login
Search
Hello gentlemen and ladies. As a Linux user for many years now (regulars user, not a programmer), I want to congratulated you all for the great work you all have done in making Linux widely supported and compatible with a lot of hardware. Recently, I was on a search to see how the Linux kernel itself compares to other Unix kernels (*BSD, Solaris, AIX, etc) in terms of *real* innovation. After reading various articles on the net about technology used in Linux and the other Unixes, especially after reading the Solaris Vs Linux articles written by Dr. Nikolai Bezroukov - http://www.softpanorama.org/Articles/solaris_vs_linux.shtml and http://www.softpanorama.org/People/Torvalds/index.shtml , I came to the conclusion (and correct me if I'm wrong on that) that Linux is not innovative at all when compared to the real Unixes in terms of technology. From what I understand from the articles is that Linux rips off a lot of technologies originally invented by other Unixes but it does very little original innovation on its own. How come? Isn't *real* innovation important anymore in Linux? Or did Linux became a commercial "fast buck bitch" for various corporations like IBM, Intel, Red Hat, etc and *real* innovation has stalled? A lot of stuff is ported to Linux, but all of this stuff isn't Linux' own innovation rather existing technology from other companies/Unixes. Solaris invented ZFS, dtrace, RPC, PAM, NFS, RBAC, etc, FreeBSD invented jails (lightweight in-kernel virtual machines), IBM/AIX invented volume manager... just to name a few. Linux' record in innovation looks extremely unconvincing for such a mature stage of development (over 10 years). What has Linux invented on its own? Ext and Ext2 were a rip off from the Unix UFS/FFS, in the early years Linux didn't even had its own TCP/IP stack, the recently announced BTRFS is a rip off of ZFS, the Linux kernel tracing tool is a joke compared to dtrace in Solaris and is hardly a Linux *real* innovation, ...
heh, I'm not a troll, I just wanted to know what the Linux people think about it and your perspectives on the issues, but it seems you all go hiding instead of explaining and making a clear stand -
On Sat, 23 Jun 2007 17:22:26 +0200 You sound like one - or very misinformed. Most of the Solaris and AIX "innovations" you mentioned are far older, other things are bogus (eg the LSB spec for Linux is based upon SuSv3 - the single unix spec), and the unix nme is payware not free for use. A few innovations that afaik first appeared the Linux kernel - Making multiple hosts appear transparently as one IP address - Futex fast hybrid locking - Single pass checksum fragment and send fragments in reverse order - Reiserfs - very innovative design, but innovation isn't neccessarily success - JFFS/JFFS2 - flash wear levelled file system avoiding all the problem patents - Loadable modules for a non-microkernel I'd argue the lack of a stable kernel internal API is also an innovation A bigger question to ask is "When is innovation good ?" The reason everyone uses ext3 or on BSD UFS/FFS is the same reason we use the paperclip today - its an extremely reliable, well understood solution to the problem space. Is every office that uses paperclips inferior - or smart ? There are also lots of big innovations in Linux donated by other organisations - from Sun NFS (The real NFS innovation was that Sun gave the spec out and let people implement it for free) through to stuff like RCU, stuff made freely available elsewhere and implemented in Linux, and tons of stuff where Linux is the one that combined them in clever and useful ways. The basis of building great free software projects is sharing and mixing, not sitting in a lab inventing something cool from scratch. Linux could have innovated its own system call interface from scratch. If so I doubt it would have caught on. Now if you want really innovative OS work go look in the lab or at projects most people have never heard of and don't run. Alan -
