Submitted by “Google is experimenting with the open-source version of Sun’s Solaris operating system as a possible long-term prelude to replacing its massive global network of Linux servers, according to sources. With dozens of data centers worldwide estimated to house hundreds of thousands of Intel servers supporting its flagship search engine, a Google move to OpenSolaris would be another of several recent votes of confidence for the platform.”
“Solaris is the executable version of the operating system, technically OpenSolaris is the source code version of the OS”. WTF?
If the rest of the article is as “well-researched” as that paragraph I’d say this is a potentially baseless, sensationalist article with no more substance than an attempt to grab headlines.
Maybe you should complete the sentence “According to Sun and other sources, a number of other companies are using Solaris 10 or Solaris Express, the executable version of OpenSolaris, which technically only refers to the Solaris source code and the community around it.”
In order to build OpenSolaris you need to install Solaris Express, so technically the statement is correct. Poorly worded but correct. And just what makes it so sensationalist? Did you read it in its entirety, or just get disgusted at the idea of Google looking at something other than Linux?
If you’re running compiled versions of OpenSolaris source code, you’re running…uh…OpenSolaris. If you need to run anaconda to install Red hat Enterprise Linux, that doesn’t mean when you’re finished you’re running anaconda, not RHEL. And I didn’t say that it was sensationalist: what I said was that if the rest of the article was as poorly researched as that part, there’s not much chance of it not being sensationalist.
Edited 2006-09-21 23:13
In order to build OpenSolaris you need to install Solaris Express, so technically the statement is correct. Poorly worded but correct. And just what makes it so sensationalist?
One note : of course there have already been people who have built all binaries and used them to set up a “pure” (no code compiled by sun) Opensolaris system (
http://www.genunix.org/distributions/belenix_site/belenix_home.html ). So you can build Opensolaris from within Opensolaris never having to see a Sun binary.
The benefit being that when you download Solaris Express you are still licensing software from Sun, which becomes unnecessary.
I think, this is a good thing since Opensolaris alot of ppl are starting to really consider Opensolaris.
and with some pretty cool advance stuff like ZFS and dtrace ect is good….
-Nex6
yes, i think the same.
Solaris is an excellent OS. It has a ton of useful features, including hardware hot-swapping, etc. A lot of mission critical orgs are already running it. If it truly becomes open source, it will give Linux a serious run for its money.
Solaris still needs to improve its hardware support. To be fair, that’s true of the BSDs and, I hear, of some popular Linux distros.
More properly, hardware vendors need to improve their support for Solaris, *BSD, etc.
True; the worse part I can’t stand is he hand wringing that occurs by the vendors, first there isn’t enough customers, and yet, every study has shown that this ‘small niche market’ not only purchase a tonne of hardware per year, they’re also willing to pay a premium for good hardware – case in point, my decision to purchase a Toshiba over the cheaper Dell.
The next excuse was “not a stable API”; here is OpenSolaris and BSD, all which have stable driver API’s, and very good API’s to write drivers for – Nvidia has grabbed the bull by its horns, and made sure the stone kept rolling; too bad AMD is full of Windows fanboys hand wrining rather than delivering good quality drivers for their customers.
Solaris still needs to improve its hardware support. To be fair, that’s true of the BSDs and, I hear, of some popular Linux distros.
Boring FUD. Please explain what server environments Solaris, *BSD and “some” Linux distros should be supporting. Detail your experiences with such if you don’t mind…
If it truly becomes open source, it will give Linux a serious run for its money.
It is truly open source now. Even the Free Software Foundation acknowledges the CDDL as a Free Software license.
Most of the source code is, but I understand that there are still little things like man pages that are not yet under an open license. It will take time for OpenSolaris to be fully packaged and accepted.
I’ve been using Nexenta from time to time and it’s great. The developers (or company?) is a little odd though. The don’t have completely open development practices.
Most people wouldn’t consider man pages to be part of the source code of an operating system, which is what was being discussed.
See:
http://blogs.sun.com/jimgris/entry/the_opening_of_solaris
Or:
http://www.opensolaris.org/os/about/roadmap/
The man pages are coming this month.
