I don't think you quite understand how shared libraries work. The reason BeOS includes this stuff in the kernel is not to boost performance, but because library loading is closely tied to program loading (programatically they are quite similar) and by putting both in the kernel you can share the common code for both. The BeOS feature I was referring to requires some knowledge of shared libraries. When a shared library is loaded, the code has to be relocated to whatever memory addresses are free. This is a time consuming job, so the dynamic linker caches the results of this process in memory. However, when the library changes, these caches have to be updated. BeOS supports all the symantics of correctly updating libraries as they are changed by the system. None of this requires any user intervention (like running ldconfig) and is handled transparently by the system. Its mainly the difference between a process that should be automatic, but is implemented as a manual one in Linux. (Just like module loading, but lets not go there...)
As for the GUI and browser, BeOS is a microkernel. It even implements networking in userspace. I think positively ugly that Linux implements stuff like audio in kernel space;)
Which circular dependencies are those? Oh, you're probably one of those people that can't manage to grasp the concept of installing two mutually-dependent packages at the same time solves this "problem".
>>>>>>>>>
Why does everyone assume that when I complain about a problem in Linux, I can't fix it? I've been using RPM quite a lot. I can figure out how to type rpm -i XFree86*.rpm in order to install all the circularly dependant packages. It is still a stupid idea. If two packages depend on each other, they should come as one package. Given the huge amount of bandwidth available these days, there is no excuse for KDE to come as a dozen different packages that only.01% of the population will install seperately.
Not a fan of modularity, or a maintainable package model, eh? Slack's fine for one or two machines, when you go beyond that, it's a nightmare to
maintain.
>>>>>>>>>>>
Slack's a pain to maintain? How? I don't know what experiences you have (I'm not a sysadmin and proud of it) but I see RPM's "modularity" as a stupid splitting of of stuff that really shouldn't. Modularity does not refer to splitting stuff up for the hell of it. There has to be a good reason for doing so. Its like if Linus seperated out the Linux VM into a seperate module. You can't use a different VM with a particular kernel, and in order to use the kernel, you have to load the VM. So what's the point of splitting the two up?
Consistent package management is key. Doesn't matter if you're talking about RedHat, Mandrake, Debian, SuSE, or even one of the "fringe" distributions like Connectiva or Immunix. All of those have good working package models. I'm sorry, but tar and gzip does NOT constitute a packaging system.
>>>>>>>>>>>>
RPM is not a packaging system. It is a POS created by people who obviously hate you. It requires a database (databases are evil!) to track packages, it doesn't mesh well with compiled software, it has stupid dependency management, and it takes too much control away from the user. That's why the packaging system I'm writing for BeOS will (hopefully) offer the power of a packager without all the crap in stuff like RPM.
It is nice, however, to see Pat Volkerding putting some *RECENT* versions of software in his distribution, at least for the most part (XFree86 is a notable exception). It wasn't all that long ago that Pat & co. shipped packages that were 3, 4 or even more versions old...
>>>>>>>>>>
How is anything in Slack 7.x NOT recent? They're glibc 2.2.3-based distros, and Slack 7.1 was the first distro to come with a KDE2 beta.
Star Trek Generations.
on
Space Diving
·
· Score: 2
Is it just me, or was this ripped straight from the book Star Trek Generations? You know when Kirk goes through his whole adventuring phase and does the space diving?
To tell the truth, I have found Slack to be the easiest Linux distro I've ever used. It doesn't do a lot of crap behind your back, it installs stuff where you tell it, and it has the cleanest/etc/rc.d. It also doesn't make you deal with the POS RPM, and all the stupid circular dependencies that come with it. While KDE on RedHat is a mess of a dozen RPMs, the same thing on Slackware is kde2.tgz.
ldconfig was, and they laughed as I explained it... until I told them they'd be rebooting on an MS box at that step.
>>>>>>>>>>>>
Except not really. Every other OS in the world is smart enough to automatically register DLLs without running programs like ldconfig. Without rebooting even. In fact, I was reading on article on dynamic linkng a few days ago, and they talked about how BeOS goes to great lengths to make sure that the cached and on-drive copies of DLLs are the same, and can add DLLs on the fly and have the relocated images updated.
Yea, the Windows registry blows. They have the right idea, enforcing policy by creating a standard API to configuration data, but screwded with the simplistic key/value format and the brain-dead storage method. Something like the registry API (more flexible, though) that would store plain text files in a config heirarchy (or wherever the user wants) would probably be ideal. Then, write a GUI program that edits these standard text files, and away go all the configuration problems (or most anyway;)
Because people have hard enough of a time getting past the "be" in my username, and even in topics where I *don't* mention BeOS, they think I'm comparing it to Linux. I'm really not in the mood to defend BeOS against a bunch of Linux weenies.
PS> Not all Linux users are weenies, just the anti-BeOS ones;)
Make the system cohesive. Get back to the UNIX roots. You know how UNIX treats *everything* as a file?
From a programmer's point of view, I agree with you (check out Plan9 from Bell Labs sometime, you'll probably like it). But from a user's point of view, that is irrellavant.
>>>>>>>>>>
Not really. A clean, consistant API breeds clean, consistant software.
PS> As for the config files, who *said* they had to be scattered binary files? I just said they should be more structured. If there was one standard for config files, and one GUI tool that could handle that standard (preferably displaying embedded comments) then Linux configuration would be a cinch. As many people have pointed out, OSX does just this.
Actually, with the NVIDIA driver (they're the only ones that have even remotely close performance to their windows counterparts) the situation is reversed. XFree86 doesn't support page-flipping (with good reason according to Carmack) so as resolutions get higher, the memory bandwidth strain of having to blit the frame buffer starts adding up. However, even at low resolutions, performance is lower than it should be, so the page flipping issue is not entirely to blame. Either way, Linux should be *much faster* than the Win2K!
Well from what I saw the difference beteen Windows vs. Linux was pronounced at low resolutions but the difference at high resolutions gets less and less as the resolutions get higher,
until it's hardly even noticeable. This was based on Quake3A benchmarks.
>>>>>>>>>>>>>>>>>>
That's even worse! That means that Linux does a fine job crunching the geometry code, but can't handle sending data to the graphics card quickly enough. That means the disply system is f*cked. (Though with XFree86, I wouldn't be surprised if this was the case;)
Considering that the file is XF86Config, and the command is ifconfig (ipconfig is for Windows NT 4.0), my guess is you don't have much experience in Linux
>>>>>>>>>>
Sorry, typo's. I've not only had a lot of experience with Linux, but I've compile XFree86 half a dozen times from source, used at least one version of every distro under the sun, and went through the painful libc5 -> glibc transition with good old Slackware 3.5
Get rid of/etc and replace it with something sane.
