Slashdot Mirror
Gentoo Linux Musings
ChaserPnk writes "Gentoo has been in the news recently. First with the news that Daniel Robbins leaving Gentoo and then with Gentoo Linux 2004.1 being recently released. Have you ever wondered how Gentoo got started? An article at IBM DeveloperWorks explains how. Get to know the history of Gentoo."
darthcamaro wrote in with a related story that suggests that Gentoo is preparing to change directions soon: "Is Gentoo gearing up to be the third major enterprise distro? That's what an article running on internetnews.com points to. They talked to the head of Gentoo's enterprise efforts. For those that think that Gentoo Enterprise is far off, Gentoo's guy figures if they had the cash they'd be up and running in 6 months."
There is something very appealing about a distro that is so source-code driven (for lack of a better tem). It embodies all the best things about open-source software.
I have been extremely happy with Gentoo. It's rock-solid stable, and its the speediest of any distro I have tried (no doubt due to all your applications being optimized for your specific system).
If they came out with an "enterprise" version I think I would give it a whirl, I can see it easily being a great fit in my server room. I wish them all the best.
bash: rtfm: command not found
no.
I've been a Gentoo user for about 9 months, and it certainly has a promising future regardless of direction.
The portage system takes one of the best features of FreeBSD and actually *improved* on the idea rather than creating a poorly ported system. Decided you want to try out a few optimizations to see what your server likes best? Just 'emerge -e world' and you've got yourself a freshly recompiled system. Dreading the release of 2004.2? No sweat...Gentoo isn't like other distros (read: Redhat/Fedora) where upgrading remotely is a nightmare...just update the system through portage and it's essentially the same system. No need to worry about how you're going to upgrade your hosting servers to the newest release or worry that it will come to an EOL and you're no longer getting your juicy security patches.
It seems the most common complaint is the time it takes Gentoo to compile anything. The flexibility this system provides is well worth the extra few minutes rather than installing *.deb or *.rpm files and entering dependency hell.
Yes, yes...let the distro wars begin.
First, I'd like to start off by saying that I currently don't even have gentoo installed on any of my systems. I am not a gentoo zealot.
That being said, while I was reading the article posted earlier regarding Linux Useability I actually asked a few friends: Does Portage have a GUI browser/installer yet?
If it did, Gentoo could instantly be turned into the single most user-friendly distro on the planet. The primary problem with Linux (besides game support, etc.) is the ease of program installation. Imagine how easy it would be to code a pretty GUI to allow you to browse the Gentoo Portage Tree (which is already split up into intuitive categories) and install whatever you need.
Gentoo is a phenominal distro. It would take very minor amounts of tweaking to make it incredibly user-friendly.
Sig.i>
I believe however, that Gentoo is even better suited to this task. A fix in the source is a fix for every distro, where as a fix in a package fixes only a single release of a single distro.
With the recent release of catalyst, gentoo makes even more sense in this role.
I guess there are two knocks against Gentoo as a 'distribution base distribution': installation, and packaging. Honestly though, packaging -- once the source has been compiled once -- now works great. That's what the knockoff distros would be doing. Installation, they've left somewhat open-ended. Every distro seems to make an installer though, so I can assume it'd be easy to make one for a Gentoo knockoff.
Gentoo's source database is simply of the highest quality. I think it is the distro to watch, but because it is so useful as a technology to create truly customized, useful distros.
Actually, it's kind of odd thinking of Gentoo in an enterprise setting. It's always billed itself as a "meta distribution" in the sense that it's something solid distributions could be culled off of.
Due to it's ever changing and rotating nature, it's about dead opposite the rock solid Debian distribution. While it *could* be a Enterprise distribution, it'd be easier to create a solid locked branch built off Gentoo and kept clean of the nasty problems that tend to have (often) entered the portage tree in the past. And then it wouldn't really be Gentoo proper.
You're reading Slashdot. Of course you like Linux and pc hardware
Businesses want support, stability and a minimum of fuss, not exactly areas where Gentoo enjoys advantages over other Linux distributions such as, say, Red Hat, SuSE, Mandrake and Debian.
At the moment, it's not positioned to compete against the major distributions for a share of the business market. It may be so at some point down the line, but it certainly isn't so right now.
"Accept that some days you are the pigeon, and some days you are the statue." - David Brent, Wernham Hogg
I suppose the other valuable lesson, though, is that he did make it that far not just because of enthusiasm or hard work, but because he had a good idea (ebuilds). I see a lot of knock-off distros--yet another CD-based router, for example--that just don't have any great ideas behind them. Sure, that's the point of Linux--I've got no complaints with people doing what they want, but it strikes me that the valuable lesson here is that a good idea can go far--but without that idea, you've got nothing.