Not even Debian has “completey open development practices,” see debian-private as an example: http://lists.debian.org/debian-private/
“Most people wouldn’t consider man pages to be part of the source code of an operating system, which is what was being discussed.”
I consider man pages and other documentation to be an important, though not vital, part of an operating system. Man pages are a key characteristic of the unix family.
It will be nice when they are opened, as I’ve missed them in Nexenta.
It’s true that Debian discussions are not completely open, particularly in regard to “political” discussions. Nexenta, on the other hand, seems to limit not just some discussions but also the software to some extent. For example, I found the post below confusing. There is probably a good reason for it, but it is contrary to what many OSS nerds are accustomed.
While I have no doubt that Google might be considering moving to Solaris (afterall, there were talks about mergers/alliances), the whole article sounded like FUD. Why is it that Linux has worked so well for so many companies for so many years and then all of a sudden it is considered a resource hog and unstable?
I will give the guy in the article who had server crashes every three days a clue, chances are it is an application bug made by your own developers. As to why it seems to work on Solaris so great; this is probably because after you migrated to another OS, your lazy programmers couldn’t point fingers anymore and had to actually go and debug your proprietary app and actually ended up fixing it.
Edited 2006-09-21 23:07
While I have no doubt that Google might be considering moving to Solaris (afterall, there were talks about mergers/alliances), the whole article sounded like FUD. Why is it that Linux has worked so well for so many companies for so many years and then all of a sudden it is considered a resource hog and unstable?
‘Resource hog,’ And ‘unstable,’ Are relative terms. To some (Open)Solaris is seemingly more stable and less piggish operating system than ‘Linux’. I can’t vouch for that as I’ve not used OpenSolaris extensively but I think that’s consensus among some.
I really don’t think you can stack them against one another. Not generally. It depends on the configuration, deployment, distribution and software running on them.
You’re right though, it is FUD. The article says Google declined to comment and that Sun only said “other companies” were interested in, or have already switched. Nothing about Google was mentioned by Sun.
Their only source was “a Solaris administrator…who recently was interviewed for a job”. No name, might not even have got the job. Big woop.
Fud indeed…
I will give the guy in the article who had server crashes every three days a clue, chances are it is an application bug made by your own developers.
If an application is causing their Linux systems to panic, you bet there’s something wrong with Linux. If code outside the kernel causes a panic, it’s most certainly a bug and a problem.
Both should exist, and both should fully cooperate with each other. Diversity is good, especially where software and operating systems are concerned. We all know the dangers of monoculture so I won’t go on a rant here.
But it does make me wonder if this isn’t backlash over GPL v3. Google likes and supports OSS, but not when it comes to Google’s internal operations. If GPL v3 forces them to reveal their tradesecrets, then I can understand Google looking now for a Linux replacement before it becomes a problem for them just in case the Linux kernel does adopt GPL v3. Interesting turn of events.
Edited 2006-09-21 23:09
But it does make me wonder if this isn’t backlash over GPL v3
The GPL only applies to software Google might release. They can do whatever they hell they like with GPL software internally.
Why all the anti-GPL sentiment (one might even call it FUD) on OSNews lately?
Edited 2006-09-21 23:11
GPL might apply to Application Service Providers who change the code.
See:
http://www.osnews.com/story.php?news_id=15487
Why all the anti-GPL sentiment (one might even call it FUD) on OSNews lately?
Nice to see, I’m not the only one who’s noticed this…
“Why all the anti-GPL sentiment (one might even call it FUD) on OSNews lately?”
I’m anti-gpl now? That actually kinda hurt. I’m very pro-gpl, and anti-Sun/CCDL.
And what I asked and said isn’t FUD. One of the points of v3 is to plug the webapp hole in the GPL, one in which Google makes extensive use of. If the GPL v3 is applied to GPL code, it will prevent Google from exploiting this hole anymore. Thus the need to use something else. All I asked was a legitimate question.
“One of the points of v3 is to plug the webapp hole in the GPL,”
The webapp hole? I’ m seriously curious here; what is the webapp hole?
“The webapp hole? I’ m seriously curious here; what is the webapp hole?”
