In an effort to spur adoption of Solaris, Sun has begun a project code-named Indiana to try to give its operating system some of the trappings of Linux. The project is one of the items on the to-do list of Ian Murdock, founder of the Debian version of Linux and, as of March, Sun’s chief operating systems officer. Though he wouldn’t confirm the name of the project, Murdock – who’s from Indiana – discussed the project’s essence at the JavaOne conference here Monday, and Sun spokesman Russ Castronovo confirmed the name.
The interesting thing is that this is the inverse of the balkanization that happened to the UNIX space during the 80’s and early 90’s.
Could this bring an end to the title of Solaris or AIX administrator, and create the “GNU Administrator”?.
And it’s also about time — the thing that made Solaris boxes annoying to work with was their userland. The kernel is largely transparent to even most admins — this is a smart move.
Edited 2007-05-09 22:13
I think Solaris seriously needs a good package management system and it’s relieving to hear Sun finally acknowledges this.
However trying to be like Linux, esp. in userland, should not be the ultimate goal for every UNIX. There are many Linux distributions serving that purpose already. BSDs are not based on GNU userland and could be argued to be preferable in many aspects. Similarly, Solaris must have its own creed not follow others.
I wish Solaris should continue pushing what it does best: provide top end features (such as ZFS) and scalability and at the same time bring smart+functional+fresh approach to issues at hand including package management and userland.
Edited 2007-05-09 22:31
“finally?” Sun engineers were working on improved package management, instalers, etc. years ago, and still are.
Also, it is important to be careful when talking about package management what part you are talking about.
For example, a Linux distribution has a “package manager” and then various frontends that go on top of that.
A debian distribution might have dpkg (the package manager), apt-get (a special meta-manager for dpkg), and synaptic (a graphical manager that uses apt-get and dpkg?).
Sun already has a good basic package management system. What they lack is something like apt-get and synaptic.
Edited 2007-05-09 22:33 UTC
“Sun engineers were working on improved package management, instalers, etc. years ago, and still are.”
As well they should. Compared to tools like apt-get, or even pup, pkg* tools are clunky. More than once I’ve had to go in manually and clean house, remove dozens of packages still in the package manager with no installed software.
I’ve never had any problems with the current Solaris package manager. To be honest, I find it easier to locate packages for Solaris than I do for say Red Hat. In Linux it seems there too many options and different developers will use different ones causing me to have to traverse between package managers to get something installed that requires dependencies.
No, unless you are talking about Nexenta and even then, still not.
Solaris will always have technologies that are unique to it and a different way of doing some things.
The closest you’ll get is Nexenta, which could possibly be called “GNU/Solaris” though that moniker isn’t wholly accurate either.
About time?
Sigh.
There are several inaccuracies in what this article implies.
First of all, an improved installer, hardware support, packaging, and many other items were already underway almost two years ago. Some of the work didn’t start until the past year, but it was well under way before Ian ever joined Sun.
It is important that people give credit where it is due.
Since no specific details have been given about this project yet, only wild speculation is available.
Also, Solaris’ userland is not nearly as bad as people make it out to be. More accurately, it is merely different, and in some cases, fully POSIX compliant while the GNU userland that most Linux distributions favour is not.
I basically had never used Solaris until the end of 2005. Yet, I had little trouble adjusting to it and I’ve been using Linux since 1996.
I just don’t buy the arguments some people make about the differences between the two.
I’d imagine that a lot of Solaris boxes out there are running GNU utilities. Stuff like Solaris tar and friends may provide some level of POSIX compliance, but most users have spoken: the GNU replacements “set the bar” for usability. (After putting up with the Solaris tools for long enough, I eventually switched to the GNU tools in order to minimise the pure inconvenience of some of the dated Solaris offerings.)
I hate the GNU tools from a CLI usability standpoint. I really like some of the GNU tools for their capabilities. I hate some of the Solaris tools for their lack of capabilities. I really like the Solaris tools from a CLI usability standpoint.
Many GNU tools are starting to force long-options (–blah), when a single character will do. Long-options don’t allow option clustering (-vld) and require significantly more typing. Many GNU tools are starting to break compatibility by disallowing options with any dashes (what the hell; why do I get an error when I use “ps -auxww” like I have for close to 20 years on UNIX?). Many GNU tools are starting to overload the functionality of commands, break the Unix simple tool model, and make it much more difficult to script complex tasks. There’s also no consistency in GNU tools options (not that there’s really a whole lot in Solaris either, but there is *more* consistency).