After trying several distributions (among them SuSe, with their rc.config madness), I was so glad when I
finally found Debian, with all of the configuration files still in/etc, at the exact location that the HOWTOs and the documentation of the individual programs tell you./etc can be pretty sane, and it's actually a lot more sane than the registry.
>>>>>
Why does everyone think in terms of the registry? The registry is a piece of junk and always will be. Take a cue from MacOS X to see how things should be done. As for Debian, it sounds nice, but the fact that the layout matches the one in the HOWTOs does't make the layout any better.
You find the directory structure stupid. It's so complicated that your dad can't install programs. Well, tell him to look for what he needs on packages.debian.org and then have him type "apt-get install packagename".
Done.
>>>>>>>>>
True. And I really don't think that sequence would be a problem for anyone. I'm pretty sure the 20-something computer elite don't know it, but these older "illiterate" users DID use *NIX terminals back in tha day...
After that, the docs are in/usr/doc/packagename.
>>>>>>>
That sounds great. If only/usr/local/doc weren't there...
And if you want, "dpkg -L packagename" tells you
what was installed and where.
>>>>>>
That sounds stupid. I'm a hands-on type person. I like the previous example better.
Try that with Windows.
>>>>>
Don't have to. Everything is nicely organized in c:\program files. Though the doc thing is nice.
And if you want to compile everything yourself, have
a look at GNU stow, which helps you to keep the different programs separated so you can easily uninstall/
upgrade them without leaving files of the last few versions lying around. You don't need to know where to
put every single file, that's make install's job.
>>>>>>
Hiding complexity behind limited interface is the first mistake of UI building.
I can't see how one can tolerate the windows\system32
directory "structure" and be confused by a Unix directory tree.
>>>>>>>>>.
I can't tolerate the windows\system32 structure. That's why I use BeOS...
Oh, nice that you mentioned how all UNIX console apps work together in a nice harmony of streams and
pipes. This is the very reason I feel more comfortable with Unix when doing real work.
>>>>>>>>>
Or BeOS;) However, I have found the *NIX text "zen" to be one of the best things about it. Like I said, I don't hate Linux (and therefore UNIX). I just can't stand some of its more inane details.
In this thread, and in pretty much every Linux-related article in the press, there is far too much emphasis on how hard Linux is to install or configure. But there are actually people who install once and work with their machine for years.
And for my style of work, a few shell windows and Emacs are the best I have found so far. And yes, I did work
with DevStudio on Windows, and did not totally fall in love with it.
>>>>>>>
I've tried VI (and still use it whenever I don't feel like opening up BeIDE) but I have fallen in love with DevStudio. It is just so damn handy. (Well, as long as you stay away from the wizards and crap.)
If you can't stand to use Linux, you're not really required to like it. Perhaps it would make more sense just to like Linux' development model or the always helpful and well-informed slashdot crowd instead.
>>>>>>>>>>>>
I *do* like Linux. I just can't stand to use it. Some of its details are what you could call "deal-breakers" for me (and many other people, or I would presume so...)
A) Who said Win2K had a good directory structure? I like BeOS's, but Win9X's is nice too. There are only 3 directories on a clean install, "My Documents," "Program Files," and "Windows." The first is your's to do with as you please, the second is logicall organized by application, and the third is crappy, but you rarely need to go there anyway.
B) Linux takes up just as much memory on my machine as does Win2K. Of course, I'm (was) running GNOME, KDE 2, and the full host of libraries, but only because I need good app compatibility. Don't even get me started on how pointless it is for GNOME and KDE's libraries to be incompatible (and thus memory wasting.)
I know you are looking for flame, or you are actually a Linux supporter attracting flames in windows. Anyway, I fell into it.
>>>>>>>>>
No flames, just honest critism. Like I said, I don't dislike Linux, I just have exceptions with some aspects of it.
If only NVIDIA would release its specifictions to linux community. After their hand-in-hand cooperation with M$ in XBox, I doubt NVIDIA would ever think of supporting Linux community in the future.
>>>>>>>>
I doubt it. History has shown that NVIDIA can drag MS around by the balls. Ever wonder why DirectX8 bears such a striking resemblance to NVIDIA hardware? There is nothing to indicate that NVIDIA will be the bitch in the MS-NV relationship.
Now that 3dfx is bought by NVIDIA....there goes voodoo*. Nevertheless, it's not Linux to be blamed, it's the attitude of those 3d chips vendors.
>>>>>>
Wrong on both counts. It is entirely the fault of 3DFx for making crappy products.
You can't deny the fact how quick Linux is be able to catch up with the performance without their
support.
>>>>>>>>
Which is why exactly 1 OSS driver is faster than the closed counterpart? (Some of Matrox's OSS drivers were faster because the closed versions had major OpenGL problems.)
B) Linux's stability won't play much of a part in this. Win2K is very stable
Win2k is stable? I bet you haven't been trying to upgrade the hardware in a Win2K box. All I got is
INACCESSIBLE_DEVICE_ERROR blue screen. Come on! I didn't even touch the disk! Just upgrade a damn CPU! It
happened in old NT, which resolved this problem long time ago, now they bring it back. Well done!
>>>>>>>>>>>>>>>
Then something is wrong with your machine. There is term in psychology (forget it offhand, its either a fallacy or the false concensus effect) that describes the overgeneralization of the characteristics of one example of a group to the whole group. This thinking is very dangerous because one's personal experiences are more or less irrelevant. I've had bad experiences with Win2K myself, particularly with relation to random crashes in OpenGL apps. However, the overall concensus (on ZDNet, Slashdot, and in other media organizations) is that Win2K is quite stable, and this idea seems to bear out in reviews of the product. So both you and I are going to have to chalk up our bad Win2K experiences to random chance and the wrong hardware. BTW> I've had consistantly bad experiences with Suse. I don't go around saying Suse sucks, because its most likely just a problem particular to my config.
1) Confusing configuration. Get rid of/etc and replace it with something sane.
Like scattering configuration files across all the disks is sane?
>>>>>>>>>>.
Who said Windows did it the sane way? Take a look at Darwin and see how they do it. Now that's a good config structure.
2) Stupid directory structure. My dad...
I don't want to be rude, but I just could point it out, that, the directory structure, which has been around for 30 years, is not
stupid, your Dad....