It is kinda similar to the email hole, web-posting hole, or file sharing hole. Basically, it is a very lame attempt that says because we shared bits through the network, I am entitled to see all your source code.
Edited 2006-09-22 14:06
Basically it is if you run the gpled software on your server and pipe the output over the net, you’re not distributing the software thus you don’t have to provide the source or upgrades.
The point of RMS’s gpl is if you’re running a piece of software, you should be able to modify it and redistribute the software. So the conflict is if someone makes a piece of gpled software with the intentions that it follows RMS’s goal, then a third party, you, comes along and lets others make use of the software in a webapp config, but won’t share the code or upgrades is violating the spirit of the gpl.
v3 seeks to remedy this. Right or wrong, that’s the basic version of it.
Linus has said publicly and categorically that he’s unhappy with GPL3; as for what “might happen”, then you could equally say that MS “might” adopt GPL2,3,4 for Vista or Vista+1, or that Sun “might” adopt GPL{2-4} for OpenSolaris. Businesses don’t go in for “this might happen” very often.
Linus does not control all of the Linux kernel code, only a tiny portion of it. Individual developers can dual license their code, and some are at least considering it. And if enough of the kernel is dual licensed, the rest can be rewritten and a new release of the kernel is possible under the GPL v3. Don’t think it can’t happen.
Oh, and I forgot to mention it, the kernel is only one little piece of code Google uses, and I wasn’t really refering to the kernel so much as I was refering to other GPL code Google uses in their webapps.
Edited 2006-09-22 09:44
Linus does not control all of the Linux kernel code, only a tiny portion of it.
No, he retains ultimate control over all of it; the kernel is copyright Linus Torvalds, a measure instituted despite his misgivings to stop exactly this kind of problem.
Oh, and I forgot to mention it, the kernel is only one little piece of code Google uses, and I wasn’t really refering to the kernel so much as I was refering to other GPL code Google uses in their webapps.
In which case they have even less reason to worry, since it’s their code.
I’m pretty sure the Linux kernel is copyrighted by the contributors.
Linus owns the trademark for Linux.
Linus does not have ultimate control over the code in the Linux kernel. He only has control over that which he has copyrights to and the terms of the gpl he has licensed his software under. The Linux trademark is Linus’, but I could go over to kernel.org and get the most current snapshot of the kernel, call it something else and fork it and there isn’t a thing Linus could do about it. Why doesn’t that happen, cause my fork doesn’t have the developer base Linux does and my fork would quickly be over-shadowed by the Linux kernel. That’s the freedom of the gpl. Once it’s out there, not even Linus himself can take it back.
The Linux kernel is unlikely to go the GPL3 way. It is apparently impossible to change that now: too many files are not licensed ‘GPL2 or later’ but just GPL2. We might sooner see Solaris under GPL3, and I’m not being sarcastic.
As for the rest of a working computing system, each and every program will go GPL3 or not depending entirely on the whims of its own developers. This situation affects Solaris just as much as much as Linux; after all, they will run the same Gnome, Apache, Zope, Bash, Python, Ruby, Java, PHP, PostgreSQL, MySQL etcetera.
They might like the effect of GPL3 on programs for service providers, or they might not, but that is not a Solaris/Linux problem.
Again, I wasn’t really talking about the kernel as I was talking about the code in the webapps. But again, Linus saying the kernel won’t be GPL v3 is wrong. He doesn’t want it, but he only controls a very small portion of the kernel code. If the other developers relicense their code so it falls under the GPL v3 as well, and if enough if it is done so, you could see a rewrite of the remaining GPL v2 code into v3 code. I’m not saying it’s likely, but not nearly as improbable as some are making it out to be.
Linus does NOT control “a very small portion of kernel code”; this is just FUD. Yes, people like Andrew Morton handle bits of code for him, but they do it at least partially because he trusts them. Implying they could relicence their “little bit of kernel code” under some other licence is as silly as saying the head of the Windows development team could “relicence their little bit of MS code.” In both cases, it’s not against the laws of physics, but it WOULD be against the law of copyright.
