Slashdot Mirror
LSB: A position paper
Ransom Love from Caldera has published
a positional paper about Linux Standard Base and what he
thinks about if the Linux community won't adopt it. I think it's a "must read" (thanks to Linux Weekly News).
Update: 03/01 03:23 by S : Seems like that LSB needs a boost: now
Intel is talking of setting up a new initiative... (see
half way down the page)
4 years ago, I started out on Linux. Then 2 years ago, I switched to FreeBSD. Why?
Well, if you look at the FreeBSD development model, it's very cohesive and centralized. I like the fact that by issueing just a few commands, I can update my entire system. I don't have to worry about compatibility or patches. It's already been done.
While some would argue that this leads to closing off the users from the developers, I disagree. The developers of FreeBSD are very open to requests and suggestions from users. They only seem to insist that you make requests clear and detailed.
I'm posting here because I feel that Linux could benifit from a development system modeled after FreeBSD. Give your users the ability to use a utility such as cvsup to update a source tree, then a simple 'make world' to build the new source and install it.
This weekend, I picked up a copy of RH 5.2, and will be installing it on my workstation to see where linux has gone over the last 2 years. I'm not holding my breath though. From everything I read on the web, it seems the community is still badly factionalized.
Were I developer and had the time, I'd love to start up YAD (Yet Another Distribution), and bring some order to the chaos. I am neither a developer, nor do I have the time, so I plead with anyone out there to do this. Only good will come with it, I promise
Think about it... you only have one problem.
If you use system calls directly you are in deap shit as it is, they are not standard in any way do't use 'em use ISO/ANSI/POSIX standard fucntions that will be in every standard c library alive(or atleast from GNU), or staticly link your application. If you program is worth billions of dollars of development time than beliave me, staticly linking the libc and other library won't hurt you at all cuz the program is going to be huge as hell as it is.
So when everything is said and done if you staticly link you program with needed library you don't have jack to worry abot(in majority of cases). And use system calls directly is bogus and stupid if you cannot update you program as they change. System calls should not be used directly anyhow unless you are writing a module.
Plus take a look at oracle... if fallows GNU LGPL to the letter and because of that it can run on any elf ditribution of linux with a not so recent libc =)
...because I trust Pat Volkerding's taste. That's my plan, anyway.
I'm not a hacker, just a de facto sys admin trying to use Linux on production servers. Slackware is tidy and I understand it, more or less. I can use pkgtool and I can use tar.gz's from almost anywhere.
Slackware = dry martini
Red Hat = fruity drink with umbrellas served in a pineapple
Debian = Kentucky bourbon
Others = ?
Forget LSB. Red Hat is the Standard Linux Base in my (and the public) view.
The other distos are nice oh yes, but I'm afraid that distos like Slackware and Debian
won't be around in a couple of years time if they're not compatible with Red Hat
(well they might be around but as hobby OSes or niche OSes). They are both
quality distobutions, belive me I have tried both, but maybe it is time to swallow
some of their pride, and just go along with the rest of us. It has nothing to do with
what distrobution are the best in an objective way. It has to do with marketshare and
where the money goes: Yes, its Red Hat. If we are going to take on microshit, we need
one big player instead of 50 small ones.
And don't come and say Red Hat is unstable blah, blah, blah. It is much more stable than
NT and that is the important thing.
And this goes also to Suse(big on my continent) and TurboLinux:
Be compatible with Red Hat or become obscure.
Just my 0.2 Euro
Quality can beat quantity when it comes
to programmers and their time.
I don't think that the most important goal
is make Linux an OS for newbies or other
point-and-click people since it bears with
it a risk of Mac-dumbing the user interface.
In my opinion this paper is spot on. We see it now, almost all major software vendors support one version of linux and thats RedHat. I am in the middle of porting a large application (> 1Million lines of code) and it is a pain in the neck looking in different places in different distributions for files and commands, sometimes with different syntax (startup and shutdown is a classic example). As such we will probaly choose to support one distribution only (RedHat). It's great to have a choice but you also need a strong base to work on. If you want choice you must support LSB otherwise RedHat will be a the only major commercial distribution. Others will only survive as niche players. Without LSB NT will always have one trump card (write once - test once)
Kevin.
As Nicholas Petreley recently observed, the tendency in Linux distributions is to move together anyway. The reason proprietary unices tended to fragment was because it was too expensive to duplicate the functionality of other unices' software. But with Open-Source(tm) software, you can just copy a competing distro's work.
RedHat is pulling a very clever judo manuver on the other Linux distributors. By giving away its software, RedHat is burning any distributor who tries to come up with proprietary alternatives to RedHat standards.
The freewheeling attitude of Linux distributors and developers is exactly what made Linux so robust and attractive in the first place. If you want uniformity today, go to BSD. Don't be surprised if nobody follows you, though.
Contrary to what some people believe, RH is NOT the most commonly used distribution. Most people I speak to hate the monster. RedHat just seems to get the most press....who the hell knows why.
Not the point. But consider this:
Be compatible with Red Hat or become obscure.
Get bent. Don't pollute Linux with an MS attitude. FREEDOM! FREEDOM! FREEEEEEE-DOM!
Well, whatever you people decide, fine...I am going my way :P If LSB sucks as much as I think it will....I won't have anything to do with it. I will sit happy in my SLack box and watch all you flame each other over which package manager should be in LSB,...if it should be KDE or GNOME...
Then, when it dies down again,...and LSB still isn't shit...I will just laugh.
I hope it does die too,...we simply don't need the thing.
Flirting with popularity is becoming a dangerous affair. For if
going mainstream freedom must go downstream, I would find less reasons
to stay with Linux. The kernel, the C library, the GNU utilities are
GPL'ed, thousands of other applications are also GPL. GPL (and BSD)
is the default license of our community.
Do not want the new world-order in Linux.
Now that would be a good move! Place Linux under the control of a commercial venture so that Microsoft can compete using their established techniques.
The real problem for Microsoft right now is that their method of competing doesn't work too well for something that is free. Once everyone thinks you have to buy from RedHat to own Linux, Microsoft is half way to defeating Linux.
Having many distributions is what protects Linux from this predatory approach. Imagine when RedHat gets really big and there is an IPO. How long will it be before Microsoft owns RedHat?
Hopefully, ISV's would be careful to require as low a level as possible, but if they really need libVeryFancyFunctions and libEvenFancierFunctions and libLotsOfEvenFancierLibs, they could specify a LSB level that included their weird requirements.
First of all, I hope you understood that he meant "test on one platform instead of X number of different distributions"? And as for startup and shutdown scripts being trivial. Sure, for a user that are familiar with the init system of his distributions, but face it: Most novice users don't even want to care about details like that. With LSB they likely wouldn't have to, because you would succeed in figuring out where to place the scripts on any compliant distributions.
> linux is already doing just fine, even without
> a prime-time Desktop Environment
What's that funny thing everybody's talking about:
"Desktop Environment".
For, say, my mom the IDEAL desktop environment
would be a fullscreen panel with big buttons on it, like:
TYPE
COMPANY'S DATABASE
EMAIL
OFF
No funny silly icons or whatever... Really!
And for me the ideal Desktop Environment is like
an xterm (rxvt in my case) and handy two-panel-like
file manager.
It's so hard to understand people trying to compete with
M$ in running-into-walls!
Newest Version of Debian... :) :)
apt-get dist-upgrade
done deal
Good stuff.
I used to do alot of slack, only slack... then about a year ago i had to start maintaining ALOT of servers running linux and I needed something a bit easier to upgrade than slackware, after checking everyone else out (rh,suse,deb,turbo) i choose debian because i feel it is the easiest to deal with on a day to day basis and by far one of the easiest to update... (2 command lines and your on your way) and the newest bugfixes and such for older versions of debian are easily downloadable the same way. good stuff.
Well, good standards are anyway. The LSB seems to me to be a good standard. I hope the different distros support it because I want more games so I can dump Windows.
Why can't ISVs distribute proprietary software as shrouded source, as used to be done all the time in the UNIX world?
Server bandwidth costs money. The only reason
that there isn't a Linux server like ftp.cdrom.com
is money. I sure wouldn't be able to fund it.
Red Hat maxes out their bandwidth all the time.
I've seen sunsite maxed out, and they support
some insane number of simultaneous users.
FreeBSD has been getting a free ride.
Popularity could put an end to that.
I believe what Linux should ultimately strive for is more that just a "Standard Base" for distributions. Why not create a standard base of functionality interfaces. Let me explain. Currently if you want to run a program, you normally install any required other programs/modules and then install your program and it works. ISV's should be able to do this as well. What I would like to see is an interface to all of the common components of a computer. The /proc filesystem goes a long way towards providing a common place to get information and such. But what about Video? Framebuffer kernels are emerging to handle this. But all of these are implementation specific and don't address functional interfaces. Linux virtually has no common printer support (Unless you consider postscript to be the only printer type) and even with postscript there are inumerable programming implementations and ways to generate it. What linux could really use is a well designed interface api to ALL of a computers hardware, at the very least, and operating system functions at the most. Included should be sound (Input and Output), video (Input and output), Keyboards, Mice, Digitizers, printers, plotters, etc. (If it's hardware there should exist a common API that fits) I don't mean to imply an API for every piece of hardware either, but where applicable there should be one API that handles various pieces of hardware. Then even with initial poor implementations of these API's, over time they could improve while the programs that need those services wouldn't need massive rewrites. In fact if the API existed then vendors could safely make an implementation of an API that didn't exist, knowing in the future that their program would work with independently developed API's.