Well, maybe because you’re using Linux ? …
I takes me a couple of days to adjust when I switch from one to the other. You’re right that some Solaris bits are a little dated or buggy (esp. tar). However I don’t see much difference from a practical standpoint – once I go back into “Sun” mode or “Linux” mode.
I don’t think it’s about which userland or package manager is better. How many IT decisions are actually based on the technical merits and capabilities of the various options? It’s about skill sets, compatibility, and mindshare. That’s where Linux is winning contracts where Solaris might otherwise be successful.
Sun is admitting defeat, in a sense, but they are determined to do as they said in the article–to make a better Linux than Linux. And I’m positive that they’ll be able to at least shake up the *nix market with this approach. We’ll finally have two operating systems, Linux and Solaris, competing purely on technical merit because they look and feel nearly identical to the administrator and to the applications. This will be as good for Linux as it will be for Solaris. A platform needs dozens of developers working in userspace for every one kernel hacker in order to sustain itself. Both platforms can use the userspace consolidation.
A very Linux-like Solaris has obvious appeal. The three main deficiencies in the Linux kernel that keep it from penetrating demanding enterprise environments are: long-term binary compatibility, enterprise storage, and RAS. Solaris, ZFS, and DTrace directly address these challenges. Linux might not be able to afford to continue its tradition of evolving APIs and fluid ABIs. Linux might have to grow up and take itself much more seriously due to pressure from Solaris.
Either way, free software is moving forward. It’s turning into a question of what kind of development philosophy is the best match for the enterprise and which is better for consumers. Sun and IBM will be duking it out for free software kernel supremacy, especially in the enterprise space. I’m not taking sides…
“Linux and Solaris are cousins that stem from the same Unix heritage, if not from the same source code.” — I’m pretty sure this person isn’t sure what he’s talking about and is on the verge of using an expression to tell a fallacy.
The article is so poorly written that it is hard for me to read. Eweek usually produces good stuff, so here is a similar piece by them: http://www.eweek.com/article2/0,1895,2127276,00.asp
I noticed this as well. About the only real heritage they share is BSD, and given how much Sun has rewritten of their OS, little of that remains.
This article is definitely on the verge of pathetic.
With respect to the offending turn of phrase, “if not from the same source code”, it’s more likely that the author doesn’t fully understand the usage connotations: the veiled suggestion that, in this case, there might be some shared source code. They probably meant to say that there isn’t any source shared between them, of course, but something like “but not from the same source code” would have been less ambiguous.
“””
“”
“Linux and Solaris are cousins that stem from the same Unix heritage, if not from the same source code.” — I’m pretty sure this person isn’t sure what he’s talking about and is on the verge of using an expression to tell a fallacy.
“”
I noticed this as well. About the only real heritage they share is BSD, and given how much Sun has rewritten of their OS, little of that remains.
“””
Oh, come on. Once you clear out all the pedantic crap from people who want to debate whether Linux is unix or not, the two share a heritage. If it looks like a duck and quacks like a duck, it’s a duck. Linux may be a mallard, and Solaris may be a wood duck. But describing the article as being “on the verge of pathetic” is just silly.
I swear, you Solaris fanatics can be just as tiresome as our Linux fanatics.
Edited 2007-05-09 22:59
Sorry, but just because you don’t fully understand the language used and the facts in play doesn’t mean they don’t exist. Bad journalism is disgusting to me so I have to call this out.
I pulled one sentence which has two examples of why the article is bad. One sentence. There is a lot more bad content in there if you want to examine it, but I’ll give you the rundown on what I chose.
First he says that Linux and Solaris stem from the same Unix heritage. Solaris IS Unix. GNU/Linux is NOT Unix, it says so in the name. It is a copy of Unix, however. So Solaris is Unix, and Linux is a copy of it. Your pedantic crap is my semantic fact — what the author said is misleading at best.
Here is Linux on a Unix family tree, note that the lines never connect: http://upload.wikimedia.org/wikipedia/commons/7/77/Unix_history-sim…
Secondly – in the same sentence – he suggest that Solaris and Unix share source code. That isn’t just wrong, it also shows a complete misunderstanding of the operating system space. Either that or the guy doesn’t know how to use the english language, and used an expression that means zig when he meant zag. In either case it makes for bad reading.
There are several other cases of idiocy in that article, and I’ll leave it up to you to find them. Or not!
“””
First he says that Linux and Solaris stem from the same Unix heritage. Solaris IS Unix. GNU/Linux is NOT Unix, it says so in the name.