>>>>>>>>>>>>
I use my dad as an average user. You're going to have to get over this elitist mentality that many computer-literate people seem to have. Its the same self-superiority trait that makes (many) New Yorkers, Washingtonions (I'm from the area BTW), artists, drama and movie people, musicians, and philosophers such snots. Anway, UNIX is perfect because it has been around for 30 years? The fact that UNIX has been around for 30 years shows that it is good, not perfect. The directory structure is asnine for a desktop user. There's just no getting around that. The fact that Microsoft, Apple, Be, and countless others all have cleaner, simpler structures can attest to the fact that random directory structure was one of the mistakes that *NIX made.
You know how to read a manual? A lot of people like you think a system is hard when it doesn't have a GUI.
>>>>>>>>>>
GUI? I'm programming my own OS and I'll be happy when I have a debugger! Anway, I can, and have read manuals. They are poorly written, full of crappy grammer, and confusing, but most of the time they work. However, there is no way in hell I'd expect a normal person to understand (or much less care) about them.
Say, configure a Apache from default installation to release to public with adaquate security only take me 5 minutes. You know how?
>>>>>>>>>>>>
How? I've never configured Apache. Although, I've configured BeOS's PoorMan server, and that takes 30 seconds. Besides, we're talking desktop here, remember?
I'm sure it helps if you use win2k more often. I'm a NT/2K admin, currently in hell.
>>>>>>>>
Don't even have 2K installed at the moment. (Or Linux for that matter.) I hosed my filesystem and have been using BeOS for the last three months. That's probably why I can't bear to use Linux. Not because Lin$ux $ux, but because I find it simply too messy for my taste, and BeOS does what I need to do, and faster.
how about no. How bout not enforcing any common structure, but have a standard. if a program requires some more complicated configuration files it can do that... sendmail.cf?
>>>>>>>>>>>
That's the same thinking that led to the current/etc problem;) The trick is to have a common structure, but make that structure flexible enough that all programs can adapt to it. Even then, there will be the occasional (very rare and low volume, certianly no desktop ones) that need more complex config files, but they can always keep a custom text file in their own directory. Sendmail is the kind of thing you set up once, on a sever that stays locked in the closet. If it can't adopt to the system structure, then nothing is really lost by keeping its configuration seperate. The problem is that when you *don't* enforce a standard structure, app developers just find it easier to store config files their own way, when they don't have to.
a) people, probably so (try to talk about environment to anyone in Scandinavia though). governments, nope. US government is trying to buy its quota of pollution from poorer countries, instead of complying with international treaties. Check out environmental policies in more environmentally advanced European countries (pretty much the entire EU).
>>>>>>>>>>>
We're not a dictatorship, the government can't just *force* everyone to comply. The people have to choose to do so, and like the people of India, China, and parts of Europe, they don't want to.
b)nothing a good 2-turn system could not solve
>>>>>>>>>>>
You heard it here folks. A crucial problem of the US democracy was solved by a one-liner from an AC that wasn't even capitalized and punctuated correctly!
While I do agree that Linux is the biggest threat to MS, that's not saying much. While on the server front, Linux has a fairly easy trip to the top, the desktop market will be a much harder road.
A) Linux still hasn't proven itself on the performance front. Even after several months, the Linux NVIDIA drivers are noticably lower in performance than their Windows counterparts. This happens even at low resolutions where the bitblit thing is not an issue. And with DirectX 8 and, later, Whistler promising huge increases in 3D performance, I don't know how Linux will deliver. Of course, MS could be overinflating their performance claims, but if NT4 and DirectX7 (which essentially made DX faster and more featureful than OpenGL) are any hint, MS *can* make good products, it just depends on if there is the motivation. When you add mediocre 3D performance to the slow desktops (in comparison to Win2K's, anyway) and high memory use (same as Win2K) then you have little reason to use Linux if you're looking for (desktop) performance.
B) Linux's stability won't play much of a part in this. Win2K is very stable, to the point where the average user (meaning one that shuts down at least once a week) won't be able to tell the difference.
C) Linux is still hard to use, and problematiclly, disunified. MS has been taking more and more steps to make configuration and control of Windows more "sane." Linux has been taking more steps in the other direction, especially with new distro like Mandrake that introduce prorietory config scripts. Adding hardware in Linux is not the simple (plug the hardware and the disk in) that it is with Win2K, and doing anything non trivial (meaning intermediate level use) is decidedly difficult. While the desktops are more or less easy enough to use, the system itself needs major retooling. Here are the problems I see
1) Confusing configuration. Get rid of/etc and replace it with something sane.
2) Stupid directory structure. My dad can handle installing programs, but with the mess of/usr,/home/_username_,/usr/bin,/usr/lib,/lib,/opt,/usr/local/*, ad nauseum, I doubt he could handle it. I barely can. (Of course I'm just terribly pampered by/boot/apps and/boot/home/config/lib)
3) Make the system cohesive. Get back to the UNIX roots. You know how UNIX treats *everything* as a file? You know how all UNIX console apps work together in a nice harmony of streams and pipes? You know how UNIX programs can be chained together to do complex work? You know how UNIX (well, BSD one's anyway) tools have more or less the same interface? Starting to get the idea? That means out go GnomeAPI and KDE-API and in comes Unified-Linux-Desktop-API (with GNOME and KDE serving as implementations) Out goes/etc, in comes a nice, organized config structure. Out goes LinuxConf, MandrakeConf, RedHatConf, DebianConf, ad-nauseam (yea, I'm making these up, but you get the idea;) and in comes UnifiedConf. Out goes modules.conf, samba.conf, XFree86Config, etc, and in comes a unified text file format. Out goes ipconfig and friends entirely. Use text files or use programs, but choose one for god sakes. Well, you get the idea.
Sadly, I enumerate all of these EVERY time an article like this comes out, and while everybody screams about how Linux has come so far, all I see are more propriatory configuration files, more stupid config programs, and a more fragmented set of APIs. Disgusting...
BTW> Don't get the idea that I don't like Linux. I really do. I just can't stand to use it.
It was a 1 year gap between the release of OSR2, and the release of the Precision Pro USB. The *functionality* difference was even less. Win95 OSR 2 *supports* USB. My speakers are USB and they worked fine under OSR2. I would have been perfectly happy paying for new features. (And I did when I bought NT.) It was an artificial trick to get me to upgrade.
