FreeBSD vows to compete with desktop Linux (uk.builder.com) [LWN.net]

BSD bigots are really just Windoze users???

Posted May 13, 2006 2:44 UTC (Sat) by pr1268 (subscriber, #24648) [Link]

Gee, as passionate as I've perceived the BSD zealots are about {Free|Net|Open}BSD, I'm kinda surprised you'd mention they use Windows 99.9% of the time.

Do you have any substance to this claim? I'm not trying to pick an argument here; I'm sincerely interested in researching more about FreeBSD (even considering dual-booting it with Linux), and I'd love to know just how loyal the FreeBSD "bigots" are to their OS.

If you have any links, journal articles, or even if it's simply your experiences at work/home/academia, etc., then might you share? Believe me, I can't tell you how irritating it is when a FreeBSD worshipper starts flaming Linux. It's refreshing to even remotely consider that they might simply be part of MS's secret FUD army (okay, perhaps that was a little too harsh!).

BSD bigots are really just Windoze users???

Posted May 13, 2006 3:17 UTC (Sat) by einstein (subscriber, #2052) [Link]

This is based solely on my experience as a unix admin and linux user over the past 12 years or so. In every single case where a BSD fan is putting down linux because it's not leet enough, and I am able to check to see what this hard core BSD guy actually uses himself, he inevitably turns out to be some using ms windoze for his computing needs - I would encourage you to check it out for yourself and see what you find.

These people say that bsd is cool in the server room, but they would never consider using anything but ms windoze to listen to mp3s, browse the web or write papers, which has always puzzled me, since I've used unix-like OSes, especially linux, for everything for some years.

(In a similar vein, I've found the same thing about people who say "GUIs are for sissies", criticize me for using kde, and claim to use only a green screen themselves. I always ask them, since they are so eleet with command prompt, how do they browse the web, watch movies or play games? At that point, they seem flustered and admit that they use ms windoze for all those things - oops. So, what they really mean to say is, the tiny little bit of time that they do spend in unix, they have learned to do with a command line, and they are proud of it.)

BSD bigots are really just Windoze users???

Posted May 13, 2006 5:02 UTC (Sat) by rsidd (subscriber, #2582) [Link]

Either this is a troll or you move in strange circles. I have moved from FreeBSD to Linux, not because any software works better on Linux, but because Debian (now Ubuntu) were easier to keep up-to-date, and FreeBSD was going through bad times around version 5. Still, for a long time I used DragonFly BSD and still have it installed in a partition. And now I'd probably put FreeBSD on a new box, and may put DragonFly on a non-critical server.

Windows? I don't need it, it turns out my wife doesn't need it, my mother doesn't need it, nobody needs it except users of certain specialised technical software (NOT mp3 players or web browsers).

Nobody I know works on a green screen, but plenty use FVWM or suchlike as their window manager, and links or w3m to browse the net. They don't use windows.

I do know one man who uses linux, fvwm, a very bare-bones desktop -- and Microsoft Office, running under Crossover. I find that weird, but to each his taste.

Not surprising

Posted May 13, 2006 16:44 UTC (Sat) by man_ls (subscriber, #15091) [Link]

Not a troll, I have seen the same thing from time to time on *BSD forums. Whenever Windows is mentioned someone always confesses: "you know, I'm not a bigot, I use Windows too on the desktop". Maybe the situation is changing lately though.

Not surprising

Posted May 15, 2006 10:20 UTC (Mon) by anLWNreader (guest, #36915) [Link]

I second this. Most BSD users' preferred interface to their BSD boxes seem to be PuTTY.

BSD bigots are really just Windoze users indeed

Posted May 13, 2006 12:34 UTC (Sat) by gvy (guest, #11981) [Link]

I second that -- those who actually use *nix on desktop are less prone to criticize it groundlessly (even if they do have quite a few words to say on bugs and idiocracies).

Among the group I communicate with I'd note a few of those who are vocal but do actually use e.g. FreeBSD on a desktop (at the time I can firmly recall only a single such person), experience with shying down ueberfolks on a browsing question is quite the same.

*gasp*

BSD bigots are really just Windoze users???

Posted May 13, 2006 15:37 UTC (Sat) by tjc (guest, #137) [Link]

I've only known one FreeBSD user personally, and he was hard-core UNIX. I mean, his wife and kids ran UNIX, at home, whether they wanted to or not. Ya just didn't bring Windows in the house.

BSD bigots are really just Windoze users???

Posted May 13, 2006 18:26 UTC (Sat) by iabervon (subscriber, #722) [Link]

For years, I've been using text-based stuff for everything except browsing the web (which always vaguely annoys me by being badly behaved) and otherwise viewing images. But I've been doing it in X, because I can't see how anyone could possibly deal with only having one text-based thing on the screen at a time. I generally have two or three xterms visible, and a dozen iconified undecorated emacs windows. I certainly don't use Windows for anything; I don't think I've managed to do anything successful with Windows in a decade at this point.

For the hardcore texties: fb, twin, screen

Posted May 14, 2006 8:53 UTC (Sun) by kmself (subscriber, #11565) [Link]

I've been doing it in X, because I can't see how anyone could possibly deal with only having one text-based thing on the screen at a time.

While I use a lot of text apps, I do so under X11 (generally using WIndowMaker). Ion (a "rat-poison"-class window manager) might be another choice.

However if you want to go fully hard-core text-based management, you can. I'd choose a high-resolution framebuffer mode in my boot configuration, run the "twin" (text windowing environment), and make heady use of screen. Note that there are framebuffer viewers for both graphics and video, and there's always aalib (download and run the excellent 'bb' demo for it) for all the ASCI-Art jonesin' you could possibly want.