Linus does not have the copyright on most of the Linux kernel, he himself has said this. BUT, all of the contributors have released their code under the gpl, thus eliminating the need to own the copyright so long as you abide by the gpl. That’s how Linus takes all these bits of code from thousands of contibutors and put them together to make the kernel. It’s not because Linus has writen the code himself or all the contributers have assigned their code to Linus, it’s because of the gpl.
Once you put it out there, it’s out there, so long as you abide by the gpl. That applies to Linus as well and he abides by it.
And the fact is, there are kernel contributers who might dual license their contributions to the kernel, and if enough is done so, it’s possible the rest could be rewriten and there could be, maybe not in name, but a fork of the Linux kernel that’s all gpl v3. And if v3 kernel is more successful and the community stands behind it, Linus and the gpl v2 kernel will be overshadowed and possibly die off. This is just possible, not what’s _going to happen_.
Yet, as I said, the webapps would be the same and would have exactly the same license problems whether running under Linux, Solaris or Vista.
One possible usage/modification of OpenSolaris, and I’m just imagining here, would be to distribute ZFS across machines/data centers. Also, Google total storage limit could very well reach the limits of a 64bit file system, while it would be extreamly hard to reach the limits of a 128bit file system like ZFS.
as someone who works with highly customized linux and solaris systems (particularly solaris 10 now) this is nonsense. while gnu/linux practically invites you to hack away and tweak a system up to your liking, solaris is just terrible for that, solaris 10 incredibly more so. (though I guess opensolaris may be a step up in that fashion)
if google is actually considering this (and I have doubts), that would be incredibly foolish of them, considering that their in house linux (which probably has as little to do with redhat as our internal fc3 (and before that rh80) systems at my work have to do with stuff that redhat/fedora actually provides. (not to mention the incredible expense of transition an enterprise THAT big from one rather functioning OS, to a completely different one (other than the fact they’re both *nixes))
solaris 10 especially breaks a lot of new ground, and time honored conventions, see particularly the whole smf structure with it’s xml manifest files and all that (even inetd isn’t really inetd anymore). sure that’s ok if you’ve bought the whole boat of sun’s system, but for folks that need a highly specialized and customizable OS, it can be a nightmare. I should know…
but for folks that need a highly specialized and customizable OS, it can be a nightmare. I should know…
It sounds like you DONT know [insert word] about Solaris 10. Maybe you could state a few cases where Solaris is NOT customizable.
If I was google, I would certainly be looking at using OpenSolaris. In the end it is their choice, and it would be foolish of them not to always look at alternatives.
is a stripped OS. And as we know with little code comes less bugs and less troubles.
Opensolaris concentrates on one GUI interface thus making their code lean, unlike lets say KDE, and other GUIs that mix up to bring a huge amount of bugs; not only this but if you like opensolaris you could get support from Sun for cheaper than redhat and get their hardware if you don’t like the preformance of your hardware, redhat cannot offer this.
In my lab installing nvidia graphics drivers is as easy as windows. Just “sh nvidia-driver.run” and reboot. that’s it.
While in linux it varies but mostly you are lost.
Code wise opensolaris is more stable and faster, but application wise it is light years behind linux. which is not a disadvantage to google who would use it as a server.
besides google might use the sun technologies and software which is more easy than in linux due to integration, and lets not forget suns powerful distributed file system.
You havent tried Nexenta have you?
It currently has 12,065 GNU applications ported from Ubuntu.
I had to stop using it though because the alpha version has far to many bugs at the moment. I’ll try it again when 1.0 is released.
“You havent tried Nexenta have you?”
Yes, I have.
“I had to stop using it though because the alpha version has far to many bugs at the moment. I’ll try it again when 1.0 is released.”
Yes, you are right that nexenta is buggy and this doesn’t show that opensolaris is buggy. If you would like to enjoy stability stay with sun OS and not a distro that uses the core of the OS and adds another features upon and make it a different OS. This is somehow like MacOSX which is not truely a FreeBSD anymore because of the modifications, though in the case of OSX the modifications are better than the original.
Sun seems to me a company the chooses its applications very wisely, not to scratch its name, in an industry that grows everyday less tolerant for mistakes.