“””
True, but totally irrelevant. All POSIX-like OSes share a heritage in a very real sense, despite the details of where the source came from.
In a very technical, hair-splitting sense, you may be correct. But that doesn’t really matter in the grand context.
It makes not a whit of difference to the meaning of the article.
Sun sees advantages to making Solaris easier to use and more familiar to that growing pool of admins, much larger now than the pool of existing Solaris admins, who are familiar with Linux and the utilities customarily found on Linux systems.
Sorry, but your “point”, which seems quite important and significant to you, looks like mere hand waiving to me.
Edited 2007-05-10 21:41
I don’t understand why Sun just doesn’t simply create its own Linux distro and incorporate portions of Solaris. I’m guessing that Sun probably doesn’t want to open source some parts of Solaris, and it would run up against resistance within the OSS community if it doesn’t; however, Sun Linux might make more sense than Sun Solaris with new Linux flavoring!
You guess wrong then. The only parts they haven’t opened are the parts they haven’t yet secured the rights to or guaranteed that they have the rights and resources to do so.
If you want “Sun Linux” go use what you find at http://www.nexenta.com. It would be pointless for Sun to have “yet another distribution.”
You guess wrong then. The only parts they haven’t opened are the parts they haven’t yet secured the rights to or guaranteed that they have the rights and resources to do so.
How does your comment contradict my original post?
I’m guessing that Sun probably doesn’t want to open source some parts of Solaris…
you want “Sun Linux” go use what you find at http://www.nexenta.com. It would be pointless for Sun to have “yet another distribution.”
Pointless? Hardly. Sun knows that Linux has more momentum than Solaris at the moment. So, it would make sense for Sun to introduce its own Linux distro, leverage the work of the community, and compete with Novell and Red Hat for the Linux market. In some ways, Sun has an advantage due to the fact that it can bundle its offerings with hardware.
Because it has nothing to with them not *wanting* to and everything with them not being *able* to. Solaris is a 20+ year old codebase. Some the companies code belongs to don’t exist anymore, and others have refused to relicense the rights.
Your claims are FUD. Stop spreading them.
You missed out on the Sun Linux several years ago. Sun tried that, and found that no one wanted a Sun Linux; they already had enough choice in the Linux space. What they wanted was Solaris, so Sun dropped their Linux distro.
Sun is open sourcing *all* of Solaris, as they get the IP rights straightened out. There’s already several million lines of code that have been open sourced, including all of the best features (ZFS, zones, Dtrace, SMF, and most of the kernel).
I think thad for Sun, dropping Solaris would be suicide. They invested plenty of money, it is great, matured product, in lots of areas more mature than Linux.
But it is great that they want to learn from Linux, and I believe that they can learn a lot. To be honest, installing and configuring Solaris for me wos a real pain. I found the userland different from GNU’s, and in most areas less intuitive. Maybe too many years of using GNU/Linux, but that is just the way it is – Linux gained power and lots of people, just like me, take for granted that commands have similar options.
On the other hand, Solaris has got so many unique features that simply dropping them and using only GNU userland would be pointless. So I hope that Sun will make it’s system easier for us, newcomers from Linux and as powerful as it has never been before at the same time.
I am really looking forward to seeing Solaris gain a little better userland than it previously had. I for one really love the Solaris kernel and really want to use it to do all my work on, however the userspace really keeps me on Linux. I’m really hoping Sun takes this in the right direction.
What really irks me about the Solaris userspace is the difference in command-line flags between the Sun utilitys and the GNU utilities. Having to memorized two different command-line option sets for the same command (3/4 if I am running the same thing on AIX or HP-UX) can get really old really quick.
Edited 2007-05-09 23:00
Remember that some of those differences in command-line options are due to POSIX requirements. GNU utilities have often thrown compatibility and standards compliance out the window in the name of convenience or carelessness.
Solaris should not make the same mistake.
Remember that some of those differences in command-line options are due to POSIX requirements. GNU utilities have often thrown compatibility and standards compliance out the window in the name of convenience or carelessness.
Good point. It is unfortunate, though, that people have to suffer (okay, not really “suffering”
) because of the “convenience”/carelessness of developers who wished to change things without regard to POSIX.
Hopefully you are right, I hope Solaris doesn’t make the same mistake (but still adds all the nice GNU features!)
“””
Remember that some of those differences in command-line options are due to POSIX requirements. GNU utilities have often thrown compatibility and standards compliance out the