(That's the best I can come up with--just trying to focus the freakin' discussion on something other than ``I like Gentoo'' ``I don't!'')
I'm pretty new to Linux in general, but am not afraid of trying out something difficult or heavily CLI-based. I started with Mandrake and Fedora, but found them too bloated / Windows-esque for my taste, and am now relatively happily using Debian sarge, and have been eagerly awaiting its release. However, due to, er, some recent stuff, I'm getting slightly annoyed with Debian, wondering if the wait for Sarge might in fact be quite long, or indeed, interminable, and am looking at trying another distro. Gentoo looks rather appealing--it seems well-documented and so on, and looks like it might be pretty fun to set up. One thing, though: I have a dial-up connection. Is it possible/desireable to easily install Gentoo this way? I've got a fast connection at the University, and it seems [from reading the docs] that one can download ISOs containing binary packages built for Gentoo. But, er, doesn't that entirely defeat the purpose of installing Gentoo? Should I take the plunge? Is it a good idea to use Gentoo over dialup? I'd be interested if anyone has any thoughts on this.
Pros of Gentoo:
Customizability.
Speed.
Perfect for that old system you want to keep using.
If you are clustering, it probably would be the way to go.
Easy to update.
Cons of Gentoo:
Installation- un-believably frustrating.
Ever even seen Red Hat's system? Its SIMPLE and it WORKS. RH and Mandrake both can get my system to boot with grub on first install and boot, but nooooo, not Gentoo.
Too tweak-heavy.
I'm sorry. Gentoo is a great special-purpose distro. If it wants to go mainstream, it must have a better install system.
Go take a look at woody's(Debian) installer, and compare with the Gentoo install scheme.
Gentoo installations are crap.
I've done Gentoo, Debian, Red Hat 8, and Mandrake 10 installs.
Gentoo is the most difficult to install.
/b
|f(x)dx = F(b) - F(a)
From the "what-I-would-like-to-see" department :)
/rant mode :)
What I would love to see in Gentoo, or any other distro that is source-based really, is a way of setting up the system from binaries and then have the system transmogrify itself.
What do I mean? Well, after the initial install the distro could start to compile the optimised packages with a preset set of flags and "replace" the existing pre-compiled binaries as it finishes the optimisations.
Why? Well I think this would offer the absolute best of both worlds. It would allow you to get a Gentoo-based system up quickly without waiting hours and hours for compilation. It would then take advantage of unused CPU cycles (and lets face it, I doubt most machines use a large amount of resources more than 5-10% of their operating lives) to compile optimised packages, thus giving the benefits that everyone loves about source-based distros?
Is it possible? I have no idea. Frankly, I don't use Gentoo or even Linux all that often, but it strikes me as very neat solution for the one weakness present in distros that have to be compiled from source.
I think it might also be quite useful in getting acceptence in the business world. Being able to get a system up and customised quickly could be an important selling point, particularly in SME business where there is a diverse range of hardware (and thus ghosting is not necessarily a good option). It such a networked environment, it might even be possible to use a distributed compilation system.
Anyway, that's my little suggestion. As I said, it may not even be practical let alone possible, but it might stimulate further ideas that make Gentoo (and perhaps linux in general) an even better solution. Again, I don't even use Linux (well, only very infrequently) but I strongly support the underlying philosophy behind much of the OSS movement.
On the desktop end, I prefer gentoo because it is more lenient with accepting non-free packages and packages with potential legal issues. I also like the optimization abilities of a source based distro. As a java developer, Gentoo is simply the best Linux distro for Java developement. The major jre's are integrated in to the packaging system and the java-config utility allows me to easily switch from multiple jres on the fly.
On the server side, debian provides stability and quality control. Contrary to popular myth, there are quite a few pay support options available for debian.
Need to upgrade to a new version of some server software because there is a vulnerability?
Actually I almost always compile key stuff from source anyway, because I want to know that features I want are compiled in.
Need new software now? Ok, wait an hour for a compile
And if you, as an admin, take less than an hour to test your rpm (or whatever) software installation, on a mission-critical server, you're not doing your job. I will give you that it takes a long time to compile most things, but in my book, it's time well spent.
bash: rtfm: command not found
We're already using Gentoo on about a dozen or so production machines. Its been great. Setup time takes about two business days (system over night, bootstrap overnight), but who cares? We have the installation procedure we use down to the point where we don't even have to look at the screen, our self-made guides have everything written down. All the machines have a common configuration this way too.