Hmm, what Nexenta is to fully get the whole thing compiling with one compiler ( 4.1) rather than having to rely on a spaghetti selection of compilers.
What there needs to be is more community contribuion – too bad I suck royally at programming
Folks-it seems like some of you are shocked google is experimenting with Opensolaris. Come on now…it’s google.
Is OpenSolaris getting that much votes of confidence?
I know lots of SMB here in Montreal using AIX or Solaris, but companies who decides to go open source usually go for GNU/Linux, hardware wise it’s easier. There’s a lot more linux admins out there than solaris admins. Beside, linux admins cost less too.
What are the big names that gave OpenSolaris some votes of confidence anyway?
Would make linux fanbois scream bloody murder.
“Would make linux fanbois scream bloody murder.”
That already started if you look at comments =)
Actually, we’re screaming “badly researched, FUDspreading article.”
Pay attention.
I think Scott McNealy said it best around 5 years ago; a win for Linux is a win for everyone.
Transplant that to today; a win for OpenSolaris is a win for all people, irrespective of whether or not they run the said operating system – its one less customer for Microsoft.
I for one am excited about OpenSolaris, its no developing as fast as I’d like it (so I can get it up and running on my laptop), but at the same time, I’m sure once it is at that moment, it’ll be a great replcaement for Windows, and a viable alternative to Linux.
…is that ComputerWorld is an IDG company. Oooooh, look at all that credibility flying away.
nice flirt
I guess if they realy need something that only OpenSolaris can provide it will be a lot wiser from a financial point of view to develop given feature for Linux instead of rolling out a new os for all their servers.
The reason why people are out of a sudden so interrested in [open] solaris is the fact that it’s probably the first unix which claims (can’t say if it’s true) to have stable driver interface ABI which ensures that even if you update the kernel (and according to Sun, even major updates) your binary drivers can stay the same. This is some major + for hw manufacturers who don’t want to give out their driver sources out and have a pain keeping up with all the hacks in linux kernel each new release.
Think nvidia and ati if you didn’t get it yet… but there are tons of others and the big thing is, I think in a year or two, opensolaris will have better hw support than linux on x86 because of closed source drivers.
Just my view of things, I’m not particularly fond of slowaris myself.
Edited 2006-09-22 08:04
What it might mean if they went the way of the proprietary-driver-written-to-fixed-api is that Sun would soon have a sackfull of outdated and unmaintained binary drivers with unfixable and untraceable bugs. (e.g. drivers for ATI cards < Radeon 9500 have reached end of road; the last version is final)
It would also mean Sun’sr hands would be tied if they found they could improve performance and/or add features and/or cure bugs by slightly or not-so-slightly changing the abi. It would make it a lot harder to add in good stuff such as XGL/AIGLX, or FUSE.
Finally, it would also mean it would come with no drivers out-of-the-box, and installers would have to endure an XP-like experience of fishing for every different driver in the net, then manually installing them (and forcing a hardware reset every time? 😉
Then, in the case of Google, think that the hardware they need supported is stuff like raid SCSI controllers and network cards which are very well supported by the current working mode in Linux.
No I strongly disagree.
If Sun follows something like freeBSD does (ABI compat in major versions) they can change ABI/API of drivers (like fBSD did between 4.x -> 5.x) but it will remain pretty stable for longer periods of time with no recompilation requirement and very little change of constant breakage.
With linux even the tiniest change requires a driver recompile, and sometimes it forces the vendors to change something too even on same minor version. That’s a bit wacked altho I agree that it also forces them to open the drivers but that’s more of a political than technical argument. (technical is what you said about old binary drivers with bugs, with which I partialy agree)
that Sun would soon have a sackfull of outdated and unmaintained binary drivers with unfixable and untraceable bugs. (e.g. drivers for ATI cards < Radeon 9500 have reached end of road; the last version is final)
Yes, just like Windows and OS X, oh wait, that hasn’t happened to the majority of users…
It would also mean Sun’sr hands would be tied if they found they could improve performance and/or add features and/or cure bugs by slightly or not-so-slightly changing the abi. It would make it a lot harder to add in good stuff such as XGL/AIGLX, or FUSE.