Instead of a common software base, how about a common programming interface that encompasses all of the computer. Everything from languages to applications would benefit greatly from this. ISV's can concentrate on making their program run instead of worrying about what machines/distributions/software versions it will run on.
apt-get update
apt-get dist-upgrade
Starting with debian slink (released tomorrow), this will update your entire system to the latest and greatest. If you want ease-of-administration, Debian and apt are the only intelligent way to go.
And Debian is actually quite like the freebsd model, in that developers are quite open to suggestions, ideas, etc.
Should we all go and develop our own binary formats and our own libc's for the sake of freedom? No, because that would ensure that apps only work on one distribution. So why not specify a standard, a "virtual distribution" if you will, that vendors can choose to be compatible with, or not, at their own discretion, to make life easier?
It lie in the nature of this concept that it's not meant to be an all encompassing standard. If that were the case, we could just take one of the major distributions, and tout it as the "standard" Linux.
It is to avoid this, that the Linux Standard Base is being worked on: So that we can have a small specification that give vendors a base that they know will be there in any conforming distribution.
It is a way to avoid all new apps from being labeled with "Supported only on RedHat" or "Supported only on Debian" at the whim of a vendor, leaving the user in the dark on whether the application will work on any other distributions, and instead inform the user that of a small set of requirements that needs to be filled for the application to run, without having to list lots of library names and version numbers, and force you to check them manually, as opposed to just checking the webpage of the distribution in question and seeing that it conforms to LSB, level , and instantly know the app will work.
It seems like some of the people whining here are so afraid of standards I find it weird they're even using a webbrowser, considering the staggering number of standards any "modern" webbrowser attempts to adhere to...
Standards does not mean you're forced into conformity. Standards are labels you can choose or not choose to apply to your product, to let your users know what feature sets you support, without handing them thousands of pages of specifications.
Standards can be bad, and so you can choose to ignore them, or they can be inappropriate for your particular product, and again you can ignore them, or they can help simplify your and your users life, and then they've served their purpose.
I doubt anyone involved with the Linux Standard Base wants to create something messy that everyone dislikes, and that are big and bloated etc. But if they do, you can ignore it, and nothing has changed, except maybe a few distribution claims compliance. No matter if the LSB becomes a mess or not, it will make it easier to choose apps and distributions.
Even if only two distributions end up supporting it, that means one feature set less to worry about when developing an app, or checking if an app is compatible with your distribution.
And thats a good thing.
No, you don't understand. Some of us here, myself definitely included, have no use at all for proprietary software. I *have* to run windows +excel+matlab+program win32 on a pretty regular basis. The last thing I want to see is this kind of mentality invade the linux world. It's bad enough I am facing an expensive windows upgrade next year.
What people like Ransom Love don't realize is that the whole idea of linux standards will be ultimately worked out by those people writing GPL code. If he wants his company to have any participation in this, he better ensure that Caldera code gets out. Posting in freshmeat would help. (Don't even start... Caldera is buidling their business on a code base valued at 2,000,000,000 dollars, which they are getting gratis. They are morally _obligated_ to return a reasonably large chunk of the favor.)
This article raises a lot of questions. Here are a few:
1) There is more heterogeneity in time than between Linux distributions. Time must surely be a bigger headache for ISVs, but I can't see any mention of this in the article. How often will LSB be revised? What will the compromise be between the need to make technically correct choices (e.g. the move to GNU Libc 2) and the "standard" environment demanded by commercial companies?
It seems like a divide-and-conquer strategy is being used by ISVs here. I can't imagine Microsoft being willing to slow up development of its OS for the sake of ISVs. Why should Linux distributors?
2) What does LSB encompass? And how limited will it's ambitions be? It might start with a standard kernel and set of libraries, but you can imagine it will grow to encompass package manager, X toolkit, desktop environment, ORB... Committees always find more work for themselves to do. Heterogeneity exists in these areas for a reason. Early standardization would be a disaster, but that's where the commercial pressure will be.
Some things are better left to the market: in this case, the free market economy and the free market of ideas.
Then you're missing a lot of the exciting new software development...
You need a modern distribution! GNOME/Enlightenment is awesome. It is no surprise that it is very popular. All that software has been written in the last 14 months while Slackware has stagnated and can't keep up. I've not much trouble trying out each gnome rpm version on my up-to-date RedHat system. I've heard that Debian tracks GNOME well too but I haven't tried it. Maybe if there were an LSB you might be able to run cutting edge software too.
Martin Sevior
What other company is he talking about? Mozilla? He never explains what the liability is. He assumes (incorrectly) it's obvious.
In two to three years, Red Hat will likely not exist as a seperate company and certainly not have the current leadership. They are getting a lot of investment from big companies and that doesn't come for free. Red Hat are currently very good at obtaining investment and provide a good service back to the Linux community. However, they make very bad business decisions and their investors won't let them do that for long.
I suspect that if the Linux market had been bigger, then Red Hat would already have destroyed it.
Best Regards
Mark
Hey kid, do you have ANY idea what you're spouting off about?
Didn't think so.
Linux existed JUST FINE for years without any big
grandiose standard to make industry happy. I dont
need any of that software to do what I want to do
with Linux. The only people who need something like
LSB for growth are the people with money in mind:
Red Hat, Caldera (why the hell do you think he wrote
this?) and pals. We can continue forever, developing
our own linux open source alternatives. We dont NEED LSB.
The money people WANT it. Don't confuse NEEDS and WANTS (remember that from 6th grade social studies)
As long as I can configure && make && make install, I dont need no steeking LSB.
Which rather ruins your troll. The only significant
player not backing the LSB is Slackware, and Pat
has basically said he's unconvinced but will take
another look when its finished and everyone is
using it.
Most of the distributions are very close already
to the LSB work so far. Most of the differences
being resolved are silly little rc3.d v rc.d/rc3.d
type stuff
Ok let's go by the points.
Expect the ISV to Publish the Source Code for Their Applications
Ok Ransom Love may have given some good points here. However what he points is only a part of the iceberg. He does not denote the unwillingness of developers to keep pace with kernel modifications. He seems to forget that people also tend to do things over minimal effort at cost of everything (and everyone) else. That's what Microsoft has been doing in its "feature" policy.
Besides in one point of view it is strange that here he does not refer Caldera as one of these non-OSS players (Caldera's Novell NDS implementation possesses non-OSS binaries which are imcompatible with recent kernels).
This situation is fundamentally important. For example I need to use kernels newer then 2.1.125 on my work. However Caldera's NDS works normally only with kernel 2.0.34-35. So I'm forced to choose between the choice of running two dozen programs or running a single utility. In fact situation is not fundamentally critical. Linux systems work relatively correct with several kernels on one single machine (yes you have to reboot it, but no one doesn't tell you you can't work).
Of course we must give the right to anyone to deal with its own proprietary ideas as they wish. But they have to agree with the price. If Linux industry runs too fast for them, they must fit on it or consider alternatives. Linux is multiple since its early times. In fact let us remember that Linux, by itself is just a small piece of software with huge tendency to be mutant. That's how this thing won the race of portability and became the most flexible OS ever created. Killing this feature means destroying Linux development basis. And besides why do we need an open source kernel for? Just to look how beautiful c files are? And here we go to the second point:
Encourage the ISV to Select a Single Kernel Version and Library
It is quite clear that under such conception, it would be easier for ISVs to port things to Linux. But is not this what we have with Microsoft? Well, probably most ISVs have not ever seen a single line of Windows internal code. But most of them don't need it also. Under such conception you only need to have clear and straightforward specs about the kernel functions, nothing else. Source code becomes here secondary.
Tying people into a single code is silly. We all know that 2.0.XX possessed a lot of inconsistencies, bugs and, most importantly, violated some conceptions on processors, specially SMP systems. 2.2.X seems to have solved a lot of these issues. Anyway 2.2.X is not perfect. We may see it on 2.2.1 story ("brown paper bag" kernel). So what we will do with such things? Hang on while LSB comitees decide to issue a "feature". Will hundreds of thousands of developers have to wait for a "decision" so that their new software instruments and tools will work with new kernels?
Besides where is the issue is being rised? In fact we all have to note that every distribution presents a very conservative view. Apart of libc5/glibc problem, the kernels presented on most distributions, differ in a subversion number. Most problems arise as _we_ - users, admins and developers, choose to upgrade kernels to the most recent ones. So who will mostly suffer with this "standartization"? It is clear that it will be the end user, not the distributor and much less the ISV.
Require the ISV to Select a Single Linux Provider to Support & The Only Answer is Collaboration
Is anyone to make here a new Tobacco Cartel? Yes there is a need to create standards over several conceptions and development trends. But we should not run over making Linux a new almost sterile ANSI standard. In fact Linux needs a RFC-like policy over several critical paths of development. And as any Internet RFC they should be given only under a recommendation basis and not as a policy of dividing Linux developers/users among insiders and outlaws.