Right. This is why I own several hundred dollars of MS stuff. (Mice, keyboards, joysticks, Windows NT, you name it.) I didn't pirate Windows 98 because I hate MS (I'm the biggest fan of NT4 and DirectX that you'll meet). I didn't pirate it because it had features that I needed (It didn't) I pirated it because I gave MS $90 for a product, and it ended up not working with an MS product that I'd already (indirectly through Dell) paid for. It wasn't because Win95 OSR2 didn't support USB (it does) it wasn't because there was a compatiblity issue (there wasn't) it was because MS tried to screw users over by playing an upgrade trick. I don't like being screwed over. It's not about free stuff not working (I haven't reinstalled NT since I hosed NTFS) its not about wanting more features than I'm willing to pay for, its about not being screwed over when you are an otherwise loyal MS customer.
It is a USB joystick... However, should I have to upgrade my entire ($90) OS just to use one joystick? My system came with a pair of USB controlled speakers. I didn't have to upgrade to Win98 to use them. And its not like a joystick is a terribly complex device that needs anything beyond basic OS USB support (which Win95 OSR2 has, btw.) Take a look at the Sidewinder USB driver for BeOS. It's OSS, and simple enough to understand that I was able to make it support my Presicion Pro after looking at the code for 15 minutes. It's *not* a terribly complex piece of software.
A) The people all over the world behave that way. The Americans don't have a monopoly on wasteful behavior. For example, people in China dump waste into the Yellow river. People in India do the same thing to the Ganges. People in developing contries waste natural gas through poor training (and laziness, it takes a while to light a gas stove.) If they could afford it, Europeans and Asians would waste just as much as Americans do. If you don't believe me, then go to Bankok, Thailand. You'd think you were in LA, sans the tight air-control standards. (BTW, I'm not talking out my ass, I've *lived* in these places)
B) Don't get *me* started on democracy. There is a reason for the electrol college which you stupid populists don't seem to get. (BTW, I'm a rabid democrat.) The electrol college assures that the majority does not totally outweight the minority. Do you really think that the politicians would give a damn about the jews in Florida if it weren't for the electrol college? How about blacks, arabs, and other minorites? Of course not! If it was just population, then they'd only be playing up to the white majority. The debate over majority rule was settled two hundred years ago during the first few presidencies. A general concensus was made (by politicians *far* more qualified than Hilary Clinton) that while the majority did in fact control the country, their power had to be kept in check to protect minority rights. The whole American democracy is based on checks and balances (and if you jaded Americans think that the US democracy is overrated and corrupted, try living in Asia or parts of Europe...) Well, welcome to one of these checks.
I wonder if enviro-natzi's will ever realize that while environmentalism is a noble goal (I'm a big environmentalist myself) it can only be accomplished through human oriented
measures. Like it or not, you can't shut the characteristics of humantiy out when trying to find a solution for a problem. Humans are human, and people will not regress to a lower
quality of life in order to do *anything* Bitch all you want about the declining social and environmental mores of society, nothing will get done unless you take the right
approach. The idea is to do your part, and get other people to do their parts. Keep progressing, don't force things on people that aren't ready for it, and eventually you'll realize
that the problem is gone. Same thing with feminatzism and ethnonatzism. Shutting out the basic flaws of people in order to promote a self-righteous and radical agenda simply
doesn't work. It would be nice if we could deal with these problems instantly, but we can't. It took a thousand years for democracy to firmly establish itself in the world (it still
isn't everywhere yet) and it took hundreds of years for slavery to die out. Humans progress, but they do it slowly. Just do your part and let the system work.
I pirate MS OSs. I have no qualms about it. They charge obscene prices for crappy products, and pull non-compatibility tricks to force you to upgrade. For example, when I bought my $90 MS joystick (say what you will, they make *good* joysticks) I had to get a copy of Win98 to use it. If they wouldn't pull stunts like this, they'd have much less piracy. Well, somebody will figure out a crack, so all hope isn't lost;)
True, not all Be developers agree that it's legal, but the general consensus seems to be that it is. In particular, the issue with OSS-ed software linking to system libraries being allowed makes sense.
RMS claims that it is illegal to load a GPL'ed driver from a closed kernel. The reason he states is that the process of linking to the kernel makes it illegal. HOWEVER, GPL'ed drivers are allowed to link against non-GPL system libraries. The kernel is a non-GPL'ed system library (more or less.) The kernel doesn't rely on the driver to work, but the driver relys on the kernel to wrok. If the logic saying that using GPL'ed drivers is illegal holds true, than the GIMP port to Windows is illegal (since they link to the closed source kernel, and closed source libraries) any GPL'ed software running on non-Open *NIXs is illegal, etc. Clearly, Windows GIMP is not illegal.
I don't think so. First, the hardware drivers should be entirely independant of the kernel. I'd really like to see drivers forked as independant projects with strong ties to the kernel developers. Second, you don't really point out *how* these developments are interrelated. As far as I can see, USB has nothing to do with the new VM. The news filesystems such as ReiserFS and the like have little to do with other kernel components other than the VFS (thought, technically, the VFS should require no changes for new filesystems...) agpgart is totally unrelated to the VM, as is DRI, etc. Starting to get the idea? While it may be true that changing one subsystem requires changes to other subsystems, it is not true that they require *major* changes at the same time. If the kernel was upgraded incrementally, then the resulting product would not only be easier to support (less stuff to break) but the transition would be easier. I remember that upgrading from 2.0 to 2.2 was a huge pain (I had to change distros) because so much stuff had changed. 2.4 will be a similarly big leap, and I can't help wondering how people will handle the 3.0 jump. Recently on BeDevTalk (the BeOS mailing list) there was some talk of what would happen if Be had to break binary compatibility (to move to GCC 3.0, if they ever do it.) There was also talk what would happen if changes to the InterfaceKit (the UI API) needed a break in binary compatiblity. Somebody (me) suggested lumping all of these breaks together. Of course I was (rightfully) chastised by a developer who pointed out that such major changes should be handled gradually, and it would be insane to make developers deal with both a new binary standard and a new API.
