Despite the relentless march of Linux, major vendors believe commercial Unix releases aren’t ready for the scrapheap yet. As Linux evolves from its single-processor roots into larger-scale applications, many market watchers have predicted that it will eventually replace the remaining commercial Unixes: Hewlett-Packard’s HP-UX, Sun Microsystem’s Solaris, IBM’s AIX and SGI’s Irix. “It costs between $150 million and $200 million a year to generate your own Unix system and you then have to say to yourself ‘Am I going to see that amount of extra revenue if I put in these features?'” Linux International executive office Jon ‘maddog’ Hall told iTnews. However, Unix vendors argue that the needs of large-scale enterprise users haven’t yet been met by Linux.
Linux not ready to kill Unix — yet
About The Author
Eugenia Loli
Ex-programmer, ex-editor in chief at OSNews.com, now a visual artist/filmmaker.
Follow me on Twitter @EugeniaLoli
A good son never kills a bad dad. Yet this is a good dad…:-)
Linux is pretty sweet, but its not AIX, IRIX, Tru64, or Solaris just yet. Its much more versatile that those OSs, and better on moderate configurations, but when you’ve got serious hardware with 16 CPUs and a multi-terabyte drive array, one of the more mature UNIXs is probably a better bet.
That said, Linux will get there sooner rather than later. SGI has already decided to go with Linux for their new large-way Itanium machines. Since there will be no IRIX Itanium port, and development of MIPS has stopped, Linux is SGI’s future. IBM has also made the statement that Linux would ultimately replace AIX on its machines. A lot of the work in 2.7 will be on scalability on NUMA machines, so 2.8 should be much better suited to huge machines.
Linux is absolutly ready to replace propietary UNIX.
Possibly you don’t know the applications proprietary Unix is used on. Linux is ready to take over some applications, but not all yet, Rayiner said it well above.
“Linux is absolutly ready to replace propietary UNIX”
You’re joking right? It’s got a few more years left before it can compete meaningfully in all areas.
Somehow many of my friends and also myself were really troubled, when SGI introduced x86 Hardware to run on WinNT. We all thought it would be much better to introduce an IRIX for x86 PCs and maybe build high end graphic cards for these beasts. Indeed the way SGI took was a mistake.
IRIX even the antique and now unsupported version 4 (I am still using 4.0.5E and 5.3 at home) is better suited for workstation than the actual Linux distributions. Well, the actual Linux distributions also have graphical administration tools. But, those supplied with IRIX can be used by users (CAD designers, graphic designers, etc.). The ones supplied with actual Linux distributions definitely not.
I developed software on a damn lot diffent UNIXoids. I somehow arrange with all of them. But IRIX (like NextStep and OS X) are superior in usability.
Just my opinion,
Carsten
http://www.debian.org/ports/netbsd/
http://www.debian.org/ports/freebsd/
so maybe;) old man + son = new OS in the future
;)we dont need to kill each other
“SGI has already decided to go with Linux for their new large-way Itanium machines. Since there will be no IRIX Itanium port, and development of MIPS has stopped, Linux is SGI’s future.”
Has SGI announced EOL of MIPS? Last time i checked, this was not true. Do you have a source?
Linux /is/ ready to replace many important areas formerly Unix’s domain, but its certainly not yet ready to totally replace it. But we are getting there. 2.6 is a milestone release, 2.8/3.0 will be even better. Unix isn’t ready to die just yet.
I’ve not found too many BSD folks outside of those Debian monstrosities (the projects you mention, not the folks behind them!) that are actually thrilled by the idea. Not that I or anyone I’ve talked to is outright against it, but the most common response to Debian BSD is “why?”
GNU utils already work just fine for those BSD users that want them, and the BSD utils are better integrated into the various projects. It seems like the whole idea behind Debian BSD is just to get a GNU userland to run on it, and it’s been done…
I assume the time frame here to be 10-15 years at best.
Unix has a long and successful life. Why would someone so eagerly change something that just works? As hardware is ridicilously cheap in comparison to all the surrounding cost proving issues, why would someone wanna step from a working Unix installation that might have worked for the better part of the decade to something else which also just might work?