Wrong. Read the many discussions from SUN engineers on this on opensolaris.org, like others, they realise that there are still ways to innovate. It’s also still possible to have new driver APIs, just like they’re doing now.
Finally, it would also mean it would come with no drivers out-of-the-box, and installers would have to endure an XP-like experience of fishing for every different driver in the net, then manually installing them (and forcing a hardware reset every time? 😉
Also wrong. Solaris has come with binary drivers out of the box for years, hasn’t stopped it so far. Same with OS X, Windows, etc. I’d prefer manufacturer supported drivers over the current no-drivers-at-all-for-many-of-them situation.
Then, in the case of Google, think that the hardware they need supported is stuff like raid SCSI controllers and network cards which are very well supported by the current working mode in Linux.
This I can agree with. If Google is thinking about switching, it isn’t going to be for hardware support. It’s going to be because of Solaris’ other features, such as greatly advanced security, DTrace, ZFS, etc.
Yes, just like windows and OSX.
Oh, wait, there is not a terribly big lot of drivers available for OSX, so it is not easy to compare …
Oh, wait, there is not a terrybly big lot of drivers for Solaris either … It must be because Solaris has only appeared on the market very recently. However, installation is very, very easy, because they provide a variety of drivers wide enough such that NOBODY is complaining about hardware support.
Oh, wait, I have this boxful of old hardware with shitty old drivers for windows 95 which don’t work well with XP. Mmmh, this scanner, this webcam, this graphics card and, this and this network cards, this audio card and this admittedly shitty pressure-sensitive digitizer must definitely all be illusions. Or it is the fact that their drivers are crap and abandoned that is a figment of my imagination.
Edited 2006-09-22 18:52
“…I’m not particularly fond of slowaris myself.”
I thought that moniker died along time ago? What in particular makes OpenSolaris, Solaris Express or Solaris 10 “slow” in performance?
Linux has been warking for decade to make sure your of-the-shelf hardware will work with it. This is biggest problem for any OS trying to replace Linux. I am yet to see OS coming anywhere close to Linux when it comes to HW support.
Do you think Google or any large Data Centre operator buys their hardware off the shelf? If they are doing anywork with Solaris, my bet is that they are buying Sun’s AMD servers or their T1 (Niagra) sparc boxes (lower power usage). In which case if their hardware is coming from SUN it’s a moot point that solaris doesn’t support as much offshelf hardware etc etc.
Do you think Google or any large Data Centre operator buys their hardware off the shelf?
Actually, if you’ve read any articles about Google, you’d know that their internal hardware group is almost as large as Dell, etc. for making servers. Most of their servers in fact, at last check, were sets of motherboards on anti-static pads hooked up to special power supplies with hard drives sitting on a tray in large racks that were wheeled around. Google’s reasoning behind this was that they assumed hardware would fail and that it needed to be cheap to replace. Whether or not this has changed, I don’t know.
Google doesn’t care about random, white box configurations, so it’s a moot issue for them.
Google mentions this in order to get cheaper support rates from Red Hat. I trust they have a business contract with Red Hat. Red Hat support is not cheap – it is more expensive than Solaris support.
I honestly doubt that Google is seriously considering switching to Solaris. It’s just a way to get a better price from Red Hat
I have my doubts they’re paying redhat anything. chances are, like yahoo’s ‘freebsd’ servers, there’s little relation to what they’re using and what the redhat provides other than that they use it as a base for their own customized in-house distro (we do the same thing here for the most part.) I would find it extremely odd that a company with the technical prowess of google (and with the degree of in-house secrecy that they have) would be dependant on an outside company like redhat for support needs in such a vast network as their’s.
Google paying RedHat for support is like Linus paying the kid that just discovered Linux 3 days ago, and now passing out Ubuntu CDs, for support.
I was always very dissapointed.
– The installer is a joke – every Linux installer is better
– hardware support is ok but cannot be compared even remotely to Linux
– GUI is that old and crappy window manager. Gnome is there, too. But it’s only second choice.
– Updates / patching is a nightmare – if you come from Debian it’s no option at all.