Having used a wide range of environments, I encourage people to use what works for them (and their hardware). That said, the text-based tools under Linux, from shells to terminals to filters and pipes to editors, all reflect a decades-old tradition of working with text, and a decades-long process of developing tools to do so quickly, easily, powerfully, and usefully. That's a really hard message to communicate to those who don't grok the commandline. But it's equally important to note that a good mix of graphic and commandline tools can deliver the best of both worlds.

It works even from a vt100: man 1 screen

Posted May 15, 2006 13:33 UTC (Mon) by hummassa (subscriber, #307) [Link]

SCREEN(1)
NAME
screen - screen manager with VT100/ANSI terminal emulation

SYNOPSIS
screen [ -options ] [ cmd [ args ] ]
screen -r [[pid.]tty[.host]]
screen -r sessionowner/[[pid.]tty[.host]]

DESCRIPTION
Screen is a full-screen window manager that multiplexes a physical
terminal between several processes (typically interactive shells). Each
virtual terminal provides the functions of a DEC VT100 terminal and, in
addition, several control functions from the ISO 6429 (ECMA 48, ANSI
X3.64) and ISO 2022 standards (e.g. insert/delete line and support for
multiple character sets). There is a scrollback history buffer for each
virtual terminal and a copy-and-paste mechanism thatallows moving text
regions between windows.

...

It works even from a vt100: man 1 screen

Posted May 15, 2006 15:50 UTC (Mon) by iabervon (subscriber, #722) [Link]

Screen is nice (in fact, I always have at least one screen session on my desktop), but it doesn't (last I checked) handle having 2 24x80 windows always at least partially visible (showing at least the bottom three lines), and having a dozen 40x80 windows, of which one or two are visible at a time. If nothing else, it's annoying to jump to windows beyond 9.

It works even from a vt100: man 1 screen

Posted May 16, 2006 14:24 UTC (Tue) by hummassa (subscriber, #307) [Link]

Screen is nice (in fact, I always have at least one screen session on my desktop), but it doesn't (last I checked) handle having 2 24x80 windows always at least partially visible (showing at least the bottom three lines),
Yes it does. ^A-S splits, ^A-Tab changes region focus.
and having a dozen 40x80 windows, of which one or two are visible at a time.
No problem here also.
If nothing else, it's annoying to jump to windows beyond 9.
I'll grant you that one ;-)

Green screen

Posted May 15, 2006 13:29 UTC (Mon) by hummassa (subscriber, #307) [Link]

I always ask them, since they are so eleet with command prompt, how do they browse the web,
lynx, links, or w3m
watch movies
mplayer -vo svgalib, or mplayer -vo fbdev
or play games?
nethack

:-)

Green screen

Posted May 15, 2006 18:43 UTC (Mon) by arcticwolf (guest, #8341) [Link]

Surely you mean mplayer -vo aa. ;)

Green screen

Posted May 18, 2006 6:43 UTC (Thu) by lysse (guest, #3190) [Link]

In fact, that was my default setup (not nethack, admittedly) until there were just too many webs(h)ites that no longer worked properly without functional JavaScript support. :( So now it's Netscape and mplayer -ao xv.

Green screen

Posted May 18, 2006 11:19 UTC (Thu) by hummassa (subscriber, #307) [Link]

elinks can use svga too, has functional javascript.

What FreeBSD needs

Posted May 13, 2006 0:12 UTC (Sat) by drag (subscriber, #31333) [Link]

Distro controlled repositories work out a bit better then that.

I don't know how to do it in fedora or using rpms, but in Debian it's pretty easy.

Say I want to upgrade to abiword aviable in Debian testing, but I am using Debian stable.

Well I just add the lines for deb-src for testing into my /etc/apt/sources.list

Then I do a apt-get update and upgrade to make sure that I am up to date. Actually I use wajig, which is a unified front end for all these debian utilities so I don't have to remember what command does what.

Then I go:

wajig build-depend abiword
to install all those dependancies needed to build it. Sometimes dependancies aren't aviable so I'd have to backport those also.
wajig build abiword

Then it produces the deb files for my compiled to use all of stable's stuff.

Then I install those and maybe run wajig fix-install to resolve any dependancies for me.

So backporting stuff is fairly straight forward most of the time for me.

What FreeBSD needs

Posted May 13, 2006 1:03 UTC (Sat) by elanthis (guest, #6227) [Link]

"Say I want to upgrade to abiword aviable in Debian testing, but I am using Debian stable.

Well I just add the lines for deb-src for testing into my /etc/apt/sources.list"

Yes. You build the entire app from source. Building any and all dependencies it might need.

Let's list the problems here:

(a) You had to some weird configuration directive that normal people shouldn't have to do, shouldn't have to know how to do, and won't do.
(b) You are potentially breaking your system anyhow due to the bad ABI practices in the OSS community; upgrading some build-dependency of an app might require upgrading all other apps that depend on that some build-dependency. Even Debian suffers from this despite their policy regarding library package versioning; some upstream libraries just can't be worked around with packaging.
(c) You had to build the app from source. That alone is just silly. Distro-neutral packages already exist for Abiword and many other apps. The *only* reason you have to build things from source in most cases is because of artificial incompatibiles caused *entirely* by the distro's aggressive, controlling packaging policy... except for the cases when the upstream ABI stuff kills you as in b.

And here's the big one:

(d) What happens when the upgrade and/or application you want isn't in the testing repository? Happens all the time. Jabberd 2 isn't in Debian testing. Gaim 2.0 beta isn't in Debian testing. snes9x isn't in Debian testing. Hell, the version of Abiword in Debian testing is still 2.4.2. That's just the first four problems I could find in a grand total of ~40 seconds of searching. Your whole "solution" falls apart because it still doesn't solve the problem of getting software your distro flat out doesn't decide to provide for you.