For most companies, IT is something that just has to work, that doesn’t necessarily mean it has to be the latest of everything (What is the whole idea behind Reactos???). Why would someone then change to Linux even if it gets on par with your unix solution? It has to be a LOT better, and that would take a substantial amount of years…
Besides, nobody knows what the future holds… Linux has grown tremendously in short time, but seriously… many would want to have heard the word reliability for 5 years before even considering changing…
Yes IBM wants everyone to change, because how would they else be able to sell consulting? Something that just works (Unix) doesn’t allow them to sell that much time. When companies start figuring this out, IBM and this “Move to Linux” thingie will have problems.
Linux indeed is ready to wipe out Unicses. We here have SMP Sparc boxes with fair amount of ram and RAID-disks. Linux 2.6 indeed is ready to work on such large iron. Much faster in high workloads as web/database server than Solaris 8. Haven’t yet compared to Solaris 9 but 2.6 is indeed very, very fast with “big” iron. I recommend everyone to test 2.6 kernel who have access to such machines. You might be very suprised (Compared to 2.4).
While the 2.4 kernel wasn’t particularly scalable beyond 4-8 CPUs, the 2.6 kernel is being run in 128 and more CPUs beast at IBM. Scalability has been increased in those machines by a significant margin. Measurements in some areas have improved by 200-300% or more. While it’s pretty sure Linux can’t run as smootly as ej: Solaris in those beasts (Solaris has been doing that for *years*), Linux has improved in only one revision more than any other SO. That’s one of the good things of Linux: The approach to solve those problems is *radical*: killl the old slow code and write new scalable code. Even if it breaks things. This is the “linux” style development. The comercial OSes have pretty much a BSD style development: “do it carefully, don’t break things”.
There’re no OSes out there which can evolve so fast as Linux does. Just look at Microsoft: They only have 2003 server betas for the AMD64, no betas for XP still. Why? Probably because they’re too busy trying to release the XP SP2. Which, BTW, should have been released in the 2003, and it has changes that were developed in linux very quickly (we have *3* different patches to avoid “buffer overflows”, and they don’t have even bored to merge them in mainstream). There’re already several patches flying around for the kernel, too. For example, there’s a “CFQ” scheduler patch from Jens Axboe. This patch allows “io priorities” (somewhat like the SGI GRIO) so you can say “this process only can have 10% of disk bandwith”. This is already working, and I bet it will be merged in the 2.6 lifetime. Guess in what LoooongOS is expecting Microsoft to get this (to make video streams more smooth)?
There’re no OSes out there which can evolve so fast as Linux does. Just look at Microsoft: They only have 2003 server betas for the AMD64, no betas for XP still. Why?
As so many peolpe often refer too, Linux is only a kernel. It evolves just as fast as many other OSes do. You see, there’s really no point in having 1000 devs working on a kernel… the team should be optimised and fewer to make it managable. Linux therefor evolves just as fast as many other OSes do.
Don’t get too high thoughts about this OSS model…
How stupid is this? Why does everyone assume that one OS must die for another to succeed? This is just dumb. Choices exist to fill requirements, and OSs don’t just succeed because they meet some basic need, they must surpass any predecessors in what they can offer to get ahead. Getting ahead doesn’t necessarily mean the death of a competitor. UNIX still has a very large development and installation base. Why use simplistic terms like “kill” when it’s just a continual race with a current winner.
“the team should be optimised and fewer to make it managable.”
XBe just doesn’t get it. Linus explicitly supports an evolutionary model for software. Proprietary software and the BSD’s support a structured managed model for software. But in practice all software develops by evolution (it is the third replicator: genes, memes, software).
The structured management model gets in the way and slows down the evolutionary process once the project gets to complex. Hence the Linux kernel evolves faster.
Don’t get too high thoughts about this closed managed model…
“the team should be optimised and fewer to make it managable.”
XBe just doesn’t get it. Linus explicitly supports an evolutionary model for software. Proprietary software and the BSD’s support a structured managed model for software. But in practice all software develops by evolution (it is the third replicator: genes, memes, software).
The structured management model gets in the way and slows down the evolutionary process once the project gets to complex. Hence the Linux kernel evolves faster.
