaded off the net, but the mysql PHP packages from Debian stable are not new enough then it's easy to selectively upgrade those specific packages you need.
Basically: If you think this sort of thing is very hard to run and maintain then you have not done it.
Posted Oct 19, 2008 2:49 UTC (Sun) by sbergman27 (guest, #10767)
Yes. The Debian Shell Game.
"Debian's old and moldy compared to Distro X".
"No! It's not! It's cutting edge! Use Testing!"
"I couldn't get 'Testing' to install."
"Then try 'Unstable'!"
"OK. Unstable installed, but it has problems 'A', 'B', and 'C'."
"Well, I've just checked and I don't see a bug report from you. Why did you not complete your assignment?"
"I wasn't really thinking of it as an assignment. It's just that everyone was telling me how great Debian was and stuff.
"Well... I still don't see a bug report. You are making me vewwy angwy!!! Try Unstable NOW and file any bug reports!"
But now Linux has trashed my computer. I'm having to post from my girlfriend's Vista box. The word "GRUB" is covering my screen and scrolling up fast. I'm starting to get a little agitated myself.
"Well, it's called "Unstable" for a reason! You should have used stable if you wanted stable! Doofus!"
Posted Oct 19, 2008 20:53 UTC (Sun) by drag (subscriber, #31333)
At work I have friends that use Fedora, just because that is what they want to use. The amount of times that he has a non-working box (ie: something he depends on is broken) after a upgrade with Fedora is considerably more often then what I have to deal with on any of my machines.
Posted Oct 19, 2008 21:38 UTC (Sun) by sbergman27 (guest, #10767)
Friend or friends? You are not clear on that point. Yes, Fedora breaks a lot. RHEL and CentOS are substantially more reliable. The Fedora guys do play something of a shell game. Nominally, Fedora is Red Hat's alpha, or beta, or whatever you want to call it. Fedora fans will try to trick you into thinking that Fedora is production ready. After all, they are not the ones who have to deal with the problems.
In general, I prefer CentOS. But if my life depended upon OS reliability, and I could only choose between Debian and Fedora... I'd choose Debian. Without the constraints, I'd choose CentOS.
Posted Oct 20, 2008 14:46 UTC (Mon) by pizza (subscriber, #46)
Are you referring to upgrades from Fedora X to Fedora X+1 here, or are you referring to the nightly-ish "download and apply the latest set of updated packages to Fedora X"
From my experience, the latter breaking things is quite rare, usually due to an upstream bug, while the former is more common, but at the same time, is to be expected.
Posted Oct 18, 2008 14:47 UTC (Sat) by maney (subscriber, #12630)
It's a nice idea, but at least for the rules of engagement that were (are?) used for Dapper I don't think it's very useful. Two years (the next LTS, which is the only usefully supported upgrade path) is at least twice too long to wait IFF volatile things like Firefox don't get some sort of new-version during LTS lifetime option. My wife's machine ran Dapper from somewhere around the time of its release (may have switched her during the extended beta, a couple weeks ahead of release) until shortly after Hardy, and there were issues with both web-based and other services that had drifted past the support in Dapper's packages (Gaim, I think, but there was another that I can't now recall that's used by a bunch of the Distributed Proofreaders gang that was an annoyance) well before the next LTS came along.
Then again, Hardy was not all roses either - the biggest one was that they insisted on sticking Firefox 3 in, without proper versioning/alternatives support, at a time when it was in our opinions grossly unready. That's the price a project pays for leaving large swaths of functionality to be added by third-party extensions: if the extensions one relies on aren't ready, neither is the base project. Oh, and pulseaudio, which Just Worked on some machines and Just Didn't on others. That has gotten better since release, though my own desktop still had some issues until I switched the old Creative card out for a low-end Diamond.
On the server side I'm using Debian Etch and nothing else. I had a test server using whichever Ubuntu release was current around the time Etch came to fullness, but saw no advantage to it, and of course there was no supported upgrade path from the existing Sarge installs to Ubuntu. I suppose I ought to reevaluate that again now, but I doubt I'll have the time for it, especially as I've had no complaints about Etch for servers.
I'm not really happy with either the six-month or the two-year cycle for desktops; the former is just too damned much churn, the latter really is too long for the inherent rate of change in services one wishes to connect to. Note that I am NOT talking about corporate desktops here, so I don't expect any commercially-funded distro to pay the least attention to this...
footnote: when I talk about a supported upgrade, I mean an in-place upgrade that doesn't need a lot of manual fixup afterwards. Having been accustomed to this level of support since I moved all my machines from Slackware to Debian a decade ago, anything less is annoying. I don't expect perfection - if the upstream has drastically altered the config file format there's not much to be done about it other than leaving things in a safe state until the admin can revise it - but it's never acceptable to just dump old config settings. Debian has come awfully near this goal repeatedly; Ubuntu seems just a bit less reliable, but the desktop stuff has gotten a lot more voluminous since I switched to Ubuntu for those machines, nor have I been keeping a written score.
Posted Oct 18, 2008 21:23 UTC (Sat) by ceplm (guest, #41334)
That's the exactly kind of experience I am after -- long-term support means supporting even beyond the upstream end-of-support, and that means backporting. And backporting means *a lot of* work. Extreme case if RHEL2.1 which has IIAC kernel 2.4.9. But even RHEL3 (which is five years old, so Canonical would still support it) has kernel 2.4.2*. I really wonder whether Canonical will find volunteers to dig into this archeological excavations if the need arises.
Fedora\\\\\\ Ubuntu and long term support
Posted Oct 19, 2008 13:21 UTC (Sun) by maney (subscriber, #12630)
Frankly, if I had it to do over again I'd have stayed on the six month upgrade cycle - it's more frequent than I'd like to have to deal with fixing the inevitable rough bits (if nothing else, making sure none of the non-supported packages haven't gone missing), but that's less annoying than waiting two years for a supported upgrade path. And yes, I've experienced the unsupported way - easier just to reinstall IME. Been there, done both, neither is fun. etc-keeper may help somewhat with restoring local configs - remains to be seen.
Ttwo years seems overlong, and five years sheer madness, given the current volatility of the Linux desktop environment - you either destabilise by trying to introduce new versions/backports (which by definition has changed the behavior somehow, otherwise you not have done it), or else you're so moribund you get wiped for a fresh install of something usable. Now servers are a little different, but I can't say anything about using Ubuntu there - mine are all running Etch. :-)
Posted Oct 19, 2008 22:16 UTC (Sun) by ceplm (guest, #41334)
Posted Oct 19, 2008 22:20 UTC (Sun) by ceplm (guest, #41334)
Did you notice, that RHEL 5.2 included rebase of all substantial desktop applications (e.g., OOo and gecko-related stuff)? Of course, it is in CentOS as well.
Posted Oct 22, 2008 16:28 UTC (Wed) by jonasj (guest, #44344)
Being a Red Hat employee I know too little about Ubuntu -- could somebody share their experience how good LTS support of Drapper really is? How many packages were updated in the last year (or last six months)?
Copyright © 2013, Eklektix, Inc.
Comments and public postings are copyrighted by their creators.
Linux is a registered trademark of Linus Torvalds