You can go the third-party repository route, but then you have all sorts of new problems:

(1) Do you trust the third-party repository? These people are not your OS vendor, and they are usually also not the upstream software vendor. Instead of just relying on the OS vendor and application vendor, you're now relying on a third party and hoping that they keep up on security updates and don't introduce malware.
(2) You have to find and manually configure these repositories. You can't go to Gaim's website and download the latest copy of Gaim. You have to go to Google, search up "'Debian 3.1' 'Gaim 2.0'" and hope you find something reliable.

We still have a UI problem with this and the main repository approach. The repository approach assumes you are working in one of two ways: either you know exactly what software you want or you find something usable given the horrendous crappy search interface the OS gives you to search through 10,000+ packages.

When I install new software after the system is setup, it is almost always because I see some blog post about it, or a reference on a mailing list, or come across some press release, and so on. i.e., I find out about it on the web somewhere. On any mainstream desktop OS, you can follow a link from where-ever you read about the software that takes you to the official website for said software which includes a download/install link (or a purchase option, as the case warrants). With Linux/BSD, the best you're going to get is a tarball (which is going to need a lot of dependencies and other tedious, redundant effort to get built and installed).

If you want to have something even remotely approaching convenience (and for those of us who don't spend every waking minute in front of a computer because we, oh, have other things to do with our lives, convenience is important), you have to use the distro's software repository. I refer you to the points above of the difficult of trying to find whatever you're looking for, if the distro even bothers to package it.

The repository model doesn't work for a desktop OS.

It's great for a "business" OS where a sysadmin manages the software installed on a couple thousand machines from a central location.

It's great for an "appliance" OS where you install the OS and use it "as-is."

That's what Linux distros, and FreeBSD releases, are right now: business appliances. You install the OS, configure it once, and the configuration (including available applications) really doesn't ever change for the lifetime of the box. If you want to go beyond that... break out the command line and the manuals, because you're going to need a moderately high level of knowledge and experience to get anywhere.

What FreeBSD needs

Posted May 13, 2006 4:29 UTC (Sat) by drag (subscriber, #31333) [Link]

It works for me.

I've back ported kernels. Kernel handling programs. I've back ported numerious applications, application libraries. I've back ported Xen. Backported OpenAFS and kerberos stuff.

All sorts of crazy stuff.

No problems at all. Never experianced any of the problems you pointed out. In practice recompiling packages usually works out very well if I need something in stable that isn't in there.

In fact I find the idea of having to use google or something like tucows to find some program to download downright distastefull. When ever I need to do something like this in Windows or in OS X I find the whole thing very unsatisfactory. *Sweet Jebus* have you ever had to find a f-king cd burning application for Windows XP that is capable of burning ISO's and actually WORKS?! That sort of crap makes me want to tear out my eyeballs so that I wouldn't have to look at another cripple-ware licensing agreement ever again.

And if you find some program that isn't aviable from your repository you just install it. Plenty of times I've done source installs and they usually go well. I've installed software on Debian using generic installers and that usually goes well.

I've got numerious commercial games. Stuff like 'marble madness' from garagegames.com, Unreal Tournament 2004, Cedega and Widnows games in there. I've got Doom3 and even 'Postal 2'. AND not only that I have all that crap installed in a NFS server so that on the couple computers I run I don't have to reinstall everything everywere.

I have a much higher track record of success in Linux then in Windows.

Seriously look at the whole host of problems Windows users have to deal with. I spend a fraction of the time in Debian or whatnot to get stuff to work then I do in Windows.

Screw all that crap. I'd much much rather have what I have now then have to deal with what I see people putting up with on these so-called 'Mainstream Desktop OSes'. Beleive me, I've also adminstrated numerious OS X machines and if you think that their install method is so much more wonderfull I've got news for you.

What FreeBSD needs

Posted May 13, 2006 5:12 UTC (Sat) by elanthis (guest, #6227) [Link]

"I've back ported kernels. Kernel handling programs. I've back ported numerious applications, application libraries. I've back ported Xen. Backported OpenAFS and kerberos stuff."

Awesome. Now go back to the part where I said my complaints where for Linux/FreeBSD as a general-purpose home-user-oriented desktop OS.

i.e., not for uber-geeks like you or any other LWN reader. :-)

What FreeBSD needs

Posted May 13, 2006 6:14 UTC (Sat) by drag (subscriber, #31333) [Link]

As it stands now it's not very user friendly way of doing stuff, but it's not that difficult either. Extending the repository way of doing stuff to make it easy for normal people wouldn't be that hard.

Plus normal people aren't realy going to give a crap weither or not they are running Abiword 2.4.0 vs. Abiword 2.4.4 anyways, if you realy want to think about it..

The thing that makes Windows easy in comparision is that there is only ONE ABI/API that Microsoft supports... And that's the latest one; Windows XP. At least for the desktop.

If Debian or Redhat was the only viable Linux distro out there then we wouldn't be having this converstation, right? Everybody would just make a version for that Linux operating system.

Well that is what people are dealing with with Windows.

As it stands now people running Windows ME and Windows 2000 simply can't run certain applications, especially games, because applications designed for Windows XP SP2 are not compatable with what they are running. Like from 'highlander': "THERE CAN BE ONLY ONE".

I garrentee you that within a year after Vista gets released people are going to be starting to be forced to upgrade for application compatability.

The same thing happens with OS X also. Plenty of software gets broken when people upgrade between between something like OS X 10.2.2 to 1.5.1 or whatever the latest version is.