Care to back up that assertion with a little thing people like to call “facts”, you know, actual examples of projects becoming so complex the whole thing self implodes, even with good management and communication.
Maybe you should actually go into the big bad world of programming and see how commercial companies operate, how their development practices are structured then come back claiming that the bizare model is better than the cathedral model used by BSD and proprietary software vendors.
Specifically, what features are in UNIX, that will not be in linux soon?
Don’t get me wrong, there may be such features, but I don’t remember the article mentioning them specifically.
The structured management model gets in the way and slows down the evolutionary process once the project gets to complex. Hence the Linux kernel evolves faster.
It may slow it down in terms of feature richness, but certainly not in coding mistakes, hence why Linux has suffered from more patches and more security holes than XP during 2k3.
Care to back up that assertion with a little thing people like to call “facts”, you know, actual examples of projects becoming so complex the whole thing self implodes, even with good management and communication.
Yes, Mozilla Seamonkey is a perfect example of this. Even the developers themselves complain about how the source is a big mess and therefor they choose to switch to a different model and smaller projects (Moz FB/TB/SB).
I didn’t say it doesn’t work, just that it’s very time consuming and has backsides to it. Therefor saying Linux will kill Unix is quite stupid. People working on Solaris kernel do a very well job, but Solaris suffer from lack of drivers, which I admit is one of those things where 1000 can work… why? Because every driver is a project of its own, the kernel isn’t.
Besides, leaving the kernel and talking about OSS apps is quite irrelevant as most of them work on all Unices and have active ports which therefor doesn’t support Linux, just the unixlike operating systems…
So what’s the Linux killing Unix point???
Linux will not kill Unix. Both will survive and prosper. Unix has been around a long time and it will always have a niche. For the visuallisation tools I use, nothing beats Irix.
“The structured management model gets in the way and slows down the evolutionary process once the project gets to complex. Hence the Linux kernel evolves faster.”
It may slow it down in terms of feature richness, but certainly not in coding mistakes, hence why Linux has suffered from more patches and more security holes than XP during 2k3.
Also, one has to point out the fact that non-sexy parts are neglected by coders in favour of developing the sexy parts or what ever the flavour of the month is.
Windows 2000 only problem was stupid defaults, however, some of the excriment that floated around regarding it were completely uncalled for. People moaning that IIS was “installed by default”, well sunshine, don’t bloody install it. If you’re automating the installation then simply specify in the configuration file that you don’t want IIS installed.
As for the break ins, how many servers out there are still not running service pack 4? go into any large organisation and look at the laziness of administrators in action.
Sure, burn Microsoft when you are hacked and you’ve done everything to the letter but don’t moan when your server is hosed because you’re rather play pong on your vintage Atari 2600 instead of admining the server your paid to do.
“Care to back up that assertion with a little thing people like to call “facts”, you know, actual examples of projects becoming so complex the whole thing self implodes, even with good management and communication.
Yes, Mozilla Seamonkey is a perfect example of this. Even the developers themselves complain about how the source is a big mess and therefor they choose to switch to a different model and smaller projects (Moz FB/TB/SB).
It had nothing to do with size and everything to do with poor management. There were never any fixed requirements and it moved from being a suite to “the stew that grew”. People adding bits and pieces left, right and centre until we see the ball of mess we have today.
There was never, “1.0 must have these features, no more, no less” and simply stuck with that agenda. Instead, however, we have programmers pushing their own agendas jumping from one market hype to another according to the weather outside.
I didn’t say it doesn’t work, just that it’s very time consuming and has backsides to it. Therefor saying Linux will kill Unix is quite stupid. People working on Solaris kernel do a very well job, but Solaris suffer from lack of drivers, which I admit is one of those things where 1000 can work… why? Because every driver is a project of its own, the kernel isn’t.
And yet, there are developers out there who are porting FreeBSD drivers to Solaris for use. There is nothing stopping anyone from creating opensource drivers for Solaris just as nothing ever stopped anyone from creating opensource drivers for BeOS when it was available.
As for the SUN Xserver, with the latest 4.3.0.1 kit from tools.de/solaris, the performance isn’t bad. The only gripe is the crappy GNOME, however, hopefully Solaris 10 will have JDS available making Solaris useful as a workstation.