We have to recognize one point. UNIX has suffered several times of its loosy characteristics. However UNIX is one of the oldest and long-livers among OSes. Linux is by itself an example of it. For those who came from other OSes this may seem quite foggy. But not for those old-DOS and UNIX hackers, who are familiar with this Universe. So, crying "wolf" here is quite suspect. It looks like some people are trying to push Linux out of its world and molding it into a more "domestic" OS. As in the previous point we have once again the same question. Who will gain with this? The Linux community? Or a few blue collars, ever searchers for the software Eldorado, that barely have an idea where they are digging in?
.
It's designed by a slow moving committee. If ISVs keep releasing Motif apps for Linux, they will continue to not interact with the KDE and GNOME desktops.
As for GTK-- I can live with it, but as for themes in widgets. It's fine if it doesn't come at a price in performance-- nobody wants sluggish apps, no matter how great they look.
My experience with GTK is that GTK themes kill performance.
I used to be a Slackware user, until I got tired of downloading and compiling every little package that I wanted.
I started using RH and Caldera, which both come with most of the apps that I had to compile by hand under Slackware (I have nothing against building packages, I just don't want to do it all day)
I don't like Red Hat much. I always have to reconfigure the system too much afterward. But they have libc6...
He said that the life of Linux will fall short. WTF? It's been around for 9 years, there's no way I am abbandoning it just because there is no "LSB". That's corny. Applications that are available work under different environments, it just requires you to compile some libs first to make them work. If you don't know how to do that then you shouldn't run Linux. Learn to walk before you run.
I think the real problem with standards compliance is RedHat. Don't get me wrong, I'm an avid RedHa user, but some of the decisions that they make regarding packages are just inane.
/usr/lib/netscape? why? please, for the love of god, why? can't it just go in /usr/local/netscape, like a good app should?
1. Why is netscape in
2. Those goddamn windowmanagers. Between Xclients here, and xinitrcs there, how the hell do you just rip out the ones that you don't want and put in new ones?
I mean, really, if I could just get the latest windowmaker source, compile, and have it do a make install that overwrote the rpms, that would be great! but instead, i have to back out the old rpms, do an install, and then pray. and when new rpms come out, how do i back out the compiled installation?! so messy.
i think that the number one problem is just the major differences between where source compilations put their files by default, where the rpm binaries put their files, and then where the distributions put their files. if everyone could just agree that apache goes here by default, gnome goes there by default, my god, it would be LSB, without killing everything and pissing everyone off.
Fine, live in the past and get out of way.
motif really sucks
Qt > *
My brother-in-law owns an auto parts store, and uses an app that runs on either QNX or PICK, I forget. Hardware, printers, network hubs, all available from only one source. You can imagine how expensive that is. Fine stability, but as expensive as anything.
Caldera should go to NAPA or CarQuest or AutoValue and tell them how they can save on seat licenses. When they have something going, start advertising in the trade mags.
You'll need some sort of text interface. How about helping to bring up to snuff the port of the Turbo Vision interface?
There are so many oppos out there. I just don't see how anybody can go wrong.
A lot of people whine about Linux (and other OSS) going "mainstream" ... "mainstream is downstream", "don't mess with my .tar.gz's", "give me command-line option wizardry or give me death", etc. Well, whatever happens, they have today's sources so they can continue to go their merry way.
However, there are a couple of reasons I would like OSS to go mainstream and eventually displace Windows, OTHER THAN the joy of world domination. The first is strategic self-interest: the longer Microsoft dominates, the more likely it is that they will gain control of Internet standards and make it impossible for OSS users to access the Internet on an equal footing. The second is altruistic compassion: the majority of the world's population cannot afford Microsoft software or the hardware that it demands, and Microsoft intends to keep things this way (note how they tried to suppress low-cost PCs). These people need lean, zero-cost, absurdly easy to use software to run on the incredibly cheap hardware that Moore's Law will be providing ten years from now.
Robert O'Callahan (roc@cs.cmu.edu)
Linux is winning. Linux is better. Companies are taking note and becoming interested. They are going to develope commercial software for it. They are going to use redhat and redhat will come to be the definition of linux. If you want to run commercial apps you'll have to run redhat or the other dists will have to copy them. And because of market forces, you'll end up using redhat or a redhat knockoff regardless of whether or not you give a damn about a package manager, standard libraries, all of that stuff. We don't need LSB because the market will define the standard, but it would be nice to have LSB and keep debian and the other distros. relevent.
What I don't understand about statements like this is the lack of room for comprimise. If the LSB was made policy tomorrow how would it effect you anyway? Would it keep you from running any particular version of any particular library? If gtk is the standard, can you not load motif and program for that? if you want egcs 1.1.2pre2 can't
you just download the thing and bootstrap it?
Do all disto's have to comply 100%? No, only if they want joe clickity as a customer. What's the big deal?
Bah. What's wrong with installing into /usr/local/* ? Who cares what format the binary comes in? tar.gz, tar.bz2, .rpm, .deb. If your system can't read those formats, that's your problem. If it uses non-standard libraries, it should come with them. How do you know it's non-standard? The majority of distros don't have it. You don't need an entity like LSB to say what's standard and what isn't. Just look at the distro's and see. We all know that at the rate Linux advances, the LSB will get out of sync with the distros, and then we'll have lots of binaries that are distro-specific anyway.
: I think it looks better, has features that Motif lacks (like tearable menus),
Uhh, excuse me, Motif 1.2 has tearable menus.
I admit that they don't work that good, and the last time I tried them with Lesstif it crashed my computer (that was 0.83 or older), but they are there.
: don't have much experience with GTK, and none with Motif, which I thought was too yucky when I looked at it.
Once you understand the basic concept of how Motif and/or Xt work then it all makes sense. What I don't understand was why GTK wasn't written on top of Xt? It seems to me that the GTK folks chopped off a toolkit layer that was proven and worked inorder to do their own thing. Why? I don't know.
One thing GTK does do is have a simpler or is that cleaner user interface. Which is less scary for the first time GUI programmer than Motif is. Is this a good thing? Have you ever seen what happens when management gets hold of a GUI builder like XDesigner? IMO, complicated is good and it keeps the undesirables (management) from writing GUI's.
This Caldera article is too late, this matter has already been decided and they know it.
The answer is Red Hat.
Many Linux users will whine and complain, but just look at the news. Who has been making all the big deals? Not Debian, not Caldera, not Slackware, but Red Hat. When the big computer manufactures start shipping Linux for consumers which Linux distro will be on it? Red Hat almost exclusively.
So from a commercial developers point of view (which I am), the only distro worth being concerned about is Red Hat. If my program works on the other distributions then fine, if not, then too bad. My time is better spent fixing bugs or porting it to some other platform.
Game over man, LSB why bother? Red Hat has won.
When the commercial apps arrive, the distro's that are not "Red Hat compatible" will perish. Its simple numbers.
So, let me get this straight. Linux, a kernel that started out as a project to learn the quirks of the 80386, and then became a "better Minix than Minix", must now bend to the whims of ISVs?
No! Well, yes. But only if it wants to break out of the server/hobbyist ghetto, and who wants to SELL OUT like that?
If I had five billion dollars in Microsoft stock, I'd liquidate a million measly bucks and buy the souls of a few stalwarts who'd call press conferences and demand the right of the individual to compile everything before using it. Cheap mother-humping insurance.
Excuse me, galactic game master, but you'd still be playing Old School Star Trek in text mode, with a digital compass rose taped to your monitor bezel, if it weren't for those evil suits and their mass-marketing ways. Mille Bourne much? Hunt the Wumpus, maybe? I have the cheats if you want them......
I STILL call them player/missile graphics, and if you want power, you'll learn assembly and like it.
Hey, all rightt! You've combatted factionalism by coming up with your own faction Which Makes All Other Factions Obsolete. I guess all the others will go out of business by next week and you'll be the only one left, eh? Where DO you come up with your ideas?
>>I guess my question would be, why should apps be interested in most of that? Video, for example. For console graphics you've got SVGAlib, you write to it's API and let it worry about the hardware. I wouldn't recommend it, though. Better is X11
That's kind of my point. You just cited two "standards" for doing the same thing. And those aren't the only ones available and they don't cover video input. If a common API interface were in place and you were using it, you could invoke a program to run in either SVGALIB or X11 or whatever environment. Maybe a better name would have been a Hardware Abstraction definition (not nessasarily layer) that would cover all hardware types.
If the API is abstracted and supports all hardware (or catagories of hardware), the only issue would be whether required libraries for an application were available. It wouldn't matter if the user were writing SVGALIB or X11 and it generally wouldn't matter what version of the API was installed, although later versions and more widely distributed versions would usually contain fewer bugs. Also the same calls would work on all platforms, making porting to other processors less painful.
You mean the power/stability of NeXT with the GUI of MacOS. Um, count me out, as neither are appealing.
Agreed. If your version number isn't really high, it means you suck. Linux is a prime example of a slow mover. We have windows 98, NT 4, solaris 7, HP/UX 11, and Linux 2.2? Guys, put some money into devel of Linux, for crying out loud!
: 1) Do we want the right to run propritary software if we want to.
If you want this, run windows.
: 2) Do we want the right to have the source to all the programs that we run
If you want this, run Debian. (I write debian here because they're very clear about what packages are free and nonfree and the main distribution includes only the free)
I'm a professional X11 programmer, and GTK+ is one of the nicest widget sets about. Combined with GNOME it has the potential to beat even the object frameworks produced by Less Palatable Companies.
For people who have never done professional X11 programming, Motif is CRAP. Everybody hates it. It was designed by a committee, and damn it shows. There's a reason it's called Bloatif.