- ALSA framework and drivers - Direct Rendering Infrastructure The userland API _is_ stable; a stable intra-kernel API would *hinder* Generally, OS kernels have adopted and improved each others' ideas since the term was coined. Simply pulling out the Linux kernel and stating it has re-implemented more features than it innovated itself simply isn't fair. The same holds true for _any_ of the others! BTW, PAM and NIS are userland. Certainly you don't want to compare even an average Linux distro with a plain solaris, AIX or *BSD* installation? Also keep in mind that the Linux kernel is highly portable (handheld to mainframe), maybe only matched by NetBSD. This requires a major amount of maintenance care and some extra work for each new feature. And BSDs are not Unix, strictly speaking; Unix has "ripped off" BSD, as you would say. You have simply fallen for some highly biased articles, if not propaganda. Torsten -
hmm, wasn't loadable kernel modules first implemented in SunOS 4.x together with the proc system ? -
Yes, but that was pretty cumbersome. You had to resolve the symbols in user space, using a hopefully matching /vmunix. Linux was first to feature an in-kernel linker and symbol table, IIRC. Torsten -
Err, uh, no- I believe that Solaris development for this at the very least predates even 0.59 linux- I think it was Joe Provino at Sun ECD near Boston who gave us a working prototype in early 1989. -
building upon or improving existing technology is as important as inventing new things. if every one insisted on dreaming up new things, i doubt we would've accomplished anything significant (not just in OS, -jb -- Tact is the art of making a point without making an enemy. -
Let's also not forget that many of the "innovative" features that Grodzan says Linux has copied to Solaris and other Unixes, were actually not invented by them. OS/2 already had dtrace in 1994 (it even had the same name), and many of the traditional Unix features were copied^Wheavily inspired in multics. -
On Sat, 23 Jun 2007 18:19:43 +0200 DRI is based on SGI work and Mark Kilgard and the SGI folks definitely Proc type stuff is a lot older than Linux or Unix AFAIK. Loadable modules ditto but the full load/unload/autoload stuff I've not seen pre-Linux. -
Representation of process state and control of that state via files on a filesystem? AFAIK, it's 80s stuff and at least one of the sources had been research branch in Bell Labs - whether you call it Unix or not... Do you have any references for that animal in earlier systems? A lot older than Unix would mean 50s or 60s... -
On Sat, 23 Jun 2007 22:02:29 +0100 I don't know about proc in that sense prior to v8 unix I was thinking about the logical device stuff and filesystem objects/namepaces that produced program generated data. -
There's a lot in Linux that was true innnovation: Alan Cox's Networking Architecture. VFS Architecture (best one out there -- even better than M$'s) Scheduler Design. Jeff -
Thanks Jeff, so from reading all the responses here I can conclude that Linux innovates stuff by itself and not only gets it from other places. Is it also right to say that other kernels, be it BSD, Solaris, maybe AIX?, also benefit from the Linux innovations? eg adding stuff from the Linux kernel into their own kernels if their licenses allow it -
Absolutely. Every operating system benefits from the cross pollination of ideas that happens on mailing lists, through white papers and at conferences. -- Politics is the struggle between those who want to make their country the best in the world, and those who believe it already is. Each group calls the other unpatriotic. -
Hey, I heard of one. I got a few friends that are sitting in an IRC channel and have been working on a complete new OS from scratch for like 10 years now (kernel, filesystem, graphics drivers, libraries - everything). I consider them to be totally nuts of course. When I ask them why are you still doing this? Can't you use linux? Then the answer is that there are still companies interested in operating systems like that, precisely because they are not well- known. It would be pretty hard to exploit vulnerabilities in such a system (or that is their explanation anyway). -- Carlo Wood <carlo@alinoe.com> -
The real innotation in Linux is that it is open source and yet popular enough that there are versions that even a windoze user could easily pick up. David Kane -
Can you name such companies so that I'll never accidentally buy some of
their stocks? ;-)
There are already more than enough operating systems available that are
less popular than Linux...
E.g. a good combination of less popular than Linux and a very good
cu
Adrian
--
"Is there not promise of rain?" Ling Tan asked suddenly out
of the darkness. There had been need of rain for many days.
"Only a promise," Lao Er said.