Is now everybody jumping on the “Linux is not ready for <you put your favourite thing here>” ??? I believe that Linux is ready for many things, and that there is atleast one Linux solution out there for every type of problem. But to come out with the “Linux is not ready for…” every 2 days is becomming lame day after day…
Unix is not strong in the desktop market, therefore this is aother “is Linux ready for the server market?” question. I doubt we would be seeing such intense FUD from the media, microsoft, and SCO if it was not.
Indeed the latest stats show an incredible growth in the server market. At who expense? From what I see, it is at the expense of both.
For example, in my previous company, we ditched the expensive Sun boxes in favor of cheap RedHat boxes because they just perform better as Oracle database servers. In my current company we are replacing Solaris and AIX boxes with SuSE intel boxes because of the awesome linux performance.
Microsoft believes that it is ready to take on the Unix world by simply offering a Unix-like environment (based mostly on the work of OpenBSD). I.e. you can run some Unix shell scripts and some unix commands in Windows. Why is that considered better than a Linux replacement of Unix?
For everyday business solutions, Linux is a much better solution that either Microsoft or Unix. Better performance and cheaper cost.
Sorry about my poor editing skillz…
Indeed the latest stats show an incredible growth in the server market. At who expense? From what I see, it is at the expense of both.
Indeed the latest stats show an incredible growth in the server market. At who expense? From what I see, it is at the expense of both Microsoft and Unix.
For example, in my previous company, we ditched the expensive Sun boxes in favor of cheap RedHat boxes because they just perform better as Oracle database servers. In my current company we are replacing Solaris and AIX boxes with SuSE intel boxes because of the awesome linux performance.
Why are you replacing Solaris and AIX with SuSE intel boxes when you could run Solaris on x86 servers provided by SUN or better yet, why did you company insist on running Oracle, which is consistantly slow on any platform it runs on? wouldn’t it have been better running Sybase? Why didn’t you consider Sybase?
Windows pretty much kills off the hope of a Unix desktop
Linux will finally kill off Unix once and for all
Average consumers eraning a WalMart salary couldn’t afford a Mac yet wouldn’t like to use an awkard linux desktop.
In 10 years, linux will try to catch up to longhorn
In *nix, it is all about disagreement.
What’s so special about Longhorn?
RDBMS as “filesystem”?
OS/400 uses DB2 as “filesystem”.
Vector graphic lib as base of the GUI?
IRIX uses GL.
NextStep uses DPS.
OS X uses Display PDF.
Carsten
Won’t it just be a matter of economics eventually?
Some factors to consider:
– The lower end X86 machines are getting more powerful and becoming cheaper (relative to processing).
– Storage prices are also dropping.
– GNU/Linux is improving at rapid rate, faster maybe then the “Unices”? The interfaces and utilities are becoming more “polished”.
– Major corporations (IBM and others) seem to be embracing Linux.
– There are enterprise strength support options available for Linux (Red Hat etc.).
– The term “Linux” is beginning to appear in the public consciousness (IBM “Linux” commercials etc- in the US at least).
– Linux may offer greater flexibility in terms of choices for both hardware and software.
– Clustering of low end machines is becoming an interesting alternative (ala Google) to individual higher end machines.
To me it looks like things are starting to equalize. If that is true and the Unix companies don’t figure some way to compete effectively in the next few years they could be in a world of hurt.
While I don’t think it means the end of Unix as an OS technology it could mean the end of it as a viable commercial model.
I think a lot of the disagreement on this issue stems from how we define ‘linux’. Linux is really at it’s core just a kernel, and by this definition I believe it’s 100% feature complete to compete in virtually all market spaces.
But there’s really quite a bit more to linux in the sense that people use the term. When we say linux, we generally are referring to not only the kernel, but also the applications built around it. And THIS is where linux is many years behind traditional unices.
The simplest example that I can come up with is some of the volume management tools available in solaris, which allow you very fine control of hundreds of hard drives, and where your indexes are stored, etc. It’s applications like these that aren’t there for linux yet I think, tho the kernel is quite ready for them I believe.
I’m sure there are other examples that the more knowledgable osnews crew could come up with.