Even the addon packages to make Motif more usable (by giving it workable file dialogs, tree views, and a drag and drop you don't have to implement 90% by hand) are buggy, slow and memory hungry.
GTK+ wasn't designed, it evolved, and in typical Bazaar fashion that seems to work better. Themes are a non-issue: in terms of ease of programming and speed of binaries, GTK+ is the obvious choice.
I am a consultant and many of my clients would be moving to Linux for all the right reasons, except the for lack of UI standards.
The problem is, Linux has standards. Informal standards, most borrowed from Unix. We use tar and gzip to zip files. The directory structure is standard. And command-line arguments have some degree of consistency (--help for help etc.). The problem with these standards is that most of them are 20 years old (or however old Unix is). The LSB wants to make these standards more modern. If something like the LSB doesn't work and OpenSource developers become stubborn, I see these things happen:
1) A certain distribution becomes standard. Programs will be developed exclusively for that distribution and patches will be needed for them to work. A Very Bad Thing.
2) We will continue to use the Unix standards. As Linux becomes more modern (GUI's, games, whatnot) it becomes more inconsistant and random. In other words, average users will find Windows as a better OS because of consistency. I will agree with them.
The LSB is the single most important step right now to World Domination (next to DE). This could help with setting a standard for programs to work on both GNOME and KDE. LSB is A VERY Good Thing.
/.ers of all people should know that technology goes very fast. If we keep with legacy standards or have a distribution as the standard, Linux will grasp where it can not reach.
Why is netscape in /usr/lib/netscape? why? please, for the love of god, why? can't it just go in /usr/local/netscape, like a good app should?
/usr/bin/.
/usr/local" Do you REALIZE that if they didn't there would be 0 files in /usr/bin, /usr/lib etc., Up until now, GNU assumed it was an Addon to the OS (/usr/local). Under GNU/Linux it is THE official tools, which means it can rightfully be installed in /usr instead of /usr/local. Hence the reason it switches to /usr.
/usr/local on a Solaris System.
I know the reason, and it obeys all of the rules of the FHS (It is not an end-user app). The end-user app is a wrapper script which is in (rightfully)
Also, I am sick of people saying "Why do they switch from the default
Of course, most of the complainers have probably not read the FHS, or do not realize the reason that the software gets installed in
-- Keith
As pointed out, Red Hat, more than most, contribute back into the Linux Community and for that, they deserve nothing but respect but picture this.
If Linux had 50% Desktop market share, had all major and most minor applications running happily, had 20 or 30 mainstream magazines devoted to it and it was becoming a desktop standard and big companies where making mega bucks with hardware and software sales on the back of Linux and Mr Gates was not a very happy person and Red Hat software had by far the most respected and profitable share of the Linux market (probably deservedly so). Picture this had happened two years ago. Picture Red Hat, suddenly dumping egcs, glibc and fvwm95 on it's users without a word of an apology. Picture it takes two releases to get all those applications working again. Don't get me wrong, egcs, glibc are obviously the way to go but you don't do it like that. Imagine what IBM, Compaq, Intel and the like would be doing to Mr Young with all their customers screaming blue murder at them. Not a pretty sight. They have shown similar bad management in simple ways like moving offices and losing their WEB site for three days. Red Hat may grow up in time but more likely will become a name owned by someone else. Debian, Slackware and SuSe will last longer, I'll bet my hat on it!
Regards
Mark
Aren't you a condesending bastard...
I build gnome just find on a slack box. I erased it all a week later cause there was nothing in it that I couldn't all ready to with much smaller, cleaner programs I had which didn't use gnome.
a.
I think he meant the venture capital they invited.
In a lot of companies that started something new
what happens is the following:
1) They start real small with their own money.
2) They make good products and get good press.
3) Venture capitalists come and fuel money into
the company.
4) The company makes an IPO and does well for a
while. The company founders are now CEO's etc.
5) The shareholders want better performance and figure it's better to bring in "real business
professionals" to run the show.
6) The founders get fired, are rich but don't
have anything to do anymore with the company
they founded.
I think a good example for this is
Cisco, and, (of course with modifications)
Apple.
Matthias (who forgot his password)
PLEASE LEARN TO SPELL!!
/., the word is "hobbyist", NOT "hobbiest"!
Contrary to custom here at
Arrrrgh!
What's wrong with GTK? I think it looks better, has features that Motif lacks (like tearable menus), not to mention that there is a fully functional free version. LessTif mostly works, but it's still not quite finished. GTK also seems to be easier to program, but I don't have much experience with GTK, and none with Motif, which I thought was too yucky when I looked at it.
Try Debian. It doesn't have support for compiling from source automatically, but you can do automatic updates with apt. Very nice.
You should try the Notif theme. Looks very much like Motif. And what you mean by loud?
./configure --prefix=/usr /usr/bin etc and overwrites the rpm-installed stuff. Or you can make an rpm yourself, I'm sure it's not that hard. /usr/local is that /usr/local shouldn't be touched by the vendor. That's in the FHS.
Now stuff goes in
The reason it doesn't go into
If the major Linux distributions agree upon support of LSB, other distributions servicing a different sector will not destroy Linux.
.
The users of a given distribution will become self-selecting. Those that have little interest in commercial applications, but value their option to alter the underlying OS in any manner they choose would follow the dissenters.
Both have a role to play - LSB and the major distributions will make Linux mainstream perhaps even to the desktop . Those that will play with the code in a non-standardized environment may be creating the next LSB standard
If this is about freedom, it is also about freedom of choice. Linux should not become 90% of the OS market.
examining the userbase and finding out what is used by the majority of those users. You know... the (L)owest (C)ommon (D)enominator?
:)
LSB v1.0 minimum specs could be 486DX2-66, 16MB RAM, 500MB HD, VGA 640x480, SB Compatible Sound, OSSFree or ALSA, libc5, X11R6 3.3.1, SVGALib, Linux 2.0.35, Mesa 3.0, TWM, linuxconf, RPM and File System Standard.
Almost every distro *I* know of provides these basic things. I include libc5 because it has the most widespread implementation (I can run libc5 binaries on my glibc system), X11R6 is a standard, SVGALib - standard, Linux 2.0.35 or better standard, Mesa 3.0 or better - standard, TWM at the very least, linuxconf is essentially a standard and distro agnostic, RPM can be used on any distro - I hear Debian's install stuff is good but more people use RPM. FSS is VERY important for compatability and crossplatform capability - isn't that what Linux is all about?
That's not to say that the distribution vendor couldn't ADD to the spec with newer versions and more choices; it would just mean that the ISV's would have a minimum system configuration that they could write to.
JUST DEFINE THE MINIMUM REQUIREMENTS. It's as easy as that. Give the ISV's additional incentive to provide us with additional sofware.
Thank you for your time.
Codifex Maximus ~ In search of... a shorter sig.
Posted by OGL:
Eveyone should be in favor of this, especially those who are always whining about Red Hat taking over the world. Instead of making products for Red Hat, companies can make products for any LSB-compliant distribution.
-W.W.
Posted by jhreingold:
...".
As someone engaged in porting a commercial application from other Unices to Linux, I would cheer at the arrival of a base set of services that I can count on in all Linux distributions. That base set of services has to be competitive with other OSes, or Linux is not going to make it. My users are used to being able to install & uninstall with some sort of package manager. I don't want to tell them "on Solaris, use admintool, but on Linux, you have to untar, copy these files, check these prerequisites,
True but vacuous. Is there any general-purpose distro without plans to support 2.2.x, egcs, and glibc in the very near future?
The devil's in the details -- for example, the libjpeg problem with its unnecessarily fine-grained version code in its header. But these are easily solvable.
As for LSB, as long as it's a reasonable discussion between techies to solve technical problems, bless them. But -- at one point, at least -- it was in danger of becoming yet another vehicle for the overblown Perens ego, to which the various distributions replied (quite properly) by ignoring it and starting side discussions.
I have no idea why Love chose to write this now, unless there's some special situation at Caldera that requires a political statement of LSB support from them. I would have thought that the whole standards thing was mostly a non-issue by now....
Craig
Even if you don't see need for LSB in order to run commercial software, I don't understand how anybody can argue against standards.
Standards are good for competition. The more things there are that are standard the easier it is for me to change distributions or work with people who use other distributions. There's less to learn, less to recompile, and less chance of frustration.
For example, every distribution has a way to stop and start daemons. Since the feature is standard there should be a standard way of accessing it. If you can't agree on init.d setup, then agree on a standard interface that each distro will implement in it's own way (ie. a set of /usr/sbin scripts that call the native commands).
Now I can go to any machine running Linux and be confident that, even though I've never seen the distro, I can perform basic administration. Also, if Slackware comes up with some awsome whiz-bang feature that I must have, I can switch to it and be instantly productive.
It is tempting, if the only tool you have is a hammer, to treat everything as if it were a nail. - Abraham Maslow
are you a programmer? have you programmed with both Gtk+ and motif? or are you just another ac spouting off when they really don't know what they are talking about?
Wrong. The Video Toaster crowd killed the Amiga, not the suits as you put it. Pull out the last year or so of Amiga World before it bit the dust and tell us just what most of the articles were about. You got it, the Video Toaster. Now for the Video Toaster crowd this may have been just great, but for people like myself who owned an Amiga 3000 as well as an A500 before that but really didn't give a damn about the Toaster it was one of the things that convinced us to jump ship before Commodore bit the dust.