I don't think that there is anything especially magical that Microsoft does with windows, except spending millions and millions of dollars testing out common user applications and putting all these special 'if then' and compatable cludges all over their operating system for backward compatability (most of which they are going to throw away with vista) that ensures that software packages and installers work across all their Windows systems.

They simply only have ONE operating system that they deal with right now and that's XP.

And even then look at all the applications that do borked up things like screwing up registry entries, install spyware and rootkits (aka Sony music players for instance), and require that users run with administrator privilages for the most asinine reasons.

In fact in the future I see Microsoft actually moving to software packages and repositories like what Debian uses with rules and dependancy tracking and all sorts of stuff like that. Except of course your going to have to put up with all this DRM restrictions and have your credit card handy when you want to use it.

Like what Valve does with STEAM and it's delivery system for various games and such that people can download from menus and such.

So I think that it's actually fairly likely that people are going to be moving more towards repositories then away from them, even in non-Linux operating systems.

What Linux needs is to have repository system and package management system that is compatable across multiple distros. More universal stuff. This is the holy grail as far as I am concerned.. It all should be aviable from menus. Maybe people working on things like SMART and as things like Portland and LSB/Freedesktop.org will figure out a way to get better compatability.

What FreeBSD needs

Posted May 15, 2006 2:46 UTC (Mon) by vonbrand (subscriber, #4458) [Link]

As it stands now it's not very user friendly way of doing stuff, but it's not that difficult either. Extending the repository way of doing stuff to make it easy for normal people wouldn't be that hard.

It is clear that you don't have the slightest clue of the problems involved in this... Distribution building (selecting packages, compatibilizing them, tracking upstream bugs and fixes while maintaining some coherency) is not easy. No, Gentoo is not a distribution, it is a bag of stuff.

What Linux needs is to have repository system and package management system that is compatable across multiple distros. More universal stuff. This is the holy grail as far as I am concerned.. It all should be aviable from menus. Maybe people working on things like SMART and as things like Portland and LSB/Freedesktop.org will figure out a way to get better compatability.

Can't have it both ways... one of the mayor strengths of Linux (or Open Source in general) is exactly the variety of software (and distributions) available.

What FreeBSD needs

Posted May 15, 2006 6:30 UTC (Mon) by drag (subscriber, #31333) [Link]

Umm...

I was talking about it in terms of End Users, not how hard it is to setup repositories by their maintainers or developers. Of course it's very involved to make sure everything works.

What FreeBSD needs

Posted May 15, 2006 19:34 UTC (Mon) by vonbrand (subscriber, #4458) [Link]

Coordinating the whole of open source software to work together, and then throwing in "I want package foo's very-bleedingest-alpha, but do not upgrade bar, while baz must stay at it (ancient) version" is a huge task, probably not even doable. In any case just not worth the effort. Either the user goes with a distribution that upgrades agressively (like Fedora), or stays with something more conservative (like Red Hat or its clones), or she goes for a distribution with everything that can somehow be compiled (like Debian), or a specialist distribution, or...

The beauty is that there are choices on each point of the spectrum here. And you sure can throw in the occasional built-from-source package (but please, go the extra mile to build your own package for your package tracker's sake; you'll be very thankful for that later) for kicks.

What FreeBSD needs

Posted May 13, 2006 15:08 UTC (Sat) by hein.zelle (guest, #33324) [Link]

I don't think this is a problem in most cases. People that are not "LWN uber-geeks" generally don't have requirements that get them into trouble either. Yes, there probably are notable exceptions, but I find that for most applications that "normal" desktop users want, there is absolutely no need to be running the latest-bleeding-edge version that gets you into trouble when using a central-repository based distribution. I myself have run into very few cases where I really wanted the latest version of something and had to do the upgrade the hard way. The only cases I can recall are things like amsn and skype, and those had statically linked versions available or a debian package that I could install without problems.

Now if you really want to be bleeding edge, what's the problem with running debian unstable (like I do on my home desktop) or perhaps the latest fedora, and keep things upgraded? I don't really see the problem in upgrading, especially on debian-based systems where this can be an almost automatic process.

What FreeBSD needs

Posted May 15, 2006 9:14 UTC (Mon) by philips (guest, #937) [Link]

> I don't think this is a problem in most cases. People
> that are not "LWN uber-geeks" generally don't have
> requirements that get them into trouble either.

You are or so very wrong.

Developer can live with bug in compiler. S/he's too lazy to upgrade compiler - and it's easier just work it around in source code.

End user, if s/he has encountered bug (*) in AbiWord X.Y and which is fixed in AbiWord X.(Y+1) - is left w/o options other than upgrading OS.

Compare time you need to:
(1) apply binary patch (as e.g. M$ Office does - and many other proprietary vendors)
(2) upgrade from binary package
(3) upgrade OS to get new version of AbiWord X.(Y+1)
(4) install development environment, install all required libraries, build new version of AbiWord

Options (1) and (2) are easy. But mostly unavailable. (3) is easy - but take hell alot of time and effort. (4) can be troublesome even for experienced developer.

From my experience, Linux end-users are opting for number 3. If one gets lucky - and proficient in hard drive formating - that can be very easy. But I have seen many normal users who just went away from Linux after several cases like that.

To conclude, end-user-wise, both Linux and FreeBSD are at moment bad. DO NOT TRUST SILLY REVIEWS ON NEWSFORGE "HOW IT IS EASY TO INSTALL DISTRO XXX" !!!! Several people claimed that BSD's ports are bit more user friendlier for application upgrades. Other have counter-claimed that Gentoo's emerge does it better. Binary package systems are very far from being really useful for end-users. I think only Debian got it right. And Debian is stable enough for second party applications (like KDE, OpenOffice.org) to maintain their own updates/upgrades repositories for Debian users.