Won’t it just be a matter of economics eventually?
Let me remind you of a term that hasn’t occured here for quite some time called TCO!
Hardware is a cost, agreed… Solaris/BSD/Linux all runs on AMDs 64bit CPU’s… so no difference there really…
Software… Solaris cost some money, BSD/Linux both = free. (Even though most would go with some contract to e.g. redhat which cost money) so solaris cost slightly more here… not a big deal, it outperforms Linux for now so it evens up on HW costs..
So why would IBM embrace Linux here? One thing that pops to my mind is the mere fact that adapting environment to Linux and making it work, keeping it up2date etc takes a lot more time and consultancy than with BSDs or Solaris… which is the very reason IBM stays alive.
so in a TCO perspective, why would IBM otherwise embrace Linux? Revenge against MS? Hardly think so… it’s simply because they earn more money that way.
So if Cost is an issue, expect Unix to survive for a very long time… and expect someone to burst the IBM bubble quite soon…
Why are you replacing Solaris and AIX with SuSE intel boxes when you could run Solaris on x86 servers provided by SUN or better yet, why did you company insist on running Oracle, which is consistantly slow on any platform it runs on? wouldn’t it have been better running Sybase? Why didn’t you consider Sybase?
Oracle is non-negotiable, since that is under the control of our DBA department. They can do SQL server and DB2, but they do not permit it. I would imagine that that was an infrastructure decision from the past.
Hosting hundreds of user connections on Oracle for Linux, using Java Stored Procedures on a Pentium box with 512MB RAM is actually quite fast. My test harness pushes the number of connections past one thousand, which makes Oracle work hard, but not enough to make it fall over. I haven’t seen this performance using Windows or Solaris (on a Sun box).
I have found that the same hardware running SQL server will thrash with 3 users running simple queries 😉
Well it also gives them an OS that runs across their entire server range, though no doubt z/os (os/390) is a better option to run on your mainframe.
Another thing to think of is that if IBM were in the future to migrate over to an “enhanced” version of linux (one containing all the features that people have quoted in their comments – both kernel and userland) they would save money from not having to do as much OS development, though that is debatable after all they do hire some of the Linux devs (rusty russel for one)
“That’s one of the good things of Linux: The approach to solve those problems is *radical*: killl the old slow code and write new scalable code. Even if it breaks things. This is the “linux” style development. The comercial OSes have pretty much a BSD style development: “do it carefully, don’t break things”. ”
Linux can currently take this approach because it’s use is very simplistic. It needs to run Apache, BIND, and Sendmail. When companies start deploying desktop solutions, databases, enterprise webservers, crm solutions, ERP’s, etc. on it, development is going to grow to a crawl. At that time, Linux would either have to worry about compatibility or lose everyone to a less “technical” superior, more “business” friendly OS.
<em>Specifically, what features are in UNIX, that will not be in linux soon?</em>
Speaking only for Solaris, there are quite a few features that will not be in Linux anytime soon:
– Debuggability. Linux’s debugging technology is still in the dark ages of operating systems — the operating system doesn’t even take a crash dump on panic, for crying out loud! This despite the LKCD patch being done many years ago by SGI — Linus refuses to buy back the patch on “idealogical” grounds. To those of us for whom kernel implementation is our livelihood, this is as if an aircraft manufacturer refused to have cockpit voice recorders or flight data recorders for ideological reasons. That is to say, it sounds completely crazy.
– Observability. Linux has atrocious observability, basically relying on old UNIXisms like top, strace and lsof as its observability story. Compare this to the army of p-tools on Solaris (pcred, pfiles, pflags, pldd, pmap, psig, pstack, ptime, ptree, pwait, prstat, preap, etc.) And then take a look at DTrace, available in Solaris Express and coming in Solaris 10:
http://www.sun.com/bigadmin/content/dtrace
DTrace is a quantum leap beyond everything else out there, including DProbes and LTT — two pieces of a nascent tracing infrastructure for Linux that Linus refused to buy back for (you guessed it) ideological reasons.
– Resource management. When machines get even slightly large, managing resources becomes very important. Solaris is already <em>way</em> ahead of Linux in this department (processor sets, pools, projects, tasks, extended accounting, etc.). And Solaris is about to make the jump to hyperspace with Zones in Solaris 10.