I'd beg to differ. If, instead of huge, long in coding major releases, the kernel had shorter, more succinct releases. then progress would be faster (less stuff to test for each release) and the product would be higher quality (less stuff that could break for each release.) When you write a major program, you don't code major subsystems at one time do you? Of course not. You code the thing incrementally. For example, if you wanted to change the VM, you change the VM and release a minor release. That's what.x releases are for. Major, structure changing releases (which 2.4 is) should increment the major version number (x.0) Of course, IANAKE (I am not a kernel engineer) so take this with a grain of salt;)
Slashdot Mirror
I don't think you quite understand how shared libraries work. The reason BeOS includes this stuff in the kernel is not to boost performance, but because library loading is closely tied to program loading (programatically they are quite similar) and by putting both in the kernel you can share the common code for both. The BeOS feature I was referring to requires some knowledge of shared libraries. When a shared library is loaded, the code has to be relocated to whatever memory addresses are free. This is a time consuming job, so the dynamic linker caches the results of this process in memory. However, when the library changes, these caches have to be updated. BeOS supports all the symantics of correctly updating libraries as they are changed by the system. None of this requires any user intervention (like running ldconfig) and is handled transparently by the system. Its mainly the difference between a process that should be automatic, but is implemented as a manual one in Linux. (Just like module loading, but lets not go there...)
;)
As for the GUI and browser, BeOS is a microkernel. It even implements networking in userspace. I think positively ugly that Linux implements stuff like audio in kernel space
Which circular dependencies are those? Oh, you're probably one of those people that can't manage to grasp the concept of installing two mutually-dependent packages at the same time solves this "problem". .01% of the population will install seperately.
>>>>>>>>>
Why does everyone assume that when I complain about a problem in Linux, I can't fix it? I've been using RPM quite a lot. I can figure out how to type rpm -i XFree86*.rpm in order to install all the circularly dependant packages. It is still a stupid idea. If two packages depend on each other, they should come as one package. Given the huge amount of bandwidth available these days, there is no excuse for KDE to come as a dozen different packages that only
Not a fan of modularity, or a maintainable package model, eh? Slack's fine for one or two machines, when you go beyond that, it's a nightmare to
maintain.
>>>>>>>>>>>
Slack's a pain to maintain? How? I don't know what experiences you have (I'm not a sysadmin and proud of it) but I see RPM's "modularity" as a stupid splitting of of stuff that really shouldn't. Modularity does not refer to splitting stuff up for the hell of it. There has to be a good reason for doing so. Its like if Linus seperated out the Linux VM into a seperate module. You can't use a different VM with a particular kernel, and in order to use the kernel, you have to load the VM. So what's the point of splitting the two up?
Consistent package management is key. Doesn't matter if you're talking about RedHat, Mandrake, Debian, SuSE, or even one of the "fringe" distributions like Connectiva or Immunix. All of those have good working package models. I'm sorry, but tar and gzip does NOT constitute a packaging system.
>>>>>>>>>>>>
RPM is not a packaging system. It is a POS created by people who obviously hate you. It requires a database (databases are evil!) to track packages, it doesn't mesh well with compiled software, it has stupid dependency management, and it takes too much control away from the user. That's why the packaging system I'm writing for BeOS will (hopefully) offer the power of a packager without all the crap in stuff like RPM.
It is nice, however, to see Pat Volkerding putting some *RECENT* versions of software in his distribution, at least for the most part (XFree86 is a notable exception). It wasn't all that long ago that Pat & co. shipped packages that were 3, 4 or even more versions old...
>>>>>>>>>>
How is anything in Slack 7.x NOT recent? They're glibc 2.2.3-based distros, and Slack 7.1 was the first distro to come with a KDE2 beta.
Is it just me, or was this ripped straight from the book Star Trek Generations? You know when Kirk goes through his whole adventuring phase and does the space diving?
To tell the truth, I have found Slack to be the easiest Linux distro I've ever used. It doesn't do a lot of crap behind your back, it installs stuff where you tell it, and it has the cleanest /etc/rc.d. It also doesn't make you deal with the POS RPM, and all the stupid circular dependencies that come with it. While KDE on RedHat is a mess of a dozen RPMs, the same thing on Slackware is kde2.tgz.
ldconfig was, and they laughed as I explained it... until I told them they'd be rebooting on an MS box at that step.
>>>>>>>>>>>>
Except not really. Every other OS in the world is smart enough to automatically register DLLs without running programs like ldconfig. Without rebooting even. In fact, I was reading on article on dynamic linkng a few days ago, and they talked about how BeOS goes to great lengths to make sure that the cached and on-drive copies of DLLs are the same, and can add DLLs on the fly and have the relocated images updated.
Yea, the Windows registry blows. They have the right idea, enforcing policy by creating a standard API to configuration data, but screwded with the simplistic key/value format and the brain-dead storage method. Something like the registry API (more flexible, though) that would store plain text files in a config heirarchy (or wherever the user wants) would probably be ideal. Then, write a GUI program that edits these standard text files, and away go all the configuration problems (or most anyway ;)
Because people have hard enough of a time getting past the "be" in my username, and even in topics where I *don't* mention BeOS, they think I'm comparing it to Linux. I'm really not in the mood to defend BeOS against a bunch of Linux weenies.
;)
PS> Not all Linux users are weenies, just the anti-BeOS ones
Make the system cohesive. Get back to the UNIX roots. You know how UNIX treats *everything* as a file?
From a programmer's point of view, I agree with you (check out Plan9 from Bell Labs sometime, you'll probably like it). But from a user's point of view, that is irrellavant.
>>>>>>>>>>
Not really. A clean, consistant API breeds clean, consistant software.
PS> As for the config files, who *said* they had to be scattered binary files? I just said they should be more structured. If there was one standard for config files, and one GUI tool that could handle that standard (preferably displaying embedded comments) then Linux configuration would be a cinch. As many people have pointed out, OSX does just this.
Actually, with the NVIDIA driver (they're the only ones that have even remotely close performance to their windows counterparts) the situation is reversed. XFree86 doesn't support page-flipping (with good reason according to Carmack) so as resolutions get higher, the memory bandwidth strain of having to blit the frame buffer starts adding up. However, even at low resolutions, performance is lower than it should be, so the page flipping issue is not entirely to blame. Either way, Linux should be *much faster* than the Win2K!
;)
Well from what I saw the difference beteen Windows vs. Linux was pronounced at low resolutions but the difference at high resolutions gets less and less as the resolutions get higher,
until it's hardly even noticeable. This was based on Quake3A benchmarks.
>>>>>>>>>>>>>>>>>>
That's even worse! That means that Linux does a fine job crunching the geometry code, but can't handle sending data to the graphics card quickly enough. That means the disply system is f*cked. (Though with XFree86, I wouldn't be surprised if this was the case
Considering that the file is XF86Config, and the command is ifconfig (ipconfig is for Windows NT 4.0), my guess is you don't have much experience in Linux
>>>>>>>>>>
Sorry, typo's. I've not only had a lot of experience with Linux, but I've compile XFree86 half a dozen times from source, used at least one version of every distro under the sun, and went through the painful libc5 -> glibc transition with good old Slackware 3.5
Get rid of /etc and replace it with something sane.