You must remember, many many users' last experience with Linux is reading cryptic README and INSTALL files from source tarballs. The LAST experience. People install OS/applications to solve the problem they have - they do not want to be faced with even more problems.

(*) I mean bug like "Saving file with style XXX with the option YYY activated replaces style with ZZZ". Distros rarely provide such upgrades (only if users annoy on forums VERY much and forum admin is NOT Havoc Pennington) but would rather claim in next version "Greatly improved office suit."

Hard drive formatting? Why?

Posted May 15, 2006 17:09 UTC (Mon) by Arker (guest, #14205) [Link]

Hard drive formatting???

Pshaw!

apt-get dist-upgrade

Done.

And debian is generally *very* good about back-porting bug fixes anyway, so you only need to do that when there's a new stable release.

What FreeBSD needs

Posted May 13, 2006 16:21 UTC (Sat) by pyellman (guest, #4997) [Link]

elanthis wrote:

> Driver upgrading/installation that doesn't require installing a whole new version of the OS

and

>my complaints where for Linux/FreeBSD as a general-purpose home-user-oriented desktop OS
i.e., not for uber-geeks like you or any other LWN reader. :-)

I'm sorry elanthis, please explain to me why your non-uber geek home user is concerning him/herself with upgrading just a driver? Isn't that precisely the type of thing that we geek-type LWN readers do? Wouldn't your non-geek user be content with pushing a button that said "Update my system"?

Really, if you're going to take the approach you are taking, to be fair you should stick with one viewpoint -- that is, geek or non-geek -- not mix and match as you see fit to deliver your criticisms. Of course this can be difficult in practice.

FYI, when I am running an OS from a repo-updated source, I do not consider every update to be getting "a whole new version of the OS", but rather, simply keeping my OS up to date. Repository-based updating makes the "OS version thing" a thing of the past. Passe, that is.

Peter Yellman

What FreeBSD needs

Posted May 15, 2006 0:47 UTC (Mon) by elanthis (guest, #6227) [Link]

"I'm sorry elanthis, please explain to me why your non-uber geek home user is concerning him/herself with upgrading just a driver? Isn't that precisely the type of thing that we geek-type LWN readers do? Wouldn't your non-geek user be content with pushing a button that said "Update my system"?"

Add hardware. More and more non-geeks are learning about the joys of upgrading video cards and the like. In general, it's a really easy process. My computer-illiterate girlfriend installed a new video card in her machine a few months ago and didn't need an ounce of help. The hardware end is pretty straightforward. And on Windows, once the new hardware is in, you just need to boot up and pop in the driver CD and things are generally working in short order.

On Linux (and BSD), you install the new hardware, boot up, download the latest kernel, compile it, and install it. The download/compile/install sequence is far more complex for regular people than managing a few screws and swapping some cards. Working with tools and such is something almost everyone has some experience with. Typing in arcane command lines and comprehending screen after screen of techno-jargon is not.

"FYI, when I am running an OS from a repo-updated source, I do not consider every update to be getting "a whole new version of the OS", but rather, simply keeping my OS up to date."

Except that most distros actually don't *upgrade* the software in their distribution between versions. The only updates you usually get with most distros are bug fixes and security enhancements. If your distro has kernel 2.6.14, chances are you're going to be stuck on that kernel until you install a new one from scratch or upgrade to the next major release of your OS. (DaveJ says Fedora does do kernel upgrades, and he would certainly know better than anyone; Fedora is in a minority here, unfortunately.)

There's a good reason for not upgrading kernels, too. Especially in the case of a kernel, an upgrade could easily cause many users' boxes to stop working. A stable OS can't just push whole new kernel versions on users without *extensive* testing.

Unforatunately, because of the Linux ABI/API flux (which BSD doesn't suffer from nearly as badly), it's almost guaranteed that if you want a new driver, you have to upgrade the whole kernel, and any userspace libraries/apps that are tightly tied to the kernel (udev being notorious for this, for example). This is *NOT* a question of binary-only modules. I'd be perfectly happy if I could go to kernel.org and grab the latest GPL drivers and install them on a SuSE 9 box. Even if they're only distributed as source, with a stable kernel API you could entirely hide that fact (think DKMS).

More and more distros are getting "backport" repositories, but those only partially eliminate the problem. You still end up with a lot of out-dated and missing software in distros even just a few months old. How much time and effort is put into just packaging the *exact same software* in ever so slightly different ways to be installable on different distros/releases?

The only way you can avoid the lock-step OS-release model is to just follow development repositories. I've done that with Fedora, Debian/Ubuntu, Arch, and a couple others. I certainly never noticed a need to upgrade my whole OS, as you pointed out. I also noticed a constant influx of new bugs, radically changing desktop behavior, and other changes that I would never expect a regular user to put up with on a daily basis. It's a wonder _I_ still put up with it. ;-)

As another example of driver installation woes, take my new machine. It has very recent hardware, hardware that Ubuntu Breezy's kernel doesn't support. The only way to get Ubuntu on it at all was to install a Dapper flight CD. I doubt Windows would have installed out of the box either, but the difference is that with Windows, I could habe popped in a floppy or a second CD and pulled disk controller drivers in to use during the installation. With Linux, you don't have that ability; not unless you build an entire new custom install disc with your own custom built kernel. Even if you have the knowledge and skills, the time it takes to do that is ridiculous.

What FreeBSD needs

Posted May 15, 2006 6:37 UTC (Mon) by drag (subscriber, #31333) [Link]

"Add hardware. More and more non-geeks are learning about the joys of upgrading video cards and the like. In general, it's a really easy process. My computer-illiterate girlfriend installed a new video card in her machine a few months ago and didn't need an ounce of help. The hardware end is pretty straightforward. And on Windows, once the new hardware is in, you just need to boot up and pop in the driver CD and things are generally working in short order.