– Real-time. Linus believes — again, as an ideology — that a general purpose operating system cannot be a real-time operating system. QNX, IRIX and Solaris (among others) have proved this to be incorrect. In Solaris, one can put a process in the real-time scheduling class on a dedicated CPU — and get determinism normally associated with hard real-time systems (e.g. < 10 usecs jitter). This is an area where Linux doesn’t even try to compete (RTAI and RTLinux are fine for tiny systems, but don’t do well for SMP real-time). Note that the SGI Australia spokesperson referred to real-time in the article as an area where SGI will <em>always</em> position IRIX instead of Linux.
There’s quite a bit more areas, but this gives at least a flavor. We don’t see Linux catching up anytime soon — or even really understanding how far behind they are. For more details on the features coming in Solaris 10, see:
http://www.aceshardware.com/read_news.jsp?id=75000449
For each of these features, look at what Linux has to offer. In some cases, Linux has a competitive offering. But in many, Linux has nothing — nothing in 2.6, and nothing in the works.
Excuse the <em>s in the above — sed ‘/em>/s//i>/g’…
“I have found that the same hardware running SQL server will thrash with 3 users running simple queries ;-)”
Oracle is a little more forgiving when it comes to poor database design. Like where i am working now the database server running Oracle 9i 64-bit(that i am not in control of) on an older SUN box, a 6 way 5000 gets slammed becasue of the poorly design database that has little to no normalazation of data and a horrid hardware setup(Mainly not enough memory and a stupid disk layout) You should see some of the queries they run its disgusting. Thats what happens when you let developers try and handle the “Systems stuff” as they put it. Saying that 3 users on a MS SQL server maxes it out tells me alot about those that implemented the setup.
You forgot to mention that Solaris has had hotplug CPU and memory for ages, whereas we’re just seeing patches now in that area in Linux.
> Linux’s debugging technology is still in the dark ages of operating systems — the operating system doesn’t even take a crash dump on panic, for crying out loud!
The distros include kgdb or LKCD. Moreover, kgdb is already in the -mm tree. May be included in mainline.
Often, the oops trace that appears in the system log is enough for debugging. Some times you *really* need a dump though – especially when the box hangs hard.
Besides, if you do get a kernel panic you can post the backtrace to LKML and some developer might reply in real time with a patch. Happens very often. (Oh, you may wanna be a little cautious if it’s a production box of course
I’m not sure how easy it is for little ol’d me to get the attention of a Solaris kernel engineer in San Quentin or wherever without a Sun Gold Spectrum support contract
> pmap, psig, pstack, ptime, ptree, pwait, prstat
pmap, pstack, pstree and top can be found in any distro. And time is equiv to Solaris ptime. Some distros don’t include sar, iostat and mpstat though.
> – Resource management.
There’s CFQ for disk IO, the network stack has bandwidth allocation policies. I did see some patches allowing a general resource allocation framework a while ago, can’t recall exactly when… But it doesn’t compare to Solaris, by far.
Most of your points are spot on.
BTW I’d really like to see Solaris on my x86 desktop
Often, the oops trace that appears in the system log is enough for debugging. Some times you *really* need a dump though – especially when the box hangs hard.
Bullshit. The only bugs debuggable from the panic message are easy ones. Hard bugs need a dump to be able to debug them. Trust me: I’ve debugged hundreds of psychotic kernel bugs, and none of them was debuggable from the panic message. If Linux wants to become an operating system for grown-ups, it will discover what we discover what we discovered in Solaris a long time ago: you get to the point very quickly where the only bugs left are psychotic ones. And these bugs are debilitating if not fixed. And as proof of my assertion that oops messages aren’t enough: look at the amateur circus around debugging of the problems cropping up due to kernel preemption in the last moments before 2.6.0.
Besides, if you do get a kernel panic you can post the backtrace to LKML and some developer might reply in real time with a patch. Happens very often. (Oh, you may wanna be a little cautious if it’s a production box of course
Bullshit, again. Try this: go to groups.google.com and search for “oops” on linux-kernel. Now sort by date. Now look at how many of those posts have no response.