/etc, at the exact location that the HOWTOs and the documentation of the individual programs tell you. /etc can be pretty sane, and it's actually a lot more sane than the registry.
/usr/doc/packagename.
/usr/local/doc weren't there...
;) However, I have found the *NIX text "zen" to be one of the best things about it. Like I said, I don't hate Linux (and therefore UNIX). I just can't stand some of its more inane details.
After trying several distributions (among them SuSe, with their rc.config madness), I was so glad when I
finally found Debian, with all of the configuration files still in
>>>>>
Why does everyone think in terms of the registry? The registry is a piece of junk and always will be. Take a cue from MacOS X to see how things should be done. As for Debian, it sounds nice, but the fact that the layout matches the one in the HOWTOs does't make the layout any better.
You find the directory structure stupid. It's so complicated that your dad can't install programs. Well, tell him to look for what he needs on packages.debian.org and then have him type "apt-get install packagename".
Done.
>>>>>>>>>
True. And I really don't think that sequence would be a problem for anyone. I'm pretty sure the 20-something computer elite don't know it, but these older "illiterate" users DID use *NIX terminals back in tha day...
After that, the docs are in
>>>>>>>
That sounds great. If only
And if you want, "dpkg -L packagename" tells you
what was installed and where.
>>>>>>
That sounds stupid. I'm a hands-on type person. I like the previous example better.
Try that with Windows.
>>>>>
Don't have to. Everything is nicely organized in c:\program files. Though the doc thing is nice.
And if you want to compile everything yourself, have
a look at GNU stow, which helps you to keep the different programs separated so you can easily uninstall/
upgrade them without leaving files of the last few versions lying around. You don't need to know where to
put every single file, that's make install's job.
>>>>>>
Hiding complexity behind limited interface is the first mistake of UI building.
I can't see how one can tolerate the windows\system32
directory "structure" and be confused by a Unix directory tree.
>>>>>>>>>.
I can't tolerate the windows\system32 structure. That's why I use BeOS...
Oh, nice that you mentioned how all UNIX console apps work together in a nice harmony of streams and
pipes. This is the very reason I feel more comfortable with Unix when doing real work.
>>>>>>>>>
Or BeOS
In this thread, and in pretty much every Linux-related article in the press, there is far too much emphasis on how hard Linux is to install or configure. But there are actually people who install once and work with their machine for years.
And for my style of work, a few shell windows and Emacs are the best I have found so far. And yes, I did work
with DevStudio on Windows, and did not totally fall in love with it.
>>>>>>>
I've tried VI (and still use it whenever I don't feel like opening up BeIDE) but I have fallen in love with DevStudio. It is just so damn handy. (Well, as long as you stay away from the wizards and crap.)
If you can't stand to use Linux, you're not really required to like it. Perhaps it would make more sense just to like Linux' development model or the always helpful and well-informed slashdot crowd instead.
>>>>>>>>>>>>
I *do* like Linux. I just can't stand to use it. Some of its details are what you could call "deal-breakers" for me (and many other people, or I would presume so...)
A) Who said Win2K had a good directory structure? I like BeOS's, but Win9X's is nice too. There are only 3 directories on a clean install, "My Documents," "Program Files," and "Windows." The first is your's to do with as you please, the second is logicall organized by application, and the third is crappy, but you rarely need to go there anyway.
B) Linux takes up just as much memory on my machine as does Win2K. Of course, I'm (was) running GNOME, KDE 2, and the full host of libraries, but only because I need good app compatibility. Don't even get me started on how pointless it is for GNOME and KDE's libraries to be incompatible (and thus memory wasting.)
I know you are looking for flame, or you are actually a Linux supporter attracting flames in windows. Anyway, I fell into it.
/etc and replace it with something sane.
>>>>>>>>>
No flames, just honest critism. Like I said, I don't dislike Linux, I just have exceptions with some aspects of it.
If only NVIDIA would release its specifictions to linux community. After their hand-in-hand cooperation with M$ in XBox, I doubt NVIDIA would ever think of supporting Linux community in the future.
>>>>>>>>
I doubt it. History has shown that NVIDIA can drag MS around by the balls. Ever wonder why DirectX8 bears such a striking resemblance to NVIDIA hardware? There is nothing to indicate that NVIDIA will be the bitch in the MS-NV relationship.
Now that 3dfx is bought by NVIDIA....there goes voodoo*. Nevertheless, it's not Linux to be blamed, it's the attitude of those 3d chips vendors.
>>>>>>
Wrong on both counts. It is entirely the fault of 3DFx for making crappy products.
You can't deny the fact how quick Linux is be able to catch up with the performance without their
support.
>>>>>>>>
Which is why exactly 1 OSS driver is faster than the closed counterpart? (Some of Matrox's OSS drivers were faster because the closed versions had major OpenGL problems.)
B) Linux's stability won't play much of a part in this. Win2K is very stable
Win2k is stable? I bet you haven't been trying to upgrade the hardware in a Win2K box. All I got is
INACCESSIBLE_DEVICE_ERROR blue screen. Come on! I didn't even touch the disk! Just upgrade a damn CPU! It
happened in old NT, which resolved this problem long time ago, now they bring it back. Well done!
>>>>>>>>>>>>>>>
Then something is wrong with your machine. There is term in psychology (forget it offhand, its either a fallacy or the false concensus effect) that describes the overgeneralization of the characteristics of one example of a group to the whole group. This thinking is very dangerous because one's personal experiences are more or less irrelevant. I've had bad experiences with Win2K myself, particularly with relation to random crashes in OpenGL apps. However, the overall concensus (on ZDNet, Slashdot, and in other media organizations) is that Win2K is quite stable, and this idea seems to bear out in reviews of the product. So both you and I are going to have to chalk up our bad Win2K experiences to random chance and the wrong hardware. BTW> I've had consistantly bad experiences with Suse. I don't go around saying Suse sucks, because its most likely just a problem particular to my config.
1) Confusing configuration. Get rid of
Like scattering configuration files across all the disks is sane?
>>>>>>>>>>.
Who said Windows did it the sane way? Take a look at Darwin and see how they do it. Now that's a good config structure.
2) Stupid directory structure. My dad...
I don't want to be rude, but I just could point it out, that, the directory structure, which has been around for 30 years, is not
stupid, your Dad....