I know what your trying to say.. but have you EVER tried to upgrade a Windows box before?

I mean not just a video card (and I know that your going to have a 90% chance in Windows if you upgrade from a ATI to nvidia card or visa versa your going to run into huge issues), but the entire system?

Take a harddrive out of a old Windows box and stick it into a all new PC and see how far you get. Unless you put a large amount of effort into preparing windows it's simply NOT going to work. It's not going to boot, it's not going to run, and if you try to install all new drivers in safe mode (if you make it that far) it's going to completely fuck up what is left of your operating system. The only way it's going to work is if you absolutely know what your doing and go through a crapload of steps to prepare your system.

Try to upgrade your harddrive in Windows some times. It's a huge headache.

Now try the same thing with a modern Linux OS. It'll 'just work'. Worst case is that you'd have to futz around with the network drivers or video drivers. If your upgrading from Nvidia card to Nvidia card it will 'just work' also, unless your using old drivers taht don't support the new card.

Plus it's been a long long time since the average Linux has had to compile a kernel. Probably 80-90% of Linux users are using the kernel that is provided by their distro.

I know I am, and I am using a Pentium-D machine with a fairly new ASUS i945 motherboard.

What FreeBSD needs

Posted May 15, 2006 7:38 UTC (Mon) by lauwenmark (guest, #37754) [Link]

Add hardware. More and more non-geeks are learning about the joys of upgrading video cards and the like. (...) And on Windows, once the new hardware is in, you just need to boot up and pop in the driver CD and things are generally working in short order. On Linux (and BSD), you install the new hardware, boot up, download the latest kernel, compile it, and install it. The download/compile/install sequence is far more complex for regular people than managing a few screws and swapping some cards. Working with tools and such is something almost everyone has some experience with. Typing in arcane command lines and comprehending screen after screen of techno-jargon is not.

Well, how long was it since you last used a "user-friendly" Linux ? Because most desktop-oriented Linux distributions are nowadays provided with most drivers already available as modules (only those related to very old, or very exotic hardware are usually left out, but a 'home-user' is highly unlikely to need them). You plug the hardware in, and the various module autoloading mechanisms recognize it.

What about the case when the required module needs to be compiled ? Well, tools like the Debian module-assistant are now available - you don't need to rebuild a whole kernel, or fiddle with complex command-line tools.

Except that most distros actually don't *upgrade* the software in their distribution between versions. The only updates you usually get with most distros are bug fixes and security enhancements.

Well, for most distributions, switching to a new version of the distribution is no big deal - for a Debian-based one, it shouldn't be harder than an update of the package sources, followed by an apt-get dist-upgrade.
Besides that, there are also distributions that have no defined "version number" - taking again the example of Debian, "testing" and "unstable" have no such bounds and receive package upgrades on a regular basis (as you underlined yourself).

The only way you can avoid the lock-step OS-release model is to just follow development repositories. I've done that with Fedora, Debian/Ubuntu, Arch, and a couple others. I certainly never noticed a need to upgrade my whole OS, as you pointed out. I also noticed a constant influx of new bugs, radically changing desktop behavior, and other changes that I would never expect a regular user to put up with on a daily basis. It's a wonder _I_ still put up with it. ;-)

First, a "regular user" wouldn't feel the need to update its system on "a daily basis" for anything else than security updates. Second, I hardly saw a lot of "radical" desktop behavior changes during the last four years of using a Debian-unstable; roughly, it happens once each time a new major release of KDE (or Gnome) happens.
The influx of "new bugs" is something that always happens with new versions of a given software, whatever the way it is installed. This is mostly related to the quality checking of the software creator. Distributions like Debian-testing try to alleviate that problem, but there is no way to expect the distribution to cope up with all the application bugs by itself.
As for the "other changes", I really wonder what they could be.

As another example of driver installation woes, take my new machine. It has very recent hardware, hardware that Ubuntu Breezy's kernel doesn't support. The only way to get Ubuntu on it at all was to install a Dapper flight CD. I doubt Windows would have installed out of the box either, but the difference is that with Windows, I could habe popped in a floppy or a second CD and pulled disk controller drivers in to use during the installation. With Linux, you don't have that ability; not unless you build an entire new custom install disc with your own custom built kernel. Even if you have the knowledge and skills, the time it takes to do that is ridiculous.

First, that's mostly not a problem related to Linux itself, but to the hardware manufacturers - they refuse to provide drivers for Linux (or, for what matters, FreeBSD). There is nothing preventing them to provide a module source code supporting the commonly used kernel lines. Of course, you could say that "Well, but in 6 months, the kernel API will change, and the disk will be useless" - maybe, but in 6 months, the driver will have made its way into the kernel mainstream code, then back to the more recent distributions.

Then, I note that you downloaded and installed a Dapper flight. Well, did you have to build a custom install disc ? No. Did you have to rely on arcane knowledge to get a copy of Dapper ? Unless grabbing an iso image from Internet and burning it is considered arcane, I bet the answer is "no" as well. Basically, your problem goes down to "I got a piece of hardware that was so recent that I had to install the latest available version of Ubuntu". I hardly see how this could change without a more active support of the manufacturers.

What FreeBSD needs

Posted May 15, 2006 14:18 UTC (Mon) by mrshiny (subscriber, #4266) [Link]

Well, how long was it since you last used a "user-friendly" Linux ? Because most desktop-oriented Linux distributions are nowadays provided with most drivers already available as modules (only those related to very old, or very exotic hardware are usually left out, but a 'home-user' is highly unlikely to need them). You plug the hardware in, and the various module autoloading mechanisms recognize it.