I’m not sure how easy it is for little ol’d me to get the attention of a Solaris kernel engineer in San Quentin or wherever without a Sun Gold Spectrum support contract
Well, you’ve got my attention for starters. And I’m not the only engineer in Solaris Kernel Development who reads OSNews — by a long shot.
Most of your points are spot on.
Phew.
BTW I’d really like to see Solaris on my x86 desktop
It is possible. I’m posting this from my Solaris x86 laptop, for whatever that’s worth…
> Bullshit. The only bugs debuggable from the panic message are easy ones. Hard bugs need a dump to be able to debug them. Trust me: I’ve debugged hundreds of psychotic kernel bugs, and none of them was debuggable from the panic message. If Linux wants to become an operating system for grown-ups, it will discover what we discover what we discovered in Solaris a long time ago: you get to the point very quickly where the only bugs left are psychotic ones.
Agreed. You’ll note that I said *often*, but not always.
> Bullshit, again. Try this: go to groups.google.com and search for “oops” on linux-kernel. Now sort by date. Now look at how many of those posts have no response.
But what of the oopsen that *do* get fixed?
> Well, you’ve got my attention for starters. And I’m not the only engineer in Solaris Kernel Development who reads OSNews — by a long shot.
OK, I have this backtrace of a recursive_mutex_enter panic from mdb on a SunFire 15K…
I’m kidding
> It is possible. I’m posting this from my Solaris x86 laptop, for whatever that’s worth…
I’ll give Solaris 10 a spin when it comes out. I have the early access beta, but never got around to installing it.
Specifically, what features are in UNIX, that will not be in linux soon?
Stability (not the “woohoo, I haven’t rebooted for 700 days kind, the “my three year old binary only application still runs on the latest release” kind), consistency, documentation, predictability, support.
Linux is a reasonable solution if you want to run basic and simple WWW, DNS, DNS, Samba, etc services (although IMHO FreeBSD is a better choice). It’s also done well at lowering the cost of entry to the *nix server market for business with small budgets that want basic Oracle (or other “mainstream” app) installs, or turnkey solutions (eg: SuSe OpenExchange) that require relatively low sysadmin overhead.
However, it is not in a position to replace Solaris on a Sunfire 15k.
Nice to see you around…
I think it’s great to see some of those “in the know” answer some of these maniac articles.
I hope you stay around and keep pushing input about how Linux frankly is years behind and killing some of the hype and keeping it to facts…
You have got to be the most informative person that’s posted on this site in ages. Good stuff! I can’t wait to try Solaris 10…
OK, I have this backtrace of a recursive_mutex_enter panic from mdb on a SunFire 15K…
I’m kidding
It got my heart rate up, nonetheless…
For whatever it’s worth, if you were serious I’d arrange to get the panic message and crash dump from you. And ironically enough, recursive mutex enters are one variant of bug that actually is generally debuggable from the panic message…
Linux has more bugs, unstable, hardware not up to snuff.
UNIX has serious hardware, vendors, support. But all of it cannot compete with MS Datacenter Server.
> It got my heart rate up, nonetheless…
😀
Well, right now the servers are ticking along just fine, so I won’t give you any trouble.
Solaris was my first love. But I will throw my weight behind Linux for a very simple reason – I can *participate*. If I hit a bug I can send in a patch (after I acquire some more brain cells, of course). I can (and do) compile my own kernel. You know what that feels like don’t you?
DaveM said it pretty nicely:
http://www.ussg.iu.edu/hypermail/linux/kernel/0309.0/1060.html
> Linux is absolutly ready to replace propietary UNIX
I’d say that never, even from Microsoft, I’ve seen such an urge to do software genocide.
What’s the goal? To drive every other OS to extinction? How come the other OSes, who are lying quietly in their corner, are seen as enemies, with the manifest example of the BSDs?
It would seem it’s not only MS who ‘can’t stand one bit of competition’.
Every OS out there has a purpose. To do this and that, for that and those users, in these or those situations, in this or that hardware. Except for 2 OSes, the W and the L, which, *especially the latter* seem to wish to be everywhere, for everyone, doing everything, and making sure no one else does the job.
(In a thread named ‘L not ready to kill U — yet’ this is hardly OT.)