This standard wouldn't be arbitrary and linux wouldn't bend to fit it. It would be a dynamic living standard that would change with the community as the community needs it to. It would be more than a package manager (hopefully it could do some god because there isn't a really good one yet, they've all got flaws) that would just be a tiny piece of the puzzle. It would probably even be an optional piece of the installation. I'm sick of hearing people complain about RPM and say this is a bad standard, you've probably already got RPM capability in your linux dist. as it is even if you never use it.
The deal is that if linux is going to be a full desktop OS it is going to have commercial source-code-less software. To support that there needs to be some kind of baseline, it's just unacceptable to include all the possible libraries with an application (what good are sharedlibs then?) and it's unacceptable for an end user to buy an application and be expected to go find, download, and build lesstif to make it run.
Not just will it benefit linux by allowing ISVs to develop software more easily (without worrying about all the 'what ifs' of installing the damn apps) but it will make it easier for OSS to distribute binaries also. And I think it is inevitable, a defacto standard (redhat linux) will become the standard or we can take control of the issue with Redhat's blessing and create a distribution free standard.
I'm willing to say the a Linux99 compatible machine can deal with RPMs, has QT (if they ever release the free version), GTK+ 1.2, GNOME 1.0, KDE 1.1, glibc, JDK 1.1.7, and lesstif on it in working order. I'm also willing to say that linux99 apps can be gnome or kde compliant but need neither to run. Essentially we'd pick some libs, name them with linux99 in the name and put them in when you install, you could upgrade to newer versions without breaking anything. Next year we will change it to match the new landscape. As component ware becomes more of a reality we can add mozilla and other components to the standard (you will need to have a mozilla component to run certain apps.)
This doesn't sound so bad to me, sure it might install a few megs of libraries I'm not fond of but if it's optional I won't install it if I don't need it. I'd much rather be able to just install and run StarOffice, Corel WordPerfect, RealAudio, Netscape, or whatever other apps on my linux box than save a few megs or be tied to redhat linux.
Or spend hours trying to force some application to work on my machine. And if I never buy any binary apps then I don't need it but I'll probably end up installing most of the linux99 standard simply because I'll need parts of it for a lot of OSS apps.
What the LSB needs to be is a standard set of tools that the machines need to be "standard." We all want our TAR to work the same and give the same flags. Something similar happened in the UNIX world a few years back. Everyone started sharing applications from one another to make a "standard" UNIX. Sun / Solaris gave out the works of NIS+. Others similarly contributed (I have no idea what though). If we can envelop a certain path to standardization be it RedHat RPM or Slackware's pkgtool or whatever, let everyone use it and Linux will get more backing from companies that want to produce binaries for Linux, but not Source (like Lotus).
Clinton made me a Republican. Bush made me a Libertarian. Trump is making me question reality.
Amen!
"Problem with that is, that RedHat is one of the main people in the LSB effort. Most of the past discussion about LSB has been about what package manager to use. I thought it had died, and was glad."
You know, I have to say, I can't take anything in this post seriously after the opening sentance. The LSB mailing list is hosted by Debian. The site isn't hosted by RedHat. The activity on the mailing list seems to show RedHat hasn't pushed for anything that much, and definately not for something that would specificly benifit them.
The LSB is NOT a behind closed doors project, and it is not going to be bulldoged by one distribution. Following your logic, the LSB would be doomed to include a commercial X server because the LSB written standard Technical lead is Stuart Anderson and he works for Metro Link, Inc. That more consistant with your logic, and we all know that won't happen.
Guilt by association is absolutely rediculus, and your stretching to even prove the association! It would be wise to read up on it, and check it out yourself at thier website a bit, join some of the mailing lists, and learn more about it before you condem something you don't seem to understand.
I don't think having a common basis would be a bad thing in itself. If you say the next releases of X and Y distribution should have spiffylib.0.63 and wowlib.1.1 or higher that just makes development easier for everyone, open source developers included. Commercial Standards make me nervous, what happens when some company comes out with a proprietary lib that makes thier programs run (vb300.dll for Linux AAAAAAAHHHH!) And then they push to make it standard in all distributions that can be called standard complient. If there is going to be a standard it must be minimal and it MUST be opensource. QT wouldn't even qualify until 2.0 is out. Ligjpeg, libungif, fine suggest a minimum version for compliance. But the first piece of closed source that makes it to the standard will stop several distributions from adhering to it.
Insert pithy comment here.
They have done something to set it apart... Netware support. Sure, there is the ncpfs stuff, but last time I heard, it still doesn't support NDS very well (if at all). Granted, they haven't updated it in a LONG time. The administration tools suck, and you only get 8.3 filenames on ncp filesystems. If they'd fix those two things, (as I have written to the begging for), they'd be moving in the right direction. Also, they need to move to glibc like everyone else in the world...
Another interesting fact, is that Caldera was the company that actually developed RPM. This happened way back in the days of Caldera's CND, which was a set of Caldera extensions over the top of Red Hat. Today, Caldera still looks a bit more like Red Hat (under the hood) than I care to talk about.
So, I guess to sum up, Caldera's biggest problem is that they are a slow-mover. RH 5.2. SuSE 6.0. Caldera 1.3?! Guys, put some money into devel!
I don't think it necessarily solves the problem for binaries-only software releases the article was addressing.
That having been said, I have to take this opportunity to say that I also like the FreeBSD approach. From what I have seen it gives you the most flexibility and control. I like its source-oriented approach (ah, the Good Old Days when everything was source only!). I would prefer a FreeBSD-like Linux distribution to those currently available.
Oracle runs on every distribution I have.
Netscape communicator 4.5 runs on RH, Debian and SuSE.
And it worked on my RH box before and after upgrading from 2.0.36 to 2.2.
Pre compiled products can statically link and run most anywhere.
If the authors are dieing to ship a smaller more elegant chunk of code... let them join the open source movement.
I'm afraid that you've got yourself the wrong version of Linux then. Red Hat aren't as conformant to the few standards that we have in place already, which is a great shame because that's the distribution that corporate-types are flocking to.
The general educated position is that Debian is, for those with the experience to cope with, far better with that sort of thing. In fact, Debian can also upgrade everything with two commands (look at apt-get), with the added benefit that everything is guaranteed to be thoroughly quality-checked before release. As a power user I'd certainly recommend it over the other distributions.
I agree on your view that Linux needs a FreeBSD style upgrade model though, as an option.
Having different files doing the same job, in different paths in different distributions is nutty. No-one should waste time trying to defend this.
/usr/local/samba, yet my original SuSE distribution had installed the old ones (binaries from rpm) somewhere else. What was the point of that? Some guy in SuSE had actually spent time shifting 'em about, making his distribution fragile to the first time I needed to upgrade samba without going back to SuSE for a new rpm.
When I upgraded samba I was surprised to see that the genuine samba make throws everything in
Perhaps rather than make a clean break to conform to the standard, filesystem links can be used to good effect to 'upgrade' old distributions' layouts to conformance without much pain.
But things like different configuration file formats are just *wanton* and one way or another will get weeded out by evolutionary forces.
One other thing, the standard clearly needs at least two levels to it. The first, core, level should talk about where files ought to live, format of configuration files, which standard-C libraries must be available and so on: issues relevant to even the smallest embedded Linux application. The guy who replied to this thread above me cursing his HDD space getting eaten with (hated by him) Gnome has a good point if your microwave or settop box is running Linux from flash. The second level could address X-Windows and window manager issues in systems where that's appropriate.
The guy way above who said that you shouldn't be frightened of these kind of genuinely open, changeable and debated standards was right: the thing is to standardize to core of things without tying people's hands to innovate and remain complient. You need to standardize a virtual base class and let people derive!
is now strain within what was once a multi-vendor agreement to support something called
the LSB.
I confess to not following these developments closely enough. Could someone knowledgeable
perhaps post an objective summary of the current state of affairs?
In general, I think that posters here have expressed fair-minded skepticism regarding the
need for--and viability of--an overly restrictive LSB. However, I think it fair to say
that Linux distributions are as compatible as they are, due to a desire to remain so.
There is real danger that some Linux distributor (eg, Microsoft, Corel?) might emerge with an agenda
to embrace and extend the system in proprietary ways. (Corel statements that they intend to develop proprietary UI "standards", and that they see an opportunity to "dominate the UNIX market", show cause for concern.) However, the danger would lie not in
the attempt to do so, but in the willingness of ISVs and consumers to accept this tactic.
I am indeed beginning to hear--from consumers and IT people--language like, "we're looking
to Redhat to provide...". This is bad news for the community, because even though I doubt
Redhat really has any intention of "consolidating the Linux market," the Redhat of 5 years
from now could find itself the "defacto standard"--which would indeed greatly stifle the
innovation and competitiveness Linux represents today.
Ultimately, the Linux community is going to have to be very savvy about this kind of move,
and even about general trends, and develop an ability to mobilize opinion against
proprietarization attempts.
NB: Applications are NOT required to choose static linkage, if they don't like the
library layout of a given system--and many large UNIX apps, such as Oracle, already do not.
An application can set LD_LIBRARY_PATH, and use libraries shipped with the application.
For GPL'd libraries, this means making source available--however, this is much less a deal
than ISVs might imagine. Savvy companies like IBM and HP will have no difficulty with
it.