Pearl S. Buck - Dragon Seed
-
Alan Cox writes: [...] > > A few innovations that afaik first appeared the Linux kernel > - Making multiple hosts appear transparently as one IP address > - Futex fast hybrid locking DEC Firefly workstation, before 1987. Nikita. -
Absolutely. We had almost 900+ not-so-productive mails on another thread recently ... then what is this? Provocation is _standard_ troll tactics. Why don't you try being innovative yourself? -
Because I've seen many times how people outside the kernel community get ignored or even labled as trolls when asking something, so I thought that provocation in this case could be better productive for me. -
On Sat, 2007-06-23 at 14:17 +0200, Grozdan Nikolov wrote: Thanks! Bernd -- Firmix Software GmbH http://www.firmix.at/ mobil: +43 664 4416156 fax: +43 1 7890849-55 Embedded Linux Development and Services -
Perhaps you should change your rude attitude towards people who are seeking for answers without actually looking for rants or flame-wars. If you have read my replies to Alan, you should know why I asked these questions.... To clarify something that might be incorrect or biased in the articles I've read so far... if you could tell me a better place to ask about Linux internal stuff, please tell me so...... As I'm not a kernel programmer I don't see the need to subscribe to the LKML, I can contribute nothing to it. Yes, I do follow the LKML by reading it (that's how I discovered the new CPU schedulers from Ingo and Con and gave them a try, great piece of software, by the way). Reading kernel "patch e-mails" doesn't really teach you who invented this stuff... there are probably a lot of technologies which I'm not aware of their inventors, hence the simple questions I asked to clarify it for myself...... but if you decide that it's trolling because I'm not part of your "kernel development team" and I don't contribute to it (maybe I don't have the skills?) then you are the one who keeps the biased or wrong articles out there live longer by not willing to answer or clarify some things to a person who's just looking for the *correct* answers Thanks !!! -
well, i would say this - put yourself into the shoes of a kernel developer who barely has time to keep track of the large volume of development work, discussions, testing, etc. Then someone who claims to be not a kernel developer, who isn't subscribed to the list comes along and says 'there is _no_ innovation in the linux kernel'. What would your reaction be? I'm not a kernel developer myself, but i think there are lots of resources on the internet where you can read watered down versions of Of course, everyone wants to learn from the gurus. But confronting them in this way hardly seems the right way ;) -jb -- Tact is the art of making a point without making an enemy. -
My reaction will be to clarify it to this person that this is not true (and thanks to the some of you who already did), even if I'm under pressure from development work/testing/patching... But this is just the type of person I am. Everyone is different so I expected some "rude" reactions. But there are people who are willing to clarify things (Alan Cox, Jeffrey Merkey, Diego Calleja on the clarification of dtrace being used on OS/2, etc). Many thanks to those... If some of you get annoyed (which is perfectly possible), then -
A few places: The LinuxChanges page at kernelnewbies: http://kernelnewbies.org/LinuxChanges The kernel section of LWN: http://lwn.net/Kernel/ Kerneltrap: http://kerneltrap.org/ Kernel Traffic (unfortunately no longer updated): http://kerneltraffic.org/ And then you have list archives like : http://lkml.org/ http://marc.info/?l=linux-kernel http://www.uwsg.iu.edu/hypermail/linux/kernel/index.html -- Jesper Juhl <jesper.juhl@gmail.com> Don't top-post http://www.catb.org/~esr/jargon/html/T/top-post.html Plain text mails only, please http://www.expita.com/nomime.html -
I fail to see what part of that reply was rude in any way. It simply offered a perfectly valid suggestion. Just because people give you something other than what you want, does not make them rude. It is also quite likely the reply was written before reading the other comments. With the volume on lkml, reading all comments in a thread before writing any replies is just not possible. -- Len Sorensen -
I'm for that (including a place for GPL discussions), but I think that people would still just overload lkml instead of using the split lists. --- ~Randy (resent due to failure on first attempt) -
Then turn it around: the technical part becomes a separate list. Sort of like netfilter and netfilter-devel. Jan -- -
would be quite difficult in practice, since people would then argue that their mails were in fact technical ;) -jb -- Tact is the art of making a point without making an enemy. -
It certainly has an innovative licence - which is why it is attracting developers and replacing most of those other unices . . . Helge Hafting -