– Solaris userland is a joke compared to GNU userland: They use old and totally outdated versions of programs.
Funny thiung is: OpenSolaris always claims to be “better” and “more advanced” than Linux. In reality they copy quite a few things. There are a couple of useful things: zfs, dtrace, the kernel. Apart from that I don’t see much that really takes off. I wouldn’t dream about changing my servers from Debian to OpenSolaris – don’t know about Red Hat servers, but would be surprised if more customers of Red Hat followed that example.
– The installer is a joke – every Linux installer is better
Ever tried gentoo’s or slackware’s?
– hardware support is ok but cannot be compared even remotely to Linux
Solaris isn’t trying to be Linux. There is a set amount of hardware supported. I personally hope it never tries to be what Linux is hardware wise. It should be better
– GUI is that old and crappy window manager. Gnome is there, too. But it’s only second choice.
Uh, no. GNOME is the first choice. CDE is the second choice, unless you’re talking about the literal selection dialog. Do you have any idea the amount of resources SUN has spent on GNOME? I thought not.
CDE is the default choice because most of SUN’s customers are used to UNIX workstations where they expect a real and positively stable work environment. While CDE is old and ugly to some people, the stability and low resource usage it has is far more important.
– Updates / patching is a nightmare – if you come from Debian it’s no option at all.
It is? Looks at GUI tool in his taskbar. Sorry, what were you saying about a nightmare? At last check, I just click a button and then tell it to update. No nightmares here…
– Solaris userland is a joke compared to GNU userland: They use old and totally outdated versions of programs.
Because it’s UNIX, not GNU/Linux. Secondly, most of the GNU userland is available under /opt/sfw/bin if you install the companion CD or even if you just use what comes with it. Last, it’s GNU userland is still way more recent than the last stable release of Debian
You are also aware that to be certified as UNIX, and to meet compatability standards that their customers demand that those “old and outdated” programs are *required* to be part of the environment? Since you’re not one of their customers (so I’m guessing) it’s obvious as to why you don’t understand.
In reality they copy quite a few things
Don’t suppose you’ll name what those are? You are aware that Solaris is UNIX, and therefore it’s Linux that is coyping Solaris, not the other way around?
Edited 2006-09-22 22:34
> The installer is a joke – every Linux installer is better
I have never been a fan of the Interactive Solaris installer, but I have just been through the process of installing all the latest Linux distros, and in comparison the Solaris installer was on par with most distros. It was just was not eye candy.
> hardware support is ok but cannot be compared even remotely to Linux
Hardware support in Solaris is getting much much better. It take a while to cater for every wierd and wonderful device that Linux does. If you are purchasing a server for a business, good sysadmins would actually look for a “certified” platform so they can get support. Last time I looked, Solaris had the most “certified” platforms.
> GUI is that old and crappy window manager. Gnome is there, too. But it’s only second choice.
In Solaris 10, gnome is an equal choice.
> Solaris userland is a joke compared to GNU userland: They use old and totally outdated versions of programs.
The outdating you speak of is true, but that is changing.
> There are a couple of useful things: zfs, dtrace, the kernel. Apart from that I don’t see much that really takes off.
Hmmm lets look. off the top of my head Solaris has
– A java that works! Not a half baked copy.
– Live upgrade, and Flash
– Zones
– Resource controls. CPU, memory, IPC. (network soon)
– Fair share scheduler
– dynamic reconfiguration
– CPU pools
– Predictive self healing (software and hardware)
– Auditing (BART)
– MPxIO
– In kernel web cache (NCA)
etc etc etc, I could keep going, but thats enough
For a company like google Sun’s Niagara processor is an ideal choice. Think of the amount of money google must pay for power, cooling and floor space. Running on hardware like Niagara would save them a huge chunk of change.
I suspect this was one of the main reasons sun was so eager to get Linux running on Niagara. So that a company like google would be willing to run their app on their hardware with minimal work.
While I would guess google is looking at what it would take to run their brand of linux on Niagara they would probably also look at what it would take to port their app from Linux to OpenSolaris and be better able to take advantage of new features released into OpenSolaris.