>>>>>>>>>>>>
I use my dad as an average user. You're going to have to get over this elitist mentality that many computer-literate people seem to have. Its the same self-superiority trait that makes (many) New Yorkers, Washingtonions (I'm from the area BTW), artists, drama and movie people, musicians, and philosophers such snots. Anway, UNIX is perfect because it has been around for 30 years? The fact that UNIX has been around for 30 years shows that it is good, not perfect. The directory structure is asnine for a desktop user. There's just no getting around that. The fact that Microsoft, Apple, Be, and countless others all have cleaner, simpler structures can attest to the fact that random directory structure was one of the mistakes that *NIX made.
You know how to read a manual? A lot of people like you think a system is hard when it doesn't have a GUI.
>>>>>>>>>>
GUI? I'm programming my own OS and I'll be happy when I have a debugger! Anway, I can, and have read manuals. They are poorly written, full of crappy grammer, and confusing, but most of the time they work. However, there is no way in hell I'd expect a normal person to understand (or much less care) about them.
Say, configure a Apache from default installation to release to public with adaquate security only take me 5 minutes. You know how?
>>>>>>>>>>>>
How? I've never configured Apache. Although, I've configured BeOS's PoorMan server, and that takes 30 seconds. Besides, we're talking desktop here, remember?
I'm sure it helps if you use win2k more often. I'm a NT/2K admin, currently in hell.
>>>>>>>>
Don't even have 2K installed at the moment. (Or Linux for that matter.) I hosed my filesystem and have been using BeOS for the last three months. That's probably why I can't bear to use Linux. Not because Lin$ux $ux, but because I find it simply too messy for my taste, and BeOS does what I need to do, and faster.
how about no. How bout not enforcing any common structure, but have a standard. if a program requires some more complicated configuration files it can do that... sendmail.cf? /etc problem ;) The trick is to have a common structure, but make that structure flexible enough that all programs can adapt to it. Even then, there will be the occasional (very rare and low volume, certianly no desktop ones) that need more complex config files, but they can always keep a custom text file in their own directory. Sendmail is the kind of thing you set up once, on a sever that stays locked in the closet. If it can't adopt to the system structure, then nothing is really lost by keeping its configuration seperate. The problem is that when you *don't* enforce a standard structure, app developers just find it easier to store config files their own way, when they don't have to.
>>>>>>>>>>>
That's the same thinking that led to the current
a) people, probably so (try to talk about environment to anyone in Scandinavia though). governments, nope. US government is trying to buy its quota of pollution from poorer countries, instead of complying with international treaties. Check out environmental policies in more environmentally advanced European countries (pretty much the entire EU).
>>>>>>>>>>>
We're not a dictatorship, the government can't just *force* everyone to comply. The people have to choose to do so, and like the people of India, China, and parts of Europe, they don't want to.
b)nothing a good 2-turn system could not solve
>>>>>>>>>>>
You heard it here folks. A crucial problem of the US democracy was solved by a one-liner from an AC that wasn't even capitalized and punctuated correctly!
While I do agree that Linux is the biggest threat to MS, that's not saying much. While on the server front, Linux has a fairly easy trip to the top, the desktop market will be a much harder road.
/etc and replace it with something sane.
/usr, /home/_username_, /usr/bin, /usr/lib, /lib, /opt, /usr/local/*, ad nauseum, I doubt he could handle it. I barely can. (Of course I'm just terribly pampered by /boot/apps and /boot/home/config/lib)
/etc, in comes a nice, organized config structure. Out goes LinuxConf, MandrakeConf, RedHatConf, DebianConf, ad-nauseam (yea, I'm making these up, but you get the idea ;) and in comes UnifiedConf. Out goes modules.conf, samba.conf, XFree86Config, etc, and in comes a unified text file format. Out goes ipconfig and friends entirely. Use text files or use programs, but choose one for god sakes. Well, you get the idea.
A) Linux still hasn't proven itself on the performance front. Even after several months, the Linux NVIDIA drivers are noticably lower in performance than their Windows counterparts. This happens even at low resolutions where the bitblit thing is not an issue. And with DirectX 8 and, later, Whistler promising huge increases in 3D performance, I don't know how Linux will deliver. Of course, MS could be overinflating their performance claims, but if NT4 and DirectX7 (which essentially made DX faster and more featureful than OpenGL) are any hint, MS *can* make good products, it just depends on if there is the motivation. When you add mediocre 3D performance to the slow desktops (in comparison to Win2K's, anyway) and high memory use (same as Win2K) then you have little reason to use Linux if you're looking for (desktop) performance.
B) Linux's stability won't play much of a part in this. Win2K is very stable, to the point where the average user (meaning one that shuts down at least once a week) won't be able to tell the difference.
C) Linux is still hard to use, and problematiclly, disunified. MS has been taking more and more steps to make configuration and control of Windows more "sane." Linux has been taking more steps in the other direction, especially with new distro like Mandrake that introduce prorietory config scripts. Adding hardware in Linux is not the simple (plug the hardware and the disk in) that it is with Win2K, and doing anything non trivial (meaning intermediate level use) is decidedly difficult. While the desktops are more or less easy enough to use, the system itself needs major retooling. Here are the problems I see
1) Confusing configuration. Get rid of
2) Stupid directory structure. My dad can handle installing programs, but with the mess of
3) Make the system cohesive. Get back to the UNIX roots. You know how UNIX treats *everything* as a file? You know how all UNIX console apps work together in a nice harmony of streams and pipes? You know how UNIX programs can be chained together to do complex work? You know how UNIX (well, BSD one's anyway) tools have more or less the same interface? Starting to get the idea? That means out go GnomeAPI and KDE-API and in comes Unified-Linux-Desktop-API (with GNOME and KDE serving as implementations) Out goes
Sadly, I enumerate all of these EVERY time an article like this comes out, and while everybody screams about how Linux has come so far, all I see are more propriatory configuration files, more stupid config programs, and a more fragmented set of APIs. Disgusting...
BTW> Don't get the idea that I don't like Linux. I really do. I just can't stand to use it.
It was a 1 year gap between the release of OSR2, and the release of the Precision Pro USB. The *functionality* difference was even less. Win95 OSR 2 *supports* USB. My speakers are USB and they worked fine under OSR2. I would have been perfectly happy paying for new features. (And I did when I bought NT.) It was an artificial trick to get me to upgrade.