DaveM said it pretty nicely:
That remark from Miller borders on the slanderous. Solaris does not go out the door “buggy with all kinds of SMP deadlocks.” I have no idea where he’s getting his data for this statement, but Miller has a long history of ad hominem polemics against Solaris — so it’s likely that his “data” is just wishful thinking. Solaris goes through a long and rigorous process before it ever ends up even in the hands of internal users, let alone external ones. Don’t take my word for it; download the latest Solaris 10 here:
http://wwws.sun.com/software/solaris/solaris-express/get.html
And see how many “SMP deadlocks” you can find.
To me it seems as if Linux worst enemy is slowly becoming its zealots. They preace about how Linux is better than everything else when they do not fully know their competion. Linux is better than MS, better than BSD, better than UNIX blah, blah, blah. They should learn to respect their favourite OS elders. Without UNIX, BSD and to a lesser degree MS, Linux probably would not exist. Hell, Linux is just a functional copy of these OS’s anway, nothing new. To all those that want to flame me about not knowing what I talking about let me tell you this, I’ve used Linux before it went v1.0, and BSD and Solaris for about the last 4 years. I’m not an expert in the OS field, this is merely my observations.
“To me it seems as if Linux worst enemy is slowly becoming its zealots”
That is really incorrect. The Linux zealots have ALWAYS been it’s worst enemy.
Hey sun.com, this gave me quite a laugh, but should be a shot in the arm for you guys:
http://uptime.netcraft.com/up/graph/?host=www.pseg.com
Check out 2003, and look at the hosting history. I’d love to hear what was going on behind the IT scenes at that company. Looks like they finally gave up and “got real”.
lol… still cracks me up.
J
Gaaah, I was referring only to this part:
> It’s not because we have some weird genius trolls writing the code, it’s because of our insanely huge testing base.
Sorry I was unclear.
I think the best part of getting rid of Linux is the fact that IBM is OFFSHORING 40,000 JOBS BY 2005.
Sorry, but this includes Linux to.
http://www.forbes.com/business/newswire/2003/12/30/rtr1194285.html
I am a little late come back as I have been busy doing other things. But to be honest I think that the responses by the worshippers at the altar of the Cathedral is pretty pathetic. I never suggested that projects implode if they get too complex, but I am asked to provide an example – RTFP.
As for the “bizare” model of software development I guess that is another reference to esr who is of course a looney gun-toting right-wing “libertarian” who’s views I would not necessarily agree with. The software development model I am referring to is Linus’s comparison to biological evolution.
Way back in the seventies ( Oh! I remember the days with the RT11 OS running on a PDP11 with an Evans Sutherland graphics system)when I was a grad student, one of the areas I was studying was protein evolution. The concepts we studied then now seem to be pretty much applicable to software development. Dawkins extension of Darwinian evolution into the area of social evolution, with the concept of memes, was a remarkable insight. Linus further extention of the idea of Darwinian evolution to software development is a similarily important insight.
Remember DNA like software can also just be regarded as a string of 0’s and 1’s. But then I guess you guys might just be creationists or something.
BTW back on topic I too think Linux has a long way to go before it replaces proprietary Unix at the top end. But then look at AIX 5L all GNU userland is there, all of the OSS desktops and utlities are there all you have to do is plug in a Linux kernel. Sometime maybe 5 too 10 years in the future IBM will replace AIX by Linux but not just yet.
BTW back on topic I too think Linux has a long way to go before it replaces proprietary Unix at the top end.
You see there is a little problem here…. because Linux WON’T kill Unix nor replace it on the top end. It might start to compete on the top end but never replace it…. well not within the next 50 years anyway….
People such as I would never even consider switching to Linux. I’d use any system which suits my needs and follows a model of development which I feel comfortable with (evolutionwise as well as socially) without starting another war here (this IMO and cannot be changed by zealotry).
I bet there is a whooole bunch of people like me, therefor Linux simply can’t kill Unix, just become a competitor.
Note!!!
I’m all pro Unix, Open Source, Solaris, Windows, *BSD, OSX, BeOS, QNX, diversity etc… just that there’s something about Linux which I dislike very much… go figure what!