Matt
is a good one. I personally think that it can become something really good, if and only if LSB doesn't try things like..."each dist will have glibc 2.1, and GNOME 1.0, and kernel 2.2.1, and blah, blah, blah." Something along the lines of the basic tools, libraries, and maybe kernel version that should be included.
if my wife asks a question, and I'm not paying attention, and then I answer wrong does it still count
Don't mistake me, I use both KDE and GNOME. Personally, I think both have a place but to say to me that I must have this interface, and thus it's supporting libraries, is not what it should be about. Like I said basic utilites and libraries is what it should be. Or if they do what to go with a standard GUI, then it should be something like FVWM, MWM, something very common and basic. Slackware will probably settle on KDE (last I read about it anyway), and that's fine, but that's something that should be left up to the maintainer(s) of the dist. Support libraries (glibc, X libraries, etc) and the like should probably be a standard, no one says that you cann't add to it, or even follow it for that matter. It's just easier for new users, and developers (not just ISV's).
if my wife asks a question, and I'm not paying attention, and then I answer wrong does it still count
That's a great idea really, layering in a way similer to Posix might not be a bad idea at all. One might ask "if we have Posix what do we need LSB for?" Answer Posix is pretty much general UN*X, LSB would cover Linux thus giving us a standard to throw at ISV's and say "you want to port your apps to Linux, here's what you need to read to work on these distros." Really, I'm seeing mainly Slackware users say how bad it is, but I have to ask you then, what makes you think that anyone is going to make you follow the standard? No one's going to kick in your door and force you to comply to the LSB.
if my wife asks a question, and I'm not paying attention, and then I answer wrong does it still count
Netscape manages to produce a decent interface using Motif. All this proves is that with sufficient work you CAN produce a reasonable interface with Motif; not that Motif helps.
...
I'm sure the Netscape folks would have some choice words to say about the quality of Motif. Certainly everyone I know of who's used Motif for serious applications hates it
I'm not sure how much better GTK is having not used it much. Netscape's major reasoning behind going to GTK is not because it's BETTER but because it's free software and it's easier to get people to work with it.
I don't know whose fault it is, but more than once I've seen /etc/passwd world writable on old Caldera boxes. More to the point, they're so preoccupied with running commercial software on Linux in binary form that they strategically leave out .h files to make it hard to compile things...
I'm just curious - why the animosity towards Red Hat? I've never really understood... What makes it fruity and umbrella-laden? The addition of GUI interfaces? Is that such a bad thing?
I mean, all of the distributions are just that, right? Distributions of 100s of independent packages. Init scripts may differ, some of the directory structure, perhaps... but at the end of the day, what is really so different? Especially after you start customizing your favorite distro to behave the way you want it to behave?
I'd be interested in some un-emotional, fact-based insight into why people believe that one distro is superior to another.
Where are the details?
What are they proposing exactly??
If it doesn't meet and exceed the best ideas we can come up with, we shouldn't standardize yet!
That has killed many things (ADA, X.400/ISO, etc. etc.). Good riddance, but let's not follow.
sdw
Stephen D. Williams
Netscape is the only program I use in Linux that is consistently ill-tempered...
It's statically linked with Motif which makes it a 11 MB monster that crashes, takes 20 seconds to load,
and sometimes decides to take off with the cpu just to run an animated gif or something.
I will be sooooo glad when 5.0 comes out.
Simply enough, the "problem" is that Caldera's semi-proprietary offerings are being beat to a pulp in the marketplace by the 70% marketshare holder Red Hat, which is putting out essentially everything under GPL or other free licenses.
Look, it takes less effort to make your app run on all major distros than it does to get it to run on Win95 and NT, and your backward compatibility isn't much different than trying to write an app that runs on NT 3.51 and NT 4.0 (gods forbid NT 3.1).
No LSB? It means the other distro makers have to play catch-up with Red Hat's decisions, instead of getting to help make them.
It also means immunity from stupid mistakes -- if Red Hat screws up, another Linux distro will replace them as the leader. If a universal LSB screws up, however, a non-Linux (and not necessarily free) OS will win.
"it's not law and it's not in stone. If LSB screws up then nobody uses it, we go back t othe drawing board and make it better."
And how long does it take to make the decisions, how long does it take to realize it was a mistake, and how long does it take LSB to correct it? And how much damage happens in the meantime?
Forget a standards committee. Let the rough, anarchic consensus of the Free Software movement and the marketplace solve it. It got us Linux, didn't it?
Is there really a chance of a major binary vendor not following the LSB ? I mean, we're probably not talking about anything more than the file system standard and C library and kernel versions here. Either the author of the article is indulging in angst for no good reason, or he knows something I do not.
The major problems running commercial packages are mostly related to the shared libraries ( in particular, libc ) . For example, Applix wants libc 5.3.12 , Star Office 4 wants libc 5.4.x , and you have a glibc distro, where all your other apps want something else. What we are seeing at the moment is a mix of the following:
(a) Big statically linked commercial apps
(b) Users, often newbies, crying out in pain as they try to configure their system to handle severl conflicting library versions.
I , for one, would be appy to see this stupidity come to an end. Even if the distributors were to do no more than synchronize kernel and libc versions, this alone would be a vast improvement.
I do not see a major advantage in having different libc and kernel releases with the different distros . There are some issues where the users want choice ( say, package management ) , and there are some areas where choice makes life difficult for everyone ( shared library versions and directory structures )
Personally, I'd love to see the distributors standardize on releases of the major ( libc and libg++ ) shared libs, and possibly do the same for the non core libs ( gtk , jpeg , gdbm , ... choose a version of qt, so that *if* qt was shipped, it should be version X ) . Is it too much to ask that linux be binary-compatible with itself ?!?!????
cheers,
-- Donovan
--
Donovan Rebbechi
No, it must give them, the ISVs, something to "bend to" -- state its own "whims" clearly, that's all.
No, but to make life easier for consumers to buy software from those ISVs.
Yeah, it's "doing just fine" among those few (relative to the myriads of Windows users) hackers and techno-nerds who use it today. But who cares about the great unwashed masses -- as long as we "L33T D00DZ" can have our toy to ourselves everything is "just fine", right?
Heh... Yes, "glaring" is a good word to use here, given all its connotations... And given those, I'm not sure it's the word you should have chosen, if you wanted to paint a rosy-red picture like it seems you did. That lovely lava-lamp mood light of media attention can soon turn into the harsh glare of oncoming headlights (mounted on a truck known as "wide-spread consumer dissatisfaction"), if Linux doesn't live up to the feel-good hype it is currently recieving.
Yeah, that will be a hit with the Mom-and-Pop crowd! Hey, what exactly are those "--switches"? And are you sure they'll work on any distribution? Mom and Pop won't be able to figure them out, if the software package they just bought is configured for a distro other than the one they are attempting to install it on, you know.
Oh, forgive me for not realizing that Linux was built for your pleasure and your pleasure alone. BTW, what exactly is it that you "already" do with Linux -- Web surfing, IRC and games, I suppose...?
The only problem being that gnumeric will run "out-of-the-box" only on Corel or Debian, KOffice on Corel or Red Hat, and AbiWord only on Debian or Red Hat, or something like that. The difference, furthermore, being that MS Word and Excel -- and Lotus 1-2-3 and WordPerfect for Windows -- run "out-of-the-box" on all the distributions" of Windows 98 and NT 4! (Yes, I know there's only one of each! :-) (That's why we need a little standardization among distros, to achieve those same advantages.)
Again, forgive me for not realizing that you and you alone have the authority to decree that ISVs who want to release proprietary software on Linux -- and people who want to run that proprietary software -- should just fuck off and die, what with Linux being your and the other "L33T D00DZ" personal playground and all...
And your needs are of course the one and only World Needs Standard -- now that seems to be a Standard you're in favour of...
??? Are you stupid, or what's wrong with you? Of bloody course it will hurt Linux's market share if people who could use Linux if there were a standard for non-OSS ISVs to write to -- can't, because there is no standard for non-OSS ISVs to write to!
No, it won't "suddenly DIE" -- it can live on among those few (yes, relatively speaking it is still a few) hackers and techno-nerds who use it today. If that's what you want, to doom Linux to such a marginal existence, and denying ordinary end-users a realistic chance at using it, then just say so. But please don't try to pass it off as striking some great blow for freedom -- it's just the opposite.
Christian R. Conrad
Opinions are MINE, not my employer's -- Hedengren, in Finland.
Christian R. Conrad
mail me at iki.fi ; same user ID as here
You know, this is exactly the kind of comment that makes me want to switch to Debian or SuSE or roll my own. Linux has gotten along just fine by being diverse for a long time, and I really think that we all ought to remember that the way to achieve World Domination (tm) is NOT by beating the old-style SW companies using *their* rules. We are getting there just fine by using our own...
The human need for a "heard" and to flock around "the leader" never ceases to amaze me, but thanks to the GPL there will always be a nice, custumizable distro for those who want it - even in a MS/RedHat/IBM/HP/Euro/whatever ruled world. That is the real beauty of it all.
This attitude is fine, if you wish for Linux to remain a hobbiest Operating System. However, since most people would like to see Linux put a serious dent into the M$ monopoly (including Linus), the answer to that is users. And users will not move to an OS that lacks the applications that they need or are familiar with. The average user doesn't care about the kernel version or which version of glibc he or she needs.