I'm currently working on a web based system to very easily keep all these systems up to date and allow us to choose which packages we want to upgrade, so we don't have to get the newest if we don't want.
I hope they do release commercial support for it, we'd be one of the first on the list to purchase!
But in the server room? Sadly, I don't see this happening. What sells there is support. And for people who don't know, when we talk about someone providing support, we talk about someone to *blame*. "Hey, the server is down, wtf? Well, I'm paying RH $XXXX, I'll let them figure it out." And for the most part, they do.
The whole philosophy of Gentoo seems to go against this though. Red Hat can support it, cause they know you are running RedHat 7.2 with the 2.4.9-31MPT-SP kernel, cause that's what they shipped with. If you buiild your own they'll have one word for you: Unsupported.
Now look at Gentoo Linux, they are at the other end of the spectrum, 100% custom. Who in their right mind is going to support that? How could they? I just don't see it.
I think it's insane to reformat and reinstall a Linux distro every year a new version comes out.
The way Gentoo is set up, you never have to do that, ever. You upgrade as you go. Gentoo 2004.1 came out, but that's just the installation CDs...I installed using 1.4 CDs months ago, and I'm up to date as one would be if they installed this weekend (I love doing "emerge -upD world" and seeing what's new).
I see the future of Gentoo not only as a meta distrobution, but also as a comunication method between developers and users
The biggest thing about Gentoo for me hasn't just been the fact that I can get anything (fresh out of the oven), but the fact that I can report bugs, and get feedback within hours....I can go to the gentoo forums and get answeres within minutes
It's because I feel the future of linux is in its ability to progess, to find new way's of doing thing s, to find new......on a five year mission to .... well you get the idea.
Gentoo is just plain FUNNN!!!!
once more into the breach
There is already a binary flag for emerge, emerge -k. Certain packages (large ones like GNOME, KDE, Evolution, Mozilla, etc... and their dependencies) are provided by what's called GRP (Gentoo Reference Platform). All you have to do is set your PKGDIR in /etc/make.conf to point to a directory where you have these prebuilt packages (which you download an .iso of off a Gentoo mirror), and you're set. Although emerge currently has the capability to fetch prebuilt packages from a mirror that provides them, there are no public mirrors which do so. If you had a bunch of computers that you wanted to run gentoo on though, you could set up an ftp site with prebuilt binaries, point emerge to the ftp, and use emerge -gK to automatically fetch/emerge the packages you want. Otherwise you have to do what I said before, which is to get an .iso with all those prebuilt packages, and simply mount it.
We will take some random people in the following magnitudes and administer an OS test to see who's really king. Now I agree that Windows has a greater advantage because of market share, HOWEVER that's the real world and the one we play in.
30, 9th graders selected at random
30, Fresh high school grads
60, members of the general population
30, persons age 30-60
30, persons age 60+
30 small business _owners_ not in IT
FYI this is 210 people.
We will have them attempt the following tasks Using the latest versions of
WindowsXP,
RedHat,
Gentoo,
Linspire
OS X
Participants will be timed and rewarded with a prize if they succeed in their tasks, say a candy bar (to simulate a work environment where they would get money)
There will be two tasks to do 1/2 of each group will do each
The first half will have to complete the tasks without any documentation other than what is provided standard ON SCREEN.
The second half with a full printed manual including screen shots and detailed step by step instructions
Our tests will be
Install the OS (I realize this isn't realistic cause every Mac already comes with it but it'll have to do)
Create 5 users
Log in as one of the users and complete the following tasks
Write a complex document with some formatting and colors and save it as a HTML document
configure e-mail and send that HTML document to someone
make a spread sheet and save it to a location and upload it to a website
Users will have to find and install all the software to do these things either durring the OS install or from the Internet, they can make 2 phone calls durring the test
Then we'll see what OS is really easiest and fastest and cheapest, we'll assume these people all cost $0.002 per second... Meaning that the commercial OSes already start with quite an expensive handicap.
I'm sure with some more time and thought one could make this more fair but I personally expect OSX (Followed by Linspire) to win the on screen only event by a wide margin even considering the heavy price tag of the OS (we'll just assume a PC that costs as much G4 to level the feild) Most of us have seen a newbie use OS X and it's almost like they know what their doing..... For the well documented test I would expect Linspire to win followed by RedHat.
Now test could be expanded to setting up a small office network typical to a small business, I once again expect OS X to clean up
the most common complain (and mistake) about gentoo is that it "takes forever compiling" etc etc yadda yadda....