Right. This is why I own several hundred dollars of MS stuff. (Mice, keyboards, joysticks, Windows NT, you name it.) I didn't pirate Windows 98 because I hate MS (I'm the biggest fan of NT4 and DirectX that you'll meet). I didn't pirate it because it had features that I needed (It didn't) I pirated it because I gave MS $90 for a product, and it ended up not working with an MS product that I'd already (indirectly through Dell) paid for. It wasn't because Win95 OSR2 didn't support USB (it does) it wasn't because there was a compatiblity issue (there wasn't) it was because MS tried to screw users over by playing an upgrade trick. I don't like being screwed over. It's not about free stuff not working (I haven't reinstalled NT since I hosed NTFS) its not about wanting more features than I'm willing to pay for, its about not being screwed over when you are an otherwise loyal MS customer.
It is a USB joystick... However, should I have to upgrade my entire ($90) OS just to use one joystick? My system came with a pair of USB controlled speakers. I didn't have to upgrade to Win98 to use them. And its not like a joystick is a terribly complex device that needs anything beyond basic OS USB support (which Win95 OSR2 has, btw.) Take a look at the Sidewinder USB driver for BeOS. It's OSS, and simple enough to understand that I was able to make it support my Presicion Pro after looking at the code for 15 minutes. It's *not* a terribly complex piece of software.
A) The people all over the world behave that way. The Americans don't have a monopoly on wasteful behavior. For example, people in China dump waste into the Yellow river. People in India do the same thing to the Ganges. People in developing contries waste natural gas through poor training (and laziness, it takes a while to light a gas stove.) If they could afford it, Europeans and Asians would waste just as much as Americans do. If you don't believe me, then go to Bankok, Thailand. You'd think you were in LA, sans the tight air-control standards. (BTW, I'm not talking out my ass, I've *lived* in these places)
B) Don't get *me* started on democracy. There is a reason for the electrol college which you stupid populists don't seem to get. (BTW, I'm a rabid democrat.) The electrol college assures that the majority does not totally outweight the minority. Do you really think that the politicians would give a damn about the jews in Florida if it weren't for the electrol college? How about blacks, arabs, and other minorites? Of course not! If it was just population, then they'd only be playing up to the white majority. The debate over majority rule was settled two hundred years ago during the first few presidencies. A general concensus was made (by politicians *far* more qualified than Hilary Clinton) that while the majority did in fact control the country, their power had to be kept in check to protect minority rights. The whole American democracy is based on checks and balances (and if you jaded Americans think that the US democracy is overrated and corrupted, try living in Asia or parts of Europe...) Well, welcome to one of these checks.
I wonder if enviro-natzi's will ever realize that while environmentalism is a noble goal (I'm a big environmentalist myself) it can only be accomplished through human oriented
measures. Like it or not, you can't shut the characteristics of humantiy out when trying to find a solution for a problem. Humans are human, and people will not regress to a lower
quality of life in order to do *anything* Bitch all you want about the declining social and environmental mores of society, nothing will get done unless you take the right
approach. The idea is to do your part, and get other people to do their parts. Keep progressing, don't force things on people that aren't ready for it, and eventually you'll realize
that the problem is gone. Same thing with feminatzism and ethnonatzism. Shutting out the basic flaws of people in order to promote a self-righteous and radical agenda simply
doesn't work. It would be nice if we could deal with these problems instantly, but we can't. It took a thousand years for democracy to firmly establish itself in the world (it still
isn't everywhere yet) and it took hundreds of years for slavery to die out. Humans progress, but they do it slowly. Just do your part and let the system work.
I pirate MS OSs. I have no qualms about it. They charge obscene prices for crappy products, and pull non-compatibility tricks to force you to upgrade. For example, when I bought my $90 MS joystick (say what you will, they make *good* joysticks) I had to get a copy of Win98 to use it. If they wouldn't pull stunts like this, they'd have much less piracy. Well, somebody will figure out a crack, so all hope isn't lost ;)
True, not all Be developers agree that it's legal, but the general consensus seems to be that it is. In particular, the issue with OSS-ed software linking to system libraries being allowed makes sense.
RMS claims that it is illegal to load a GPL'ed driver from a closed kernel. The reason he states is that the process of linking to the kernel makes it illegal. HOWEVER, GPL'ed drivers are allowed to link against non-GPL system libraries. The kernel is a non-GPL'ed system library (more or less.) The kernel doesn't rely on the driver to work, but the driver relys on the kernel to wrok. If the logic saying that using GPL'ed drivers is illegal holds true, than the GIMP port to Windows is illegal (since they link to the closed source kernel, and closed source libraries) any GPL'ed software running on non-Open *NIXs is illegal, etc. Clearly, Windows GIMP is not illegal.
I don't think so. First, the hardware drivers should be entirely independant of the kernel. I'd really like to see drivers forked as independant projects with strong ties to the kernel developers. Second, you don't really point out *how* these developments are interrelated. As far as I can see, USB has nothing to do with the new VM. The news filesystems such as ReiserFS and the like have little to do with other kernel components other than the VFS (thought, technically, the VFS should require no changes for new filesystems...) agpgart is totally unrelated to the VM, as is DRI, etc. Starting to get the idea? While it may be true that changing one subsystem requires changes to other subsystems, it is not true that they require *major* changes at the same time. If the kernel was upgraded incrementally, then the resulting product would not only be easier to support (less stuff to break) but the transition would be easier. I remember that upgrading from 2.0 to 2.2 was a huge pain (I had to change distros) because so much stuff had changed. 2.4 will be a similarly big leap, and I can't help wondering how people will handle the 3.0 jump. Recently on BeDevTalk (the BeOS mailing list) there was some talk of what would happen if Be had to break binary compatibility (to move to GCC 3.0, if they ever do it.) There was also talk what would happen if changes to the InterfaceKit (the UI API) needed a break in binary compatiblity. Somebody (me) suggested lumping all of these breaks together. Of course I was (rightfully) chastised by a developer who pointed out that such major changes should be handled gradually, and it would be insane to make developers deal with both a new binary standard and a new API.
I'd beg to differ. If, instead of huge, long in coding major releases, the kernel had shorter, more succinct releases. then progress would be faster (less stuff to test for each release) and the product would be higher quality (less stuff that could break for each release.) When you write a major program, you don't code major subsystems at one time do you? Of course not. You code the thing incrementally. For example, if you wanted to change the VM, you change the VM and release a minor release. That's what .x releases are for. Major, structure changing releases (which 2.4 is) should increment the major version number (x.0) Of course, IANAKE (I am not a kernel engineer) so take this with a grain of salt ;)