If the user cannot *use* his or her machine in an effecient manner, then they will not change, technical superiority not withstanding. Period.
The media has shined a very large spotlight on our community. Very large companies have voiced support and some have even begun to provide very good products. However, without an increasing user base, how long do you believe this support will continue?
You know, before reading your post, I thought that having LSB was generally a good thing as long as it was open and flexible to new developments.
However, you bring up a couple of really good points. One of the things I like about Linux is that it is co-operative (collaborative? collective?) effort of people with many different points of view, but one larger (yet amazingly hard to define) direction.
I then had a realization...
The LSB could easily be the first step to the Linux Cathedral.
Yes, we need standards. Yes, we need protocols and APIs. We've developed those using the current structures. Do we need an arbitrary group of people determining our direction for us?
I think this could be one of the more interesting questions for the next year or two, as the different allies in the Linux effort (distributors, users, hackers, ISV's, and proprietary software companies expanding into the new market) realize that their own goals are subtly (or not so subtly) different from all the others, and want to have some control (or lack of it) over the direction of Linux.
This could easily lead to the fragmentation that has plagued Unix and other great endeavours in the past.
--
A.
--
The Internet is the Suppository of All Knowledge. You get it in the end.
Xt has some significant bulkiness and design issues. There are large portions of Xt that just feel unpleasant and unwieldy to use, in addition to requiring too much code to implement. (Having event translations as a user-alterable resource springs to mind.)
I don't have a good idea of how GTK stacks up as I find the idea of bringing out yet another UI library that's designed to do OO in C instead of C++ to be rather revolting. I do recall GTK being rather more simply and cleanly implemented when I gave it a glance awhile back though.
Need a Python, C++, Unix, Linux develop
Xt (and hence Motif) has some significant bulkiness and design issues. There are large portions of Xt that just feel unpleasant and unwieldy to use, in addition to requiring too much code to implement. (Having event translations as a user-alterable resource springs to mind.)
I don't have a good idea of how GTK stacks up as I find the idea of bringing out yet another UI library that's designed to do OO in C instead of C++ to be rather revolting. I do recall GTK being rather more simply and cleanly implemented when I gave it a glance awhile back though.
Need a Python, C++, Unix, Linux develop
I dislike GTK because it isn't written in C++ and it should be.
One problem I see with LSB is that too much standardization is bad. A good OS will succeed in a marketplace (ecology) because it can adapt to fill a lot of niches. There's a shrink-wrap application/home user niche. There's a heavy duty file server niche. There's a web server niche, and there's an imbedded applications niche. There are TONS of niches. Linux is adaptable enough to fit into all of them.
The most important niche from a standardization standpoint is the shrink-wrap niche. That niche is large enough to have several players, and they will only badly hurt eachother if they don't cooperate to some extent. I see GNOME, and possibly LSB as a step in the direction of getting the shrink-wrap people to cooperate to the extent necessary to create a stable marketplace niche for Linux.
LSB is bad if it attempts to make ALL Linux distributions conform to the standards necessary to compete in the shrink-wrap niche. Distributions should be clear about the needs they try to fill, and standards should be flexible enough to let them fill them.
Need a Python, C++, Unix, Linux develop
Do not see the problem here:
Even slackware has rpm utilities. And all systems have tar and gzip.
All systems shipped at the moment have both libc and glibc (with minor glitches like wich one is preferred) so no problem there either.
All systems ship with a 2.0.35+ kernel and none with 2.1.x or 2.2.x
So if an ISV is a moron and does not see the fact that a properly linked package will work on any system, then to hell with it and its venerable product.
If the ISV is not a moron, than there is no problem here and Oracle and IBM are proving this pretty well.
And there is more than enough standards as of now to follow (i mean those that are not completely implemented). Overlapping them with a new and artificial one is simply stupid...
Baker's Law: Misery no longer loves company. Nowadays it insists on it
http://www.sigsegv.cx/
Do not see the problem here:
Even slackware has rpm utilities. And all systems have tar and gzip.
All systems shipped at the moment have both libc and glibc (with minor glitches like wich one is preferred) so no problem there either.
All systems ship with a 2.0.35+ kernel and none with 2.1.x or 2.2.x
So if an ISV is a moron and cannot design a properly linked package that will work on any system, then to hell with it and its venerable product.
In btw: this is what mostly makes windows life so troubled. So letting these guys into linux will quite windowize it... It is better to keep this kind of lousy "developers" OUT!
If the ISV is not a moron, than there is no problem with having 4-5 distributions. Oracle and IBM are proving this pretty well.
And there is more than enough standards as of now to follow (I mean those that are not completely implemented into Linux). Paying attention to a new and artificial one is simply stupid...
Baker's Law: Misery no longer loves company. Nowadays it insists on it
http://www.sigsegv.cx/
I would suggest you take a look at Debian. They have developed a wonderful utility apt-get (soon to get a nice gui front end). Whenever I wish to get the latest changes I type two commands:
apt-get update
apt-get upgrade
Everything else takes care of itself, dependencies are checked and updated, old programs are removed, and the system works as a Debian distribution should (which is very stably). It is possible to add these commands to cron, and your system will remain up-to-date without any routine intervention.
Enjoy,
cartographer
The only reason I still run Windows at all is to play games. It might be wishful thinking, but it would be nice if games (like baldur's gate) were made for linux. Then I could drop windows completely. If there was some stardard that would make it easier for companies to make sure that their programs would run on all distributions, I think they may consider porting more seriously. A company isn't going to want to spend money to port a program to linux and then have it only run on one distribution. If you think interplay or or other commercial game companies are going to release the source to their current games so we can fix the problems your crazy. Also, a company doesn't want to have to deal with making multiple package files. If a company is going to sell a shrink wrapped game in a store they aren't going to package three cd's, one for rpm, one for deb, and one in gz because it costs more. This is assumning that the game is large enough that all all 3 package types wouldn't fit on one cd...
For the above reasons I think a standard would help.
So, let me get this straight. Linux, a kernel that started out as a project to learn the quirks of the 80386, and then became a "better Minix than Minix", must now bend to the whims of ISVs?
:-), linux isn't going to suddenly DIE because a few ISV's decide it's too hard to support Red Hat, SuSE, AND Debian.
Linux, possibly the most successful GPL'd project outside of GNU, must stifle experimental development, and lock itself to an arbitrary standard (beyond POSIX), to make life easier for ISVs to keep their source code proprietary?
Well, I only abandoned WinNT 6 months ago or so, but it seems to me that Linux has been doing just fine before all the big names started porting and supporting Linux. And, now that the media attention is glaring, I see no reason for the current system of distributions and flexibility to change.
./configure --switches; make ; make install
After all, even if IBM, Intel, and everybody suddenly stops supporting linux, us Free software and Open source types will still be able to do with linux what we already do with it.
Only, as time goes on, free software alternatives grow stronger and stronger. Any Day Now (tm), gnumeric or KOffice or AbiWord or something will be able to do all that MS Word or Excel does for people. Real Soon Now (tm) Corel or Red Hat or somebody will make a distribution that truly is as comfortable for users to install and configure as Windows 98.
Those of us who like linux are content to bite the bullet and use Windows when we have to, and wait until linux gets all the productivity software it needs, with source code included.
Those of us who have looked at perl, egcs, emacs, KDE, and more know that we don't need no stinkin' linux-specific standards, and we don't need no stinkin' proprietary software.
It may sound like wishful thinking; many people do need proprietary software to get their jobs done. But, those people who need that software aren't using linux right now anyway. So, if ISV's don't develop for them, then it won't hurt linux's market share at all.
linux is already doing just fine, even without a prime-time Desktop Environment. So, barring some foul play on the part of transmeta's flying saucer technology
About playing well with other unices, wasn't it informix who said, when questioned on how much work it took them to port to Linux, "we typed make". I.e. the source that they were using for sun compiled cleanly and worked under Linux?
They laughed at Einstein. They laughed at the Wright Brothers. But they also laughed at Bozo the Clown. -- C. Sagan
RedHat is basically the leader in the Linux community... Redhat went to rpm.. and so did Caldera, TurboLinux, SuSe, and most others. RedHat went glibc2, and now the rest of the distros will follow.. and are..
.. other than that the distros may have a few different libs....
... threads are good and in many programs they can improve performance... and on multi processor machines (like mine) they can only be a good thing....
Although Debain and Slackware are good distros, and WILL be around for a while, they are both a bit behind.. debain lacks in a good configuration tool (as of 2.0), so does slackware...... they are both excellent tools for learning th einternals of the Linux system, as they require you to edit almost every file by hand, rather than thru a GUI...
Redhat is a good distro, but because they are often the first one to go to the bleeding edge Linux stuff they have had some stability issues in the past. but I think they have overcome most of them (RH 5.0 & 5.1 are known to be a bit unstable in some areas.. duh they released 5.2 to fix these)..
I use SUSE.. and it does follow Redhat model in some ways.. however the installation is the main difference and the setup tool..
there was an article on 32bitsonline that that compared distros.. and there results were that most distros are different in the setup tool and installation
to port to linux here is what I'd recommend
1) most distros are moving to glibc2. is support threads and some other things that are 'modern' in programming and necessary.. so if you are going to port to linux go with glibc2 as the c lib... (now libc2 officially)
2) the kernel version.. if they are just coming on board with linux..start with 2.2 kernel series.. 2.0 is stable but even Alan Cox and Linus T, will probably tell you that eventually they like to see 2.0 die and 2.2 improve...