:)
:) i would never, ever trade it for other distro....
this is BS....
first: I have like 20 servers running gentoo, the oldest of them is a pentium3-1ghz...
even on this machine mostly everything compiles just fine (doesn't take long).
2nd: for the things that WOULD take a lot to compile on this hardware, I can always resort to the binary packages (emerge -k)... kde/openoffice/gnome/etc gets installed in seconds....
3rd: most my servers don't need kde/X/gnome/etc...
4th: if there is a package i use often, and it's not avaliable as a precompiled package... i can just have emerge "create" one and store it on the network... if i do an emerge things get compiled from source... if I do emerge -k , the portage will first look into my packages dir to see if it finds a precompiled version, and if it does... use it...
5th: distcc is your friend... i have 5 xeons 3.06ghz on my distcc farm... talk about fast compiles
6th: gentoo rox
I for one, welcome our new hot grits... PROFIT!
I love gentoo but still wouldn't run it on a critical server because of the compile demands.
I feel an enterprise version of gentoo needs some sort of master compiling server that can build binary packages (perhaps optimized for each arch in the company). That way, every 90 days (or whatever period, the IT department can build a 'cutting-edge' stable release and subject it to their quality control procedures.
Once it has passed, they need to produce the binary packages, and every system in the company can then emerge those (binary) packages on a nightly basis.
It doesn't make sense to have all your workstations and servers compiling everything for themselves.
Enterprise users [are lazy people with no time to do anything with their computers]
What crack are you smoking? It's the enterprise users who have stuff that's so mission critical they buy Windows source code from Microsoft and do stuff with it. You know, because tons of money is at stake.
It's the poor schmoes with three computers and a network hub that just want to plug things in and make it work, because one person-week wasted is a significant percentage of the company's time.
You obvious have no clue. If the enterprise users can score any kind of 10% improvement enterprise-wide with merely a few thousand man-hours invested, that's a good deal.
Is that really the point of Gentoo though? Gentoo is a META distro, NOT a full distro. If you read the handbook, there is a lot of emphesis on making Linux YOUR way. What's to stop an IT department from taking Gentoo as a basis, and implimenting it to the demands you specify? Gentoo already has a way to lighten the compile burden through distributed compiling, you can already set up your own portage system in a corperate style intranet for easy, fast access to packages, that, surprise surprise, can be hand picked by the head IT guy.
The bottom line is Gentoo is about CHOICE, thats why the things so damned hard to install for newbies, (even with the wonderful Gentoo Handbook).
For myself, I've tried other Linux distro's and have run into a lot of frustration when those distro's don't follow a regulated norm, (IE they liked to make up directories not specified in a program original make file, or other stupid things like that.) Gentoo is the closet you can come to a LFS (Linux From Scratch) system, except with Gentoo you have a way to deal with installed packages in a semi-organized fashion, (instead of having to remember every little tid bit about where every single file, etc is stored, so you can hunt it down to delete it so you can do that upgrade you want).
I think we should level the playing field just a tad. First of all, if you're taking that broad of a selection of people (well, it's not THAT broad but broad enough for testing), you have to consider that a decently high percentage of said people have trouble installing a program on their computer and that's with propmting, so, I would personally ditch the OS installation idea right off the bat.
Next up, creating the users. I can count on my hands how many people on their home based computers actually have more than one user created on their systems (and this is regardless of OS), let alone 5 users. If we're going to keep this real world, we have to look at real world situations.
Third. The setting up of email is a good one. Everyone basically has to do that at some point and time (except people using AOL basically) so that is a good test. Another good one would be setting up the internet connection, and I am talking about making the people setup a dialup connection. Broadband is cheating in some respects and a bit more difficult in others.
Fourth, navigation of the OS/GUI. Make them find various programs and give the location. Nothing really obscure, but make them have to use the search functions of the OS/GUI. This will test how well the various OSs handle searches and how intuitive they are to people (if you're wondering, I'm looking thoroughly and only at usability here).
Fifth, ask the users to create a folder in a given location, create a document to put into it, save this document to the removable media of your choice, and hand it to another person to open. This will test interoperability between platforms/programs. It is cheating to put the same Office Suite (hell, leave out the office suite, just use supplied text editors) on every system, regardless of availability.
I can go on and on with this and I am seriously going to try and carry out these tests in the not too distant future. Some of these things I would use to gauge how well students were comprehending what I was teaching during Linux and MCSE courses. Others are jujst ramblings off the top of my head. hehehe Anyways, it's time to eat and I'm hungry...
CliffH
sigs are like a box of chocolates, they all suck remove the underscores to email me
I fail to see the benefit of gentoo in a work environment. In my experience, it requires nearly as much tinkering to "get working right" (ie, trying multiple package versions) as LFS. emerge simple streamlines some of the steps normally taken with LFS.