Of course that only works for hardware that had a driver available at the time that the kernel your distro uses was released. Frankly I think it's not likely that distributions will have timely support for new hardware unless the hardware makers are distributing drivers for their hardware, and sample hardware, to the distro makers, otherwise there is no way that a driver writer at some hardware company can make sure the product is tested and the driver distributed to the customers.

It's also pretty much impossible these days to maintain drivers out of tree in any sane fashion since not only does the ABI change potentially with every kernel release, but so does the API. Here windows has a huge advantage for third parties: It is a stable target. The driver model doesn't change from minor revision to revision and changes little from major revision to revision. This requires more work for the NT developers but allows driver developers to maintain their drivers out of the NT source-code tree and it allows them to target all users of a Windows NT kernel version with a single driver. In Linux the ONLY solution for this is to get your driver into the main kernel sources, but that can't happen until the driver is finished, and likely only after the hardware is available. This means that a linux hardware vendor can not reliably target ANY existing linux users with his hardware, until they have updated their kernel to the latest version. This makes life easier for the kernel devs but DEFINITELY makes life harder for the users.

What FreeBSD needs

Posted May 15, 2006 19:49 UTC (Mon) by vonbrand (subscriber, #4458) [Link]

Re: Stable driver interfase in the kernel: Go take a look at Documentation/stable_api_nonsense.txt in your nearest Linux source tree (or go to the online version). There is a reason for not having a stable in-kernel API/ABI; besides, being open source (and having all relevant drivers in the same tree) a stable API/ABI is not nearly as needed as in closed-source systems.

What FreeBSD needs

Posted May 15, 2006 20:42 UTC (Mon) by mrshiny (subscriber, #4266) [Link]

I've read that document and I disagree that a stable API or ABI isn't a desirable feature. As a software developer I am aware of the challenges posed by a commitment to backwards compatibility. However that doesn't change the fact that backwards compatibility is still desirable for users and hardware vendors. Not everyone can or will want to put their driver in the mainline Linux kernel. But those that want to provide a Linux driver have little choice if they want to maintain the driver themselves. Either they go the NVIDIA route and basically work around the kernel, or the make a nice driver that plays well with the kernel and only works on kernels nobody has. With Linux 2.4 you can basically target the same kernel for all versions, with Linix 2.6 you need to target 16 different versions as of today. While the current situation stands, device support in Linux will be limited to devices that have been released for several months or devices with binary drivers. And home users will never be able to reliably buy new hardware and expect it to work.

What FreeBSD needs

Posted May 15, 2006 21:23 UTC (Mon) by vonbrand (subscriber, #4458) [Link]

The current model gave me suppport for a lot of hardware that isn't supported anywhere else. Something must be right with this development model... Besides, being able to fine tune the kernel to your exact machine for maximum performance does have its costs, like making a stable ABI next to impossible. Also, a binary blob that is abandoned by the vendor when the next widget model hits the market is of no (real) use. Yes, I do have hardware that needs ancient versions of Windows to install their "update driver", and the "firmware" is clearly not maintained anymore for a year or so, and might dissapear from the 'net tomorrow.

If you need a stable kernel API so badly, you are welcome to fork a stable-api tree. It is a free world.

What FreeBSD needs

Posted May 15, 2006 22:09 UTC (Mon) by mrshiny (subscriber, #4266) [Link]

I'm not saying that I PERSONALLY need a stable ABI. And I understand that binary drivers have problems. Heck, I'd be happy for a stable API, nevermind an ABI. Realistically right now I need a binary driver to use my workstation in Linux. I'm not thrilled about it but it works just fine for me. But it is very annoying that I can't run out and buy just any hardware without first conducting an extensive search to see if this hardware is supported by Linux, and if it is, which versions of what software I'll need to make it work. But at least I am able to do that, and for me that poor usability is worth the benefits of using linux. But the original question was something along the lines of: Does the Linux driver situation need to improve in order to please more desktop/end-users? The answer is clearly YES.

An end user expects to be able to buy a video card, bring it home, install it, and have it just work, or ask for the CD that comes with the video card, at which point the user can stick it in the drive and the computer does the rest. You can not tell me that this is the case with any Linux distribution, for any hardware that is newer than 6 months old, unless that hardware HAPPENS to work with some older driver.

Even if there IS a driver for the user's new hardware, it LIKELY is only in the latest kernels, which means a kernel upgrade, which is not something that should be done lightly. Lots of things need to happen to support a kernel upgrade. Lots of things need to be tested. Can you drop a 2.6 kernel into a 2.4 system? Heck, you can't even drop a 2.6 kernel into another 2.6 system all the time. And nevermind that most users use distro kernels, which are not always kept up to date with the upstream vanilla kernels.

I'm not saying the situation in the Windows world is perfect either. Old hardware lacks support from many vendors. But hell, Fedora recently stopped supporting serial mice! I mean, ok, serial mice are obsolete, but hardly impossible to find. Most users will be indifferent about this, no user will be happy that serial mice are no longer supported, but some will likely be upset that their old mail server/appliance/whatever no longer has a working mouse. But that's life, I guess. We've got the source code, we can always change it, right? But I can guarantee that no average end user will do that.

What FreeBSD needs

Posted May 15, 2006 2:51 UTC (Mon) by vonbrand (subscriber, #4458) [Link]

If you want any notion of a dependable system, you will not fuzz around installing stuff that has not been QAed with the rest of the distribution.

Sure, I do install vanilla kernels and the occasional package from source (mostly for testing or because the functionality isn't available elsewhere), but I know why I'm doing that.