3) now if they ar porting from windows to X they shoudl pick a tool kit and stick with it.. here I'd recommend wither gtk+ for C or qt for c++.. I use gtk+ and find it very easy to use.. but I am a c programmer..
.. and I think most people with a clue will agree that this will be Linux 99... glibc2, rpm, kernel 2.2.. although I have not move up to it yet myself.. by the end of the first quarter of 99 I will have upgraded from suse 5.2 to 6.0 + updates....
Only 'flamers' flame!
Blech. I (and most of us I believe) aren't looking to just be better than NT, why be so AMAZINGLY short-sighted? We're looking to be the best we CAN be, which is alot more than just better than NT. Not all of us are just microsoft haters. Also, this hardly seems a distribution issue, but more a library/kernel issue. There are only two libraries in general to be concerned about, libc5 and glibc2. There are two kernels to be concerned about, 2.0.x and 2.2.x (I realize there are obviously subtle differences in each release, but those generally don't affect what the world sees). This doesn't seem all that bad as I suspect in the future libc5 will phase out (Assuming glibc2 becomes more stable and such), as will 2.0.x, as 2.2.x becomes more stable and accepted. This may take 6-12 months, or maybe even a year or two, but it'll happen.
Obviously the community is not going to listen to a committee or company that attempts to impose some kind of standard on the community. In fact, such attempts can only undermine the reputation of Linux and will _cause_ splinters in the community (as this Slashdot discussion surely indicates).
If I was an LSB supporter, I would come up with your standard, and recommend it to the community (so far, I haven't seen a standard yet, just a lot of talk about why there should be one). If your idea is good, the community will adopt it and assist in making it better. If not, well, the community will tell you where you went wrong. Show the code, or in this case, show your standard. Noone in the community is likely to take you (the LSB supporter) seriously unless you provide some solid recommendations. If they make sense, then great! I'd be happy to support them.
The problem here is that an LSB standard must stand on its own merits, not on theories about what should be right for Linux. Until that is understood, no progress will be made.
Jason
> The real matter is what can be if corps really
> move to Linux ? Without them, I am sure we will
> never rise at M$ desktop standards.
The question is do we want that.
In the early days of computing, the time of the
C64, ZX Spectrum, BBC etc, computing was a lot
of fun. There was this whole bunch of hackers
that tried to figure out the systems and program
for fun. Then the suits came in, developed the
PC, and Windows, and took away much of the fun
and orginal spirit. Now with Linux computing
and computers are fun again. I am afraid all
that will be gone again if we let the big
companies over again, and try to be everypersons
OS.....
look at first Unix versions. It factionized, and it was the mess.
If there is no minimal Linux version, big corps will distribute their own versions (isn't Corel starting this already ?). This will bring :
* either chaos with more distributions
* or corps apps will end to being tied to their distribution (and say bye-bye to freedom). Just imaging that M$ distribute some Linux version that can easely run windows apps. I am sure they will if Linux is threatening them. Years ago Billy said : "Internet is nothing, we have MSN" but one year later he embraced Internet... You might be horrified by that idea, but plenty of M$ clients would be interested... Another argument in this favor : Win is a mono user environement, M$ never believed in thin clients, but they changed their mind with NT4 Terminal Server & win 200?
The world belongs to those who get up early. - I'm far from being the king of Earth then
People just don't seem to get the difference between a kernel and a distribution. IMHO, nearly everybody who has a linux uses a distribution with (g)libc.
If people want to make programs that do things differently, they'll have to provide statically linked binaries. So what?
Linux wise, I am also not distribution-centric yet. So all of the flame wars between Debian, Slackware, Caldera, RedHat, and SuSe strike me as being alot of wasted use of bandwidth and mindspace. As are the non-stop commentaries about KDE vs. Gnome, Gtk vs. Motif/LessTif.
As a developer, what I want is a clear set of API's and/or published specs which allow me to write and test my code, then point it at a target platform, knowing that as close to 100% as possible of the code will function properly on that target platform.
Which is what the Win32s API, etc. do well (that is, not perfect, but enough to get the work done.) I have written apps in the past that worked without any code changes on platforms from Win3.11 (with the Win32 library), Win95, and WinNT. They even functioned correctly under OS-2/Windows without problem.
I guess what it boils down to is that if the LSB initiative results in a more consistent API across platforms, then I'm all for it.
...Open Source isn't the only answer -- but it's almost always a better value than the alternatives...
Only experts can test new applications, because there is nearly no real computer newbie who uses Linux. But a big corp can put money in these kind of tests. Therefore, if we want Linux to be stronger, and being a real alternative to M$, we need support from corps.
Your argument is valid but not sound. I think I read here that GNOME is going to be installed on pretty much every computer in Mexican elementary/secondary schools. You couldn't pay for a better usability lab than that.
In theory I think it's a good idea. I doubt the open source projects need this, but the binary-only software could really benefit from this. It would go a long way in helping ISV products because they could test against a standard, as opposed to testing all major distributions on all the platforms. It would also solve the problem of "splintering" versions of Linux that a lot of people outside of the Linux community seem to constantly predict.
I doubt, however, that it would actually work. The problem is that Linux software is changing so rapidly that any standards base would be swamped trying to keep up. A lot of software changes every few weeks, not every few years like traditional software. By the time any panel could agree on the Linux standard it would change and they would need to start over.
If people really want something like this, I think a greater benefit would be gained by participating in other Unix standards organizations and meeting their certification tests. A lot of the new vendors supporting Linux now also run on other Unix platforms. If making life easier on them is the goal, let's play nice with everyone instead of creating yet another standard.
Unfortunately the issue adressed in the paper is a concern. Maybe I dilusional, but lately more software seems to only work with certain versions of other libraries. For example, Window Maker, GNOME, & the Gimp seem to be very version specific on some of their libraries. A third party vendor would have a nightmare trying to ship a product against constantly moving target. The question is how do you "standardize" without slowing development to a halt?
Kernel - Choose version closest to the pure Linus Torvalds Linux kernel. I'm personally planning to get Linux for my PCI PPC PowerMac, and I've chose LinuxPPC over other versions, such as MkLinux, even though it is harder to use than MkLinux because it is more pure. MkLinux is slower, and is built on top of the Mach microkernel. Mac OS X was originally meant to be on top of Mach, but now it appears that Apple intends to make Mac OS X pretty much a version of Unix with the Mac user interface. Apple is trying to migrate the Mac towards being a Unix box.
DES Khaddafi KGB genetic jihad Uzi Rule Psix Qaddafi cryptographic Peking Mossad Legion of Doom Albanian Serbian Saddam
Saying that POSIX compliancy, SVGA, X Windows, and Motif Libraries should be all ISVs count on locks us into old technology. What LSB is all about is providing a framework that can be used for the future. We will move beyond glibc, and I want better than Motif (which is actually even uglier than Windows).
Today Linux has a shot at what advocates have always wanted, widespread acceptance. And the only way it is going to continue to succeed is with modern commecial applications with a modern look and feel. (and locking ISVs to Motif, or Lestif, certainly isn't going to do that.)
You may consider package managers to be "stupid" but again if Linux is going to get widespread acceptance, it needs to be easy for the average user to install and remove programs. I really think that RPM does that. Use one of the GUI frontends, and you've got a pretty easy to use system.
What you don't seem to understand, is that this whole movement doesn't have you as its target. I'm not its target, Joe Slashdot isn't its target. What this will do is bring more Windows users to Linux. Most of whom expect the software that they just bought at Best Buy to run right out of the box. (Which isn't too much to ask) He's not going to mess with libs, he calls tech support immediately when he sees DLL or VXD.
If Linux provides only what the "old gaurd" sees as bare essentials, then it is not going to succeed. What RedHat is doing for Linux right now is going to give it a chance to really compete in the OS market. I am guessing that the next redhat release has some real potential to be the breakthrough release. I think with Gnome 1.0 it has the potential of attracting a lot of consumers.
Oh, and GTK rocks. I mean come on, how can a 100% themeable toolkit not? If you like Motif, use the motif theme... Windows, the Redmond theme, whatever. Personally I like the BeOS theme... And I have been told that it is simple to use, but haven't used it myself, us web developers don't get out much...
Can you believe the arrogance of that bunch of suits?
From the article:
Intel likes the idea of open source software but is afraid of too many cooks in the kitchen.
What do those idiots think "open source" means, anyway?
For this reason, the company will announce next month the formation of a consortium to help ensure a unified vision and common implementation of the Linux operating system, company officials said last week.
Yeah, right. These assholes finally figured out that the Linux community has something worthwhile, and is trying to position themselves to be in control of it.
While Intel execs believe in the open-source model, they're concerned about potential fragmentation caused by too many disparate implementations.
Bullshit. Fragmentation of what? the kernal? the libs? The beauty of open source is that if something doesn't work with the system that you have you can fix it so it does work.
"Not everything has to be shared all the time," said Ron Curry, Intel's director of marketing.
Now that one really takes the cake. Thinking like this is the reason that Linux is kicking Microsoft's ass.
So now all the bandwagon jumping suits-without-a-clue have decided that the grass just might be greener in Linuxland and have opted to join our parade. Be afraid.
Symlink