Some serious shortcomings in gentoo besides the above mentioned which make it inadequate for such a task:
- It's time consuming to install. Time is money. Companies don't like spending money if they don't have to.
- emerge doesn't do dependency checking when removing packages. For example, if I accidentally remove libc instead of glibc (for example), I've just fscked myself.
- there doesn't appear to be any significant review process as Debian and RedHat has in terms of stability - Debian in particular. For instance: Someone used the fact that gentoo only requires the updating of the source code to update all gentoo machines. This isn't a good thing - it doesn't allow for a sufficient review of the code to make sure that there aren't serious problems with it. Contrast that to the armies of reviewers that debian has - even to the relatively new packages which are currently in sarge.
My personal experience with gentoo is that it's too much of a hastle to install - only marginally more irritating than LFS. The only reason to do LFS, IMO, is if you're an anal retentive control freak, have some sort of philosophical bent, or you're doing it for the learning experience - once.
I do know experienced users that use gentoo, however the majority of them are of the "I used Redhat for a short while, it sucked and broke a lot. Then I used slack, because it's leet, and now I'm using gentoo because it's leeter." Not many of them have even tried debian; several that I've convinced to try debian have started to turn their backs to gentoo to some degree. Nearly all of the people that I'd trust to babysit my servers run either debian predominantly or run multiple distros and have experience with all of them. I'd likely not want to work with someone that's so reckless to put such an untested system as gentoo in a critical role.
~/ssh slashdot.org ssh: connect to host slashdot.org port 22: too many beers
If you are having touble installing gentoo or does not understand the handbook manual on the oficial webpage of gentoo i recommend you to take a look at the anaconda-gentoo graphical installer that victor padra make just go to: http://gentoo.vidalinux.com/?q=node/view/35
From here:
"[M]y new machine wasn't very stable.
Obviously my first reaction was to go back down to 2x366Mhz. But now I experienced an even stranger problem. As long as my machine kept the CPUs chugging away, the machine didn't lock up. But if I left the machine idle overnight, there was a good probability that the system would lock up completely. Yes, an idle bug -- argh!"
And thus Gentoo was born: as a way to prevent idle bugs by keeping the CPU active 24/7!
Posters recognized by their sig,
Thanks for coming out! </flame>
And if you, as an admin, take less than an hour to test your rpm (or whatever) software installation, on a mission-critical server, you're not doing your job. I will give you that it takes a long time to compile most things, but in my book, it's time well spent.
As much as I love Gentoo, this comparison makes no sense. When you're compiling, you're still not testing your software installation. You're sitting there waiting for the installation to complete. The Redhat administrator can install and test his installation before you're even done compiling.
And who said anything about a mission critical servers? Most people aren't running mission critical servers, especially Gentoo users. They're tend to be more desktop oriented, so you're talking about a niche market. If you're running a mission critical server, you always have a completely seperate box to test new software and there's usually no rush to upgrade, barring security updates. In fact, I'd venture to say you may be more vulnerable with Gentoo on a mission critical server because you need to take the time to compile, leaving your exploitable machine open while your Redhat friend took 15 seconds to install the latest security update.
All that said, I love Gentoo. I use it on my desktop. It's by far my favorite distro. I just realize that Gentoo's approach isn't flawless in itself, and that compiling everything doesn't always make sense. But for myself, I like it.
If you're doing this you probably have a test server which is identical to your production server (at least if you have a lot of servers which are essentially clones). In any case, that is the only way to be sure you won't have dependancy problems when you deploy to production.
In that case while you are doing the initial compile on your test server you do an emerge -b , so that it tar's your binary files. Then just distribute the tar to your production servers once you are happy and do an emerge -k. You can skip the distribution step if you have network mounted filesystems (which would probably be the case in an environment like this).
Gentoo supports binary packages as readily as source-based ones, just not by default...
No looking for parts here and there, just "emerge mplayer" and BAM! It's all there, working great with all the codecs in one shot.
Fast and clean. Gentoo rocks.
"And a voice was screaming: 'Holy Jesus! What are these goddamn animals?'" - HST
Why do so many of the gentoo-zealots only mention the speed optimizations as a reason to run gentoo? :). The speed improvement (that atleast I haven't noticed) is just a bonus.
The reason why *I* run gentoo is because it allows me to very easy customize the features of all packages. Also it very easy to apply custom patches and still have the package managed by portage (therefor knowing when there is a new version out
--
In soviet russia gentoo compiles you!