What FreeBSD needs

Posted May 13, 2006 0:43 UTC (Sat) by malefic (subscriber, #37306) [Link]

'The distro-controlled central repository model' DOES work. It's the only way to make sure that packages you install work well with the distro and other packages. I think you meant the fact that major distro makers prefer to stick to some version of the package in that particular release of the distro. It's easy to explain. OSS world is in a constant flux - things are changing rapidly. It's a feature of OpenSource. So many distros do not risk the system stability in favour of some new version of some package.
Many, but not all. Take Gentoo - there's no such thing as 'whole system upgrade' there, and you can install any new version of a package as soon as it is available. Note - install from 'the distro-controlled central repository'. But if you screw-up your system - it's your fault.
So it's just a matter of choice. Either you choose stability, or you choose living on the bleeding edge and risk 'some inconvinience' :-). Another freedom of OSS :-)

What FreeBSD needs

Posted May 13, 2006 1:30 UTC (Sat) by davej (subscriber, #354) [Link]

If you install Fedora 5, and want to upgrade your sound card to a model that just recently came out, you're again forced to upgrade your entire OS to get kernel packages with the latest in-tree drivers

This is completely untrue. Fedora rebases its kernel to the latest upstream release constantly and releases them as yum-able updates.

What FreeBSD needs

Posted May 13, 2006 2:30 UTC (Sat) by elanthis (guest, #6227) [Link]

s/Fedora/stable distros/ ;-)

I didn't realize Fedora updated kernels that often. That's pretty interesting. Did/would Red Hat every do that for an OS that it offered paid support for?

What FreeBSD needs

Posted May 13, 2006 11:05 UTC (Sat) by joib (guest, #8541) [Link]

AFAIK they backport bugfixes and new drivers for RHEL. No rebasing.

What FreeBSD needs

Posted May 15, 2006 0:49 UTC (Mon) by elanthis (guest, #6227) [Link]

They don't do it on my RHEL servers. I've only just started migrating a few of them to RHEL4, though; maybe it's something they've only recently started doing?

It certainly isn't done on my Debian or Novell boxes.

Fedora kernel updates

Posted May 14, 2006 4:12 UTC (Sun) by markhb (guest, #1003) [Link]

This is still FC4:

[boot]$ pwd
/boot
[boot]$ ll vmlinuz*
-rw-r--r--  1 root root 1639539 Jun  2  2005 vmlinuz-2.6.11-1.1369_FC4
-rw-r--r--  1 root root 1634997 Jul 15  2005 vmlinuz-2.6.12-1.1398_FC4
-rw-r--r--  1 root root 1641264 Aug 26  2005 vmlinuz-2.6.12-1.1447_FC4
-rw-r--r--  1 root root 1644645 Sep 22  2005 vmlinuz-2.6.12-1.1456_FC4
-rw-r--r--  1 root root 1558522 Nov  9  2005 vmlinuz-2.6.14-1.1637_FC4
-rw-r--r--  1 root root 1559267 Nov 27 03:35 vmlinuz-2.6.14-1.1644_FC4
-rw-r--r--  1 root root 1550959 Dec 13 21:42 vmlinuz-2.6.14-1.1653_FC4
-rw-r--r--  1 root root 1551206 Jan  5 22:20 vmlinuz-2.6.14-1.1656_FC4
-rw-r--r--  1 root root 1603152 Feb  2 17:35 vmlinuz-2.6.15-1.1830_FC4
-rw-r--r--  1 root root 1603157 Feb  7 13:45 vmlinuz-2.6.15-1.1831_FC4
-rw-r--r--  1 root root 1603703 Mar  1 23:52 vmlinuz-2.6.15-1.1833_FC4
-rw-r--r--  1 root root 1598219 Mar 28 12:43 vmlinuz-2.6.16-1.2069_FC4
-rw-r--r--  1 root root 1598436 Apr 19 15:42 vmlinuz-2.6.16-1.2096_FC4
-rw-r--r--  1 root root 1602889 May  5 00:13 vmlinuz-2.6.16-1.2108_FC4

Looks like I need to do some spring cleaning! Believe me, it makes keeping the CiscoVPN module in sync a real bear. Yes, I know about vpnc.

lack of motivation I guess

Posted May 13, 2006 12:37 UTC (Sat) by gvy (guest, #11981) [Link]

> 4) Administration tools for CLI-phobes.
Developed by whom? I doubt these fine people vowing to "run up with Linux" are enough to do that.

--
mike@,
who recently ack'ed a "80%" patch to the "20%"
sysv service control module for ALT Linux alterator framework
-- that was rather a piece of cake to write, but polishing
demanded much more time (so thanks dmi@)

oh no, latest/stable mixer again

Posted May 13, 2006 12:50 UTC (Sat) by gvy (guest, #11981) [Link]

> 2) Ability to install the latest version of some application
> without upgrading to a whole new version of the OS.
Hey, that versiomania isn't to be cured technically. It's human's disease, and if one wants to get on a bleedin' edge, s/he wants to put some effort into it.

For one, I do backports as I need them and API issues are at some reasonable level. If some clueless user wants "the latest AbiWord", maybe he either needs "the distro with it", or "skill to backport it", or "someone who can do it". Most likely, he just can't tell *why* exactly the *latest* version is critical; in fact, it usually isn't.

PS: s/wants/needs/g is another story, of course.

latest apps on any distro... klik!

Posted May 13, 2006 19:06 UTC (Sat) by jabby (guest, #2648) [Link]

Isn't this was klik was made for? Basically, a klik package is a mountable filesystem containing everything necessary, including the libraries. I just looked and Abiword is not available as of yet. But, as Linux desktop use increases, more and more packages will be provided in this form.

AFAIK, there's no reason that klik couldn't work under *BSD as well. The apps would just have to be compiled for *BSD.