OSNews

"I tried Solaris 10. It is good only as a server OS, with still very limited hardware support and everything needs to be done from the CL. What is the point?"

Sounds familiar 3=)

It is good only as a server OS, with still very limited hardware support

Linux runs on every cheap device you cant think of.

In the realm od Solaris, you first get the OS, then look in the HCL, to pick the right hardware. Or, while buying a machine, you already buy an appriopriate machine for solaris.

It's jst in not a OS, to put on an 15-year old 80486 stanting under your deas and make it route your packets.


and everything needs to be done from the CL.

Do you suggest, that it's a drawback for a server?

"Do you suggest, that it's a drawback for a server?"

In a sense...
Why give up the comfort of the tools you get if you use RHEL or SLES (OK, $$$$)

Solaris is probably (with the exception of mainframe, OS/400 os's etc) the best server operating system on the market.

Personally I'd trust my systems (and do!) to Solaris on Sparc.

Linux is fine on the desktop and as a general purpose workstation O/S and non-mission critical serving.

...and everything needs to be done from the CL.

It would be incorrect to state that "everything" needs to be done from the commandline.

1) webmin is included with Solaris 10 so you can do quite a few administration tasks from a browser

2) Sun Management Console is included and is a graphical desktop application for doing many system management tasks.

3) updatemanager is included and is a graphical desktop application for installing / managing system updates.

So, your assertion that "everything" has to be done from the command line is grossly incorrect.

1) webmin is included with Solaris 10 so you can do quite a few administration tasks from a browser

Webmin is a horror at times. It is good if you know what you are admining works with Webmin and if you have some jr. admins that you can't train immediately. Otherwise, I avoid it unless I've checked it completely.

Once Sun releases their code under an acceptable license such as the GPL I will look into their Open Solaris. Until then I stick with Linux.

You mean, untill they release Solaris 10 under a license that's compatible with the GPL.

The CDDL and GPL are almost carbon-copies of one another.

> The CDDL and GPL are almost carbon-copies of one another.

http://www.gnu.org/licenses/license-list.html#GPLIncompatibleLicens...

"This is a free software license which is not a strong copyleft; it has some complex restrictions that make it incompatible with the GNU GPL. That is, a module covered by the GPL and a module covered by the CDDL cannot legally be linked together. We urge you not to use the CDDL for this reason.

Also unfortunate in the CDDL is its use of the term 'intellectual property'."

Not to say that it is wrong to be in-compatible with GPL. But just to state the fact that it is in-compatible and some people do take in that fact when deciding what OS or software to use.

Not to say that it is wrong to be in-compatible with GPL. But just to state the fact that it is in-compatible and some people do take in that fact when deciding what OS or software to use.

I guess these same "people" must also refuse to use anything under a Mozilla or Apache Licence because these are alos incompatible.
So they will refuse to use Firefox and Apache. Wonder what they have left to use?

Yes, because which Open Source license Solaris is released under really affects the users of the software.

I suppose you don't use Mozilla either?

Honestly, grow up.

If you were planning on helping develop the software, then I could see the concern. But as a user, it really does not affect you.

Mod this man up! He has a point.

Choosing software based on philosophies/licenses, as opposed to which is the best tool for the job, is moronic.

I never understood why people who have no principles need to make fun of people who do. You don't have to agree with them, but calling them moronic is, well, moronic. And no, using the best tool for the job is not a principle.

I STILL don't understand why people who have no principles need to make fun of people who do.

Pretending politics don't affect the future of your decision is naive.

Right, that's why everyone is using M$ windows, because it's t he best tool for the job: for those who don't install squate, who just want to write some papers and use the internet, guess what??? oh yeah, linux is way better, don't know about other systems since I haven't used them. The reason I switch to linux is because of it's license, and it's its licence that will keep it alive: the gpl allows a natural growing of software, sometimes doing double work, but that's no problem, there's no such thing as one solution.
Besides, there is nothing moronic about choosing software based on philosophies/licences. After using such software, you might never go back, cause it might fulfill the needs as well. There is nothing more moronic than using overkill, how many people use more than 1% of the feautures ms word can? (and they even pay for lots of stuff they don't NEED).

Yes, yes it does. Try reading, it'll do ya some good.

If you can't understand why a good license is important....

Yes, yes it does. Try reading, it'll do ya some good.

If you can't understand why a good license is important....

OK, what is so bad about the CDDL?

You need to get real.

Neither Apache nor Firefox are GPL compatible. Do you refuse to use these too because they are "unnaceptable".

CDDL is based on the mozilla license anyway, don't see anyone rejecting Firefox because of it's license do you?

Firefox is lisenced under MPL/LGPL/GPL. So your theory sucks.

You are just silly, have you even read the license? Why can't you agree to it more than the GPL? I have more trouble agreeing to GPL than the BSD license, i still use GPL-products. GPL isn't the finest there is and never has been.

"Linux killer"? Oh please. The guy's network card didn't work without compiling a driver and he had to manually set everything up.

I distinctly remember having to do this very thing on linux maybe 4-5 years ago. Linux was touted as a windows killer then!

Also, linux killer means in the x86 server space where the number of no name nics are fairly small and contained. No one in thier right mind is claiming Solaris will kill linux on the desktop. There are other OSes already doing that, like Windows and MacOS X.

Also, linux killer means in the x86 server space where the number of no name nics are fairly small and contained. No one in thier right mind is claiming Solaris will kill linux on the desktop. There are other OSes already doing that, like Windows and MacOS X.

Linux came from 0% 10 years ago to 2 or 3% on the desktop, and is still growing, and somehow being killed at the same time?

Growing slowly != Losing market share/being killed.

It's funny isn't it? I think Solaris will have a bit of an app shortage though...

I rather liked how they talked about Wine crashing Linux in the article as if that related over to Solaris? Did they try Wine on Solaris? I don't see that as being very fair.

The virtual kernel stuff looked very cool. Sounds quite a bit like user-mode linux though, I didn't see that mentioned. I understand that user-mode linux eats a lot more resources.

It also seems a bit off to compare Solaris to kernel 2.4. I mean, if they were using a 2 year old version of Solaris that'd be fine; but seriously 2.6 was released about 18 months ago! Everyone but Slack is using it! Even RedHat switched from their barely 2.4 kernel.. But of course, Windows sucks because of 9x you know. And Macintosh is awful because of pre-OS X problems (like the lack of memory protection).

Most of the numbers that I have seen that are not blatently doctored have shown KDE at about 60% of the total OSS desktop market

In any event, I don't think this is a KDE/Gnome thing.

Solaris's problem is hardware support, it is horriable, I have 4 systems, I could only get it to install on one of them and on that system my network did not work properly without some major troubleshooting.

Solaris cannot kill linux simply because you cannot kill and idea

Nope, but, of course, when articles refer to Linux and the Market, they're not talking "Linux" in the big view bubbling petri dish that Linux is, that will never go away. They're talking RedHat and Novell.

So, while Solaris may have little impact on Linux as a social meme, Sun and Solaris/OpenSolaris CAN have a great impact on RedHat, which is where Linux meets with Money.

Damage RedHat, and you damage Linux, even immortal Linux.

I do not know about the iPod and power management, but nVidia has a nice 3D accelerated driver for FreeBSD.

Nvidia 3d acceleration works, and i can synch my iPod just fine...

Linux has failed on the desktop because of the myriad of distributions. And since Solaris has a much better license than the GPL there's no problems with binary drivers.

You do realize licensing is not the main problem with binary drivers, right?

Each distro using a different GCC cannot have the same binary drivers. If Intel releases a driver compiled with GCC 3.2, and you are running a GCC 3.4 compiled kernel, the module is going to have problems functioning correctly. This is why the Nvidia driver installer compiles the kernel module right on the desktop, and only the OpenGL libraries are closed source.

Also, what in the CDDL is stopping different distros of Solaris from popping up, using different versions of GCC, and causing the same problems? Ever heard of Schillix?

Once Open Solaris has different distros, which is bound to happen, and can be compiled fully using GCC, the same problems will no doubt pop up.

(If you don't believe me about the GCC problem, take a look at the Nvidia driver read me, it discusses the problems I speak of).

Oh, and Nvidia/ATI have binary drivers. I haven't noticed any lawsuits, have you?

With the HURD, however, the FSF would sue companies left right and center for binary drivers since they own the copyrights, which is one reason HURD will never be big on the desktop.

You do realize licensing is not the main problem with binary drivers, right?

Wrong. If you take the time to read half the discussions about licensing issues and binary drivers on the Kernel mailing lists you'd know that statement is incorrect. Many of the interfaces that used to be available to all kernel driver developers have been made "GPL only" in recent years.

Each distro using a different GCC cannot have the same binary drivers. If Intel releases a driver compiled with GCC 3.2, and you are running a GCC 3.4 compiled kernel, the module is going to have problems functioning correctly. This is why the Nvidia driver installer compiles the kernel module right on the desktop, and only the OpenGL libraries are closed source.

Also wrong. The main reason there are problems is because Linux doesn't have a stable driver API. Different GCC versions rarely if ever have an effect on the C binary ABI, only on the C++ ABI. So, yet again what you say is incorrect.

Additionally, since the Sun Studio compilers are available now and provide a much more stable ABI. In fact, the ABI for Solaris is so stable that drivers compiled and built for Solaris 9 work years later without change on Solaris 10!

Once Open Solaris has different distros, which is bound to happen, and can be compiled fully using GCC, the same problems will no doubt pop up.

Only if they diverge from the Solaris ABI.

(If you don't believe me about the GCC problem, take a look at the Nvidia driver read me, it discusses the problems I speak of).

You're right, I don't believe you. You don't sound like someone that has actually helped develop drivers for multiple operating systems. I have.

"With the HURD, however, the FSF would sue companies left right and center for binary drivers since they own the copyrights, which is one reason HURD will never be big on the desktop."

I'm not sure what binary drivers has to do with copyright ownership. The GPL states you can do whatever you want, and the driver development model is not fully established. It's not illegal to make a binary driver. It's what you do with it that can be a problem.

You do realise that Solaris isn't built with gcc, surely.

Solaris is built with the SunStudio compiler.
Sun have this thing called backwards-compatability, which is unheard of in the GNU/Linux world. Not just API, but ABI also.

A Solaris 2.6 binary runs fine on Solaris 10, and vice-versa. Regardless of the compiler, OS, libc version. That's the difference with Solaris.
Not the only difference, I'll grant you - Linux could mimic IPMP without too much difficulty, and the relative ease of SVM within a year or two. MPxIO (STMS) could take a while, and the Dynamic Reconfig features of Solaris are unnecessary on typical x86 systems which don't support hotplug, so they're almost guaranteed not to make it into Linux for lack of users.

Solaris is a different world from Linux. I use both every day; Linux on my laptop, Solaris on servers. Linux is a great desktop, and a perfectly adequate small server (DNS, File/Print, small Database, etc). Truth be told, we recently replaced an (old) Sun Solaris DB server with an Apple XServe, with great performance improvements. That's largely because the old server was an ancient single-CPU 360MHz Netra!

Steve

You do realise that Solaris isn't built with gcc, surely.

Not quite true. All AMD64 libs/binaries were built with gcc, due to the Sun Compiler not being ready before Solaris 10 GA went FCS. You can read more about this on opensolaris.org...

If you look at the code and doco available at www.opensolaris.org you'll see that the amd64 stuff was in fact built with gcc. Sun has committed to getting their required changes back into the mainstream of gcc, so you need not worry about that. The latest I've seen from the SunStudio10 compiler is that it is much better for amd64 than gcc.

I'm surprised you replaced a single-cpu netra with an XServe, if only because I would have expected you to see at least comparable performance numbers from an equivalent Sun box, and have the added benefit of keeping your investment in Solaris.

The parent seems to be a personal opinion, and not posted as part of an "official" OSN staff function. I agree with the previous respondant to your post, that comments such as that should not be exempt from the mod system. It was a valid criticism. I found it interesting that he ended up getting moded all the way down.

For the record, I did not touch his comment. But it was completely off topic, which other admins frequently mod down.

Anyway, it has been explained many times why all admin comments are exempt from moderation. I really don't mean to be mean, but I will be blunt: deal with it. It's a necessary evil. If the administrative staff posted moderatable comments, they WOULD be modded down by trolls. We are reviewing the moderation system, but it is extremely unlikely that admin comments will ever be publically votable.

I think it's for the best that they cannot be moderated. There should largely be no useful purpose in moderating admin accounts. If an admin is engaging in inappropriate behavior, it's fairly easy for you guys to deal with.

My point is that if the admin is posting as simply a poster, contributing to the conversation, rather than as an Admin, then not only can their posts not be modded down, they can't be modded up as "interesting" etc. either, that's all.

If they're posting in an "official" capacity, then, by all means.

But if they're participating as part of the community, then why not let the community participate back.

Sounds like a reasonable suggestion to me.

Depends.. if you compile your app staticly it can run on any version of linux on the same architecture. If you use system libraries, like glibc, then you have to use functions that maintain backward compatibility between glibc versions or risk a compatibility problem.

Recently I updated glibc on my Mepis system, using a Debian glibc package. If what you say is true all the apps on my system should have stopped working. They didn't.

Depends.. if you compile your app staticly it can run on any version of linux on the same architecture.

Sadly, that isn't always true either. For example, if a developer foolishly decides to statically link libc, and a particular libc does kernel calls or makes assumptions about /dev or /proc that are no longer true in later versions, then the application may no longer function.

Static linking isn't a solid guarantee of the binary compatability or usability.

If you use system libraries, like glibc, then you have to use functions that maintain backward compatibility between glibc versions or risk a compatibility problem.

Recently I updated glibc on my Mepis system, using a Debian glibc package. If what you say is true all the apps on my system should have stopped working. They didn't.

You got lucky. For example, Oracle 8 and 9 required relinking when installing on different Linux distributions or different versions. The reason being symbol differences between versions. Additionally, most of the time you'll see ABI differences with Linux applications with C++. This happens especially when GCC versions change. The C++ ABI for GCC has changed three times as far as I know in recent years.

The same problems exist on most platforms; if Solaris makes the situation better then that's great. Usually people get around it by either statically compiling or just shipping multiple binaries.

You act like it changes monthly, it's not nearly that often .

I'm not attempting to flame you or anything, but how does Solaris not have a good installer? I found it to be quite pleasing in fact (because it worked). I'm not sure what the worst Linux distro installer is... does anyone have any thoughts on this? I heard Debian was a pain to install, but I didn't think so. I breezed through the Gentoo install. I've read that many people are chased away from FreeBSD because of the installer. Took me a few minutes. So I don't get it, what makes an installer bad and/or hard to use? I could understand if it dropped you into a shell and expected you to magically know what to do (similar to Gentoo, except Gentoo comes with extensive documentation), but the Solaris install is very easy and straight-forward. So what makes it "not good"?

I'm not attempting to flame you or anything, but how does Solaris not have a good installer?

I believe that was sarcasm.

I believe that was sarcasm.

You say it's sarcasm, I say it's a valid question. I'm not a Solaris zealot, lambasting someone for saying it doesn't have a good installer. I was attempting to ask the poster why he felt the installer wasn't good. Nobody gets anywhere by saying "program x's installer sucks, but program y's is good"... you need to find out why it's not good. That's all I was attempting to find out. There was no sarcasm (nor none intended) in my post.

Solaris is incredible... once you have it installed. The install process is horrible. It takes hours, it is incredibly error prone, and the documentation is quite poor if you stray off the beaten path.

A few weeks ago I installed Solaris 10 on an Opteron machine. I mistakenly select the "GUI" version of the installer. Unable to back out of my decision, I elected to continue and save time.

During the install process I HAD to use a floppy to install a patch that, without, Solaris would only recognize 18g of my 73g hard drive. There were no options other than the floppy, I could not use the CDROM, as it was in use by the installer. USB floppy drives, cd drives, hard drives, and network patch installations are not supported. Later in the installation, the process froze and I had to restart the machine.

During the second installation attempt, accidentally selecting the GUI again (my bad), the installer froze at the same point. There were no messages in the existing console window, but I finally found that if I manually opened a different console terminal, there was a text prompt asking me to re-insert the floppy (which was still inserted). The installer would not continue without my correspondance in the terminal. It was almost luck that I opened up this window, given the existance of the other window named "Console" already on the desktop.

This is one sad tale, but I have several from various installs of various Solaris version of various hardware. I feel that being anything less than an entry-level Solaris Guru requires time and frustration.

I sincerely hope that the Open Solaris folks take a page from Debian's book and provide a powerful but easy to use installer that given a decent net connection, can be installed with only a few hundred megs of download, rather than four CD's worth.


Thankfully, once installed, Solaris can run for years without problems. Before last September, before a two day power outage, I had solaris machines that had uptimes of over two years but had never required any maintenance beyond simple security patches.

And we have several HP-UX and Solaris machines which haven't been rebooted literally in years, running internally. 770+ days uptime on one that I can remember.

It always amazes me when people post statistics such as these. They are not impressive at all but admittance to being a sloppy System Administrator.

I do not administer those systems, number one.

Number two, these are internal systems.

Three...if the systems are properly secured/isolated, rebooting is very very rarely required. Shutting down and restarting upgraded services should be enough in nearly all cases.

To suggest otherwise implies that kernel level patches are required for security or stability. If they are, there are more fundimential problems either with the supplier (regaurdless of type; commercial or not) or with the admin's abilities.

That said, if the admin is just lazy they likely have not properly secured either the system(s) or the network.

Well this machine is on an isolated LAN even within my organization. I presume the fact it has not been patched is a matter of priority, number one, given its extreme isolation and two, the system is fairly stripped down to the bone as it is.

That being said, yeah, if you run a networked system with remotely exploitable kernel flaws, you'd better patch and reboot. And then of course there are the issues (not an issue here) of local compromises as well - people walking by.

Nevertheless, it is a rough, if imperfect indicator of its stability that it continues to do massive data crunching on a daily basis, and has not had to be touched in quite some time. I mean, I rely on it.

What it does not do is have dozens and dozens of simultaneous users nor have to deal with overwhelming network load, so that has to be taken into account.

The CPU and memory are however pushed hard by local processes.

My point in even mentioning this was that the machine has not wedged in over 2 years of 24/7 use; no more than that. No one's trying to win some kind of uptime contest, but as an indicator of reliability, you can glean something from an uptime. It just doesn't say that a machine that has been rebooted regularly is necessarily less stable.

On this subject as per Linux, I have wedged it a few times, but all of these problems were tied to video. Eventually I turned off render acceleration in the X configuration file and it's not wedged once since.

Overall then I've just not noticed any big stability issues with any of these UNIX-like OSes; it would not be a factor - for me - in choosing which to run based on what I know.

What would matter to me is timeliness of security updates (How is this for Solaris? I ran a Cobalt RaQ for awhile - a whole different product line, running Linux - and the timeliness was dismal). Debian had its problems recently...

Then, ease of being alerted of them and applying them would be fairly important. I'd prefer not to have to dick around too much to get them applied quickly.

All of this is of course trumped by what applications you need to run, if it is not a standard cross-platform application like Apache.

And then there's cost. I mean, look, some businesses will spend whatever is asked for in IT, but others will not. Sometimes it's organizational. Sometimes it's a matter of, "We can do this, but not as well, for a lot less money." and sometimes that wins.

There is never a hard and fast rule in business regarding the up-front costs of hardware and software. Sometimes you can say, "Hey I can do this without vendor support for a fraction of the cost on hardware we already have with a free OS" (not so much an issue now with Solaris I guess), in which case sometimes the answer will be yes, and sometimes the answer will be no - no meaning, we require vendor support, and will not deploy a system without it.

My company suggests and recommends Solaris hardware and OS, but tolerates, say, Linux, in certain instances. In two instances we deployed remarkably useful resources because we could do it for (virtually) free. We could not have done it otherwise.

Now with Solaris being free, this consideration changes some.

Three...if the systems are properly secured/isolated, rebooting is very very rarely required. Shutting down and restarting upgraded services should be enough in nearly all cases.

To suggest otherwise implies that kernel level patches are required for security or stability. If they are, there are more fundimential problems either with the supplier (regaurdless of type; commercial or not) or with the admin's abilities.

That said, if the admin is just lazy they likely have not properly secured either the system(s) or the network.

I have had this argument time and time again fact is best practice in well maintained environments is regular quarterly reboots if possible. Truth is you don't have to reboot Unix but you may work in an environment that is not architected for proper maintenance and/or you user base is unreasonable with their expectations of uptime and availability.

"It always amazes me when people post statistics such as these. They are not impressive at all but admittance to being a sloppy System Administrator."

What, reboots are necessary? The only people that consider rebooting these machines "normal" are those that grew up using Windows. Nuff said.

So you're telling me that you can update the Linux kernel without having to reboot?

Get a clue, fanboy.

*cough*kexec()*cough*

That said exactly how often do you think a production system gets a kernel upgrade?

I think you should always that everything is in working order after you updated that includes checking if the system will reboot if for example the power fails. Or in case that it does crash that the disruption is limited to the time it takes to restart.

Always check if your update works or you are a sloppy admin

jerven

Sloppy? Could you elaborate?

We had a Unisys 2200 mainframe when I worked at Northwest airlines that handled all of our type B message traffic. Lots of text messages send in real time to/from other airlines and facilities mainly via dedicated comm lines.

A critical OS patch came out just before Y2K that had to be applied, and they took the system down to apply it, but that was the first time that the machine had been down for any reason in almost three years.

Why? Because the continued up time of this particular system was a lot more important to the airline than the application of patches that weren't considered critical (taking the system down for even a few minutes would result in a backlog of hundreds of messages that could take an hour or five to work off the queues, and many of those messages were time-critical for flights).

Moral of the story: not all systems are administered using the same criteria; mapping your own local admin experience or procedures to all systems will sometimes result in serious failures.

Maybe folks should stop putting obvious troll phrases like "Linux Killer" in the headlines...?

As lot as that typemof crap continues, you'll continue to see flamage.

Amen to that. There is too much sqabble between BSD/Linux/Commercial UNIX. We should work together to achieve our somewhat common goal.

Solaris demands a higher end I/O subsystem not because it is slow but because it does not cache writes as Linux does.
Care to expand on this claim?

Writing to disk is generally a driver issue, not a kernel issue. Of course, drivers are linked to the kernel, but it's not a policy decision.

I have worked with Linux for 8 years and Solaris for 6 years, and I honestly have no idea what you are talking about.

At the application level, you write to disk (I assume you are talking about disk), and consider it done. The kernel passes that down to the driver, which deals with the write as appropriate.

A typical Solaris system would hand that down to the volume manager (SVM/VxVM), which would pass it to the qlc driver (via MPxIO/DMP to send the request down the optimal path) to write the data to disk (or disks, if RAID'd). Those disks themselves may be LUNs on a (possibly SAN) array, with their own RAID configuration.

I don't understand at what level you claim that cacheing comes into this, for either OS, the method is pretty similar.

I don't know about Solaris demand[ing] a higher end I/O subsystem but I do know that Solaris turns off the write cache on scsi disks. This is one of those data integrity thingies that Sun folk seem so keen to harp on about. I'm pretty sure you can't turn off the write cache on IDE disks. Of course if you equate cost with "higher end" then the OP's statement is true.

Personally, if I'm gonna worry about data integrity then I'll invest in more reliable hardware. And yes, for me, scsi is more reliable hardware.

Rule of thumb - if you get modded down into minus figures and no one replies to your post then you're right ;-).

As far as I can see I'm the only one here who actually read the article, itemised it and replied to it. The rest of you are just arguing about absolutely nothing, as usual.

I generaly do not participate in discussions about licenses, but I'll make an exception. For the sake of the argument, you should re-read your post and think about what you wrote : your reasoning is extremely flawed, especially at the end where you're making a common mistake.

"Stealing the work of others to your own benefit is NOT a freedom, it infringes on freedom of others, and as you should know, freedom stops where it infringes on others' freedom. What you described (the stealing) is a dangerous power, not a freedom. I'm sorry to say that your logic seems a lot more flawed than RMS one."

using source code under the BSD license for you're own closed source app isn't stealing its both completely legal and completely ethical for things under BSD. nor is it infringing on anyone else's freedom to use the original source code.

Very true. To go around blathering about how a perfectly acceptable use of the code under the copyright holder's chosen license is "stealing" is farcical. The copyright holder knew the terms and implications of the license when they elected to use it.

You just forgot the most important people in the equation : the one that produced the first code, the one that provided the primary functionality, the main developer !!

Imagine this scenario: you need to develop some real-time software for this company. You're a hired developer. They work with embedded systems for medical devices and, consequently, they have IP issues and have proprietary software, and won't touch GPL. You convince them to put stuff under BSD code. This being your area of expertise, the next time you work for such company, you can simply suggest to use the same code.
Imagine you work with for an internet company. They need you to develop some C++ socket library. You convince them to use the BSD license. This being your area of expertise, under your next job, you can still use your library.
Imagine you participate in a project that's under current development under the GPL, but that you didn't start. You submit patches, they get accepted. This software is primarily being developed by a company. The choose to dual-license the software. They make money, you don't.
Imagine a scenario like the above. The original developer says "thank you very much" and decides to stop OSS development and releases under a proprietary license. You have worked for free.
Imagine you're a 20 something with no real expertise, and a no-good job. You scream and scream about the GPL. You use a lot of your mom and dad's money, but then you release stuff under the GPL. When you finally learn something, you come to realize you've worked for free. It was fun, but it was dumb. Unless...you get a job a Linux company, like RedHat. Oh, but you couldn't, you really don't grok UNIX that much. If you did, you'd think it sucks when you $man hier in Debian, and you learn that the man page has nothing to do with the reality.
Why don't Linux people understand that the true beneficiaries of the GPL are big corporations? They're so greedy, they needed the GPL; but you know what? The BSDs do just fine. Now, if only some people would stop refusing to be brain washed by the GPL, and accept the fact the IBM, HP, etc are only commoditizing their complements?

But, on a bright note if it's UNIX, it's all good !

Do you know what you are talking about? Have you actually used Solaris? Solaris x86 is already a "player" in virtualization with Zones and Containers, which is vastly superior to anything Linux has to offer when used together. Sun is also contributing to the development of Xen, see the discussions on www.opensolaris.org.

I think you misundertood the original poster. That post was pro Solaris. But I admit that it was a little difficult to understand what was really meant.

Then again rereading the post I think. You were right, Robert. The first half is anti but the second half is pro.

Anyway, Virtualization has many forms, some like VMware are already supporting Solaris (note the recent deal). Xen is actively being worked on from the Soalris guys. So yes Robert was right in pointing out an obviously ignorant post.

Been a long day ;(

You said "and its not like anyone is complaining about stability for either OS for general use. "
I care about stability alot; that's why I use linux not windows even for general purpose desktops.
Solaris or linux do not crash on me with a bad CD-ROM; or If a drive fails it will not take the whole OS with it.
And let's just remember, if an OS like Solaris is a little bit more stable than linux then this is not because of its design but lake of hardware support.

At least I can unplug the keyboard on Linux without taking down the server.

I know you meant that in jest but removing the keyboard traditionaly on SPARC workstations does a break and drops the system to the ok prompt. This was done to enable one to drop to the debugger and break out of deadlocks because the break caused a high priority interrupt. This was meant as a debugging feature. Which is no longer true on systems with USB keyboards.

Like doing the 3 finger salute on a linux box. You can happily do a Ctrl+Alt+Del on a sparc box with out bringing it down but can't on a linux box.:)

Like doing the 3 finger salute on a linux box. You can happily do a Ctrl+Alt+Del on a sparc box with out bringing it down but can't on a linux box.:)

Comment out one line in /etc/inittab and Ctrl-Alt-Delete does nothing but wear out those keys.

>solaris ahead:
>*AIO

Linux has it.

>*threads

Solaris went to 1:1 threads after they saw how good Linux's were.

>*testing

Err, yeah Linux is tested.
You have companies like IBM, Intel, SGI, HP, OSDL, Novell, RedHat, etc continually running regression tests and performance tests on the kernel. Along with all the regular users who test beta kernels.

>*guaranteed APIs(impl. & iface)

Yeah Linux implements POSIX too.

>*deterministic performance scaling (up&downside)

Linux can run on systems without MMUs and 1MB of memory. It can run on a MMU based system with a full TCP/IP stack and busybox in 2MB with a few hundred K left over.

Linux can run on systems with 512 CPUs and 6TB RAM.
Intel regularly runs TPC performance tests on systems with over 10 000 disks.

>*regression testing

See: testing.

>*/dev/poll

epoll

>*good docs @ docs.sun.com (wirtten by devs and
>technical staff, not end user tutorials like TLDP)

Plenty of good Linux docs.

>*POSIX compliance (single unix spec 3)

See: APIs

>=> real unix
>*sun compiler suite (gcc is teh crap!)

It cannot support nearly so many architectures as gcc can, nor can it give the performance of icc.

>*dtrace (the best thing since sliced bread/unix)

IBM's kprobes do something similar. They need a decent user interface however.

>*scalable (not only by the factsheet, your telco runs
>its billing/call processing/network monitoring/...
>system on it)

Err, that's worse than a factsheet. It is simply anecdotal rubbish. How do you know what my telco runs?

>*tree letter acronym agencies use it

Wow, they probably use Sun's SPARC CPUs too, which are crap. Too bad about your tax dollars.

>*zones/partitions hw resources are nice to manage

So are lpars and VM guests. IBM's been doing that with Linux before places like Sun caught on. In fact, IBMs been doing that stuff since the 70s I think.

> - a linux users since 96

- a solaris user until it got overtaken by linux.

Solaris went to 1:1 threads after they saw how good Linux's were.

WRONG. Solaris has had a 1:1 thread library (secondary) since Solaris 8. The switched to it as the primary library in Soalris 9. Years before the linux guys could even decide which 1:1 library to pick.

>WRONG. Solaris has had a 1:1 thread library (secondary)
>since Solaris 8. The switched to it as the primary
>library in Soalris 9. Years before the linux guys could
>even decide which 1:1 library to pick.

Doesn't matter. They still had the libraries there, and the concept was proved. This is why Sun went to 1:1.

Doesn't matter. They still had the libraries there, and the concept was proved. This is why Sun went to 1:1.

It does matter. The point is that SUN had 1:1 threading long before Linux ever did. Linux had nothing to do with SUN having 1:1 threads.

Doesn't matter. They still had the libraries there, and the concept was proved. This is why Sun went to 1:1.

My point was Solaris had a 1:1 library ages before the linux guys even dreamed of having one.

Solaris 9 with the 1:1 thread library was released in May 2002. The linux guys were still doing benchmarks isn september 2002. What you claim not only didn't happen it was near impossible to have happened.

Look at the Solaris development cycle, it takes a couple of years to develop or release a major Solaris release. Solaris 8 was released in 2000 and had a 1;1 threading library. The linux guys didn't do squat to influence Solaris. It was the other way round.

>My point was Solaris had a 1:1 library ages before the
>linux guys even dreamed of having one.

Do you even know what 1:1 threading is? Linux has had support for 1:1 threading basically since it first became available (at least since version 1).

When was Solaris first released? 1993? Gee, Linux may have had 1:1 thread support before Solaris even existed.

Anyway, the point is not who had what first, the point is that the Linux community first proved that high performance 1:1 thread libraries (matching M:N ones) could be done.

Anyway, the point is not who had what first, the point is that the Linux community first proved that high performance 1:1 thread libraries (matching M:N ones) could be done.

No, the UNIX world proved it could be done and Linux followed. Solaris had M:N threading before Linux did as well. Heck, a lot of operating systems had M:N threading before Linux did.

Do you even know what 1:1 threading is? Linux has had support for 1:1 threading basically since it first became available (at least since version 1).

Obviously you don't. Linux never really had threads they had processed which were done using clone(). Since Torvalds contention was that linux's context switch was as quick and lightweight as threads in other OSes linux didn't need threads. However the user level pthread library was bound to one process and scheduled by the library. Now with 1:1 threading each user thread becomes a linux process consuming different PIDs.

Solaris always had the lowest schedulable quantum as a thread bound to an lwp. each user process can have one or more lwps allowing for M:N threading using soalris' own thread library, pthreads were scehduled by the library. In Solaris 8 an alternate pthreads was introduced to do what the linux guys eventually did.

You really shouldn't talk if you don't understand how stuff works.

Obviously you don't. Linux never really had threads they had processed which were done using clone(). Since Torvalds contention was that linux's context switch was as quick and lightweight as threads in other OSes linux didn't need threads. However the user level pthread library was bound to one process and scheduled by the library. Now with 1:1 threading each user thread becomes a linux process consuming different PIDs.

I think you should go relearn your Linux history. Torvalds did think that the Linux context switch was fast enough to not require a dedicated threading system. This is why LinuxThreads (which has been around forever) and NPTL both use the clone system call. clone() creates "threads" and "processes" (as defined by other OSes) because it lets you specify whether you want a shared address space between the new and old process or not. So both LinuxThreads and NPTL are 1:1 threading systems. They both create new "threads" with new PIDs. The difference is that NPTL brought along kernel changes which made NPTL follow the POSIX standards closer, particularly regarding signals and it provided a significant speed increase. There was (is?) a user land pth library which follows the pthreads standard, but it is cross platform, not Linux specific. And to the best of my knowledge, no large application ever really used pth.

So, to repeat myself, Linux has used a 1:1 threading model pretty much since the day the LinuxThreads library was written.

In Solaris 8 an alternate pthreads was introduced to do what the linux guys eventually did.

If by "eventually" what you mean is "in the mid 90s" then yes, you are correct.

You really shouldn't talk if you don't understand how stuff works.

That has to be my absolute favorite comment when someone posts something which is completely inaccurate. Now, if you'd like to go learn how Linux threading worked historically, go read up about LinuxThreads. And as a little helper, here is the first sentence from FAQ entry K1 (at http://pauillac.inria.fr/~xleroy/linuxthreads/faq.html):

"LinuxThreads follows the so-called "one-to-one" model: each thread is actually a separate process in the kernel."

Now, I have no idea whether or not the Solaris kernel guys even thought about Linux when they went to a 1:1 threading model. Either way, Linux's "native" thread system has been 1:1 for a long, long time.

Solaris is not a TOY for HOME users. Go away, we don't need you, and yes i am arrogant & stubborn.

Given this attitude, I'm uncertain how you intend to see the Solaris user base expanded?

If I can't pay with the OS on my own time for evaluation purposes, it's going to be a lot harder for me to learn about the platform, and I'm not going to recommend anything to my employer w/o at least some general experience with it.

One man's "hobbyist" is another man's employee who is simply interested in learning more about other technologies during off hours...

Read this:

http://docs.sun.com/app/docs/doc/817-0552/6mgbi4fh7?a=view

You can't expect Sun to support EISA forever, the last EISA machine I used was junked 5 years ago.

The Digital Server 3000 does have a PCI bus which is used. My point is that it could have ignored the EISA bus.. and still loaded, instead of giving up as soon as it saw the EISA bus.

And how old are those machines exactly? The kernel still has to have support for EISA whether the machine has PCI slots or not, so I can easily see the installer choking on that. This is one time where I think Sun has it right in removing overhead for things that are not widely used, like EISA.

And your qualifications for making that bold blanket statement are?

In the example of Maple, Maple dropped the product support on Linux (if I understand you) and it will not run on new-ish distributions. Those old distributions still exist, BTW, but what is important to note here is that the proprietary software company left you high and dry in terms of future proofing for its customers - this is *not* the fault of Linux.

I think the point he's making is that Linux doesn't have a stable ABI. Solaris can still run unmodified Solaris binaries that were built properly all the way back to 10 or more years ago! Linux can't do the same thing...

I have Linux games released by Loki that will no longer run on newer Linux distributions because of ABI differences. It isn't the application's fault. It's because the Linux platform doesn't even have the concept of a stable ABI.

Things like GTK+ are more likely to break things now.

That's why I can't run Firefox at all under Mandrake 8.2, for example.

You're confused. If you are talking about a user-land application, you almost certainly mean GLIBC which changed twice in big ways since the first release of Linux back nearly 15 years ago.

No, I'm not confused. I'm talking about the Linux platform in general. From the Kernel Driver API to the ABI used for C++ programs, to the ABI for commonly used libraries such as glibc, etc. the general lack of release engeineering principles or proper ABI stability is daunting. I've been using Linux since 1994. I think I know a thing or two about it

Not Linux' fault at all, though most distributions ship GLIBC. If you were talking about drivers, I'd partially agree with you, but you're not.

It is the fault of Linux. The fault of Linux distributions, library authors, architects, engineers, and designers. Incomplete design gives incomplete results.

As a developer of Linux, BSD, and Solaris applications over the years I have learned more than I ever wanted to know about the darkside of various operating system platforms. The ABI issue on Linux has stung me more than once. Driver, glibc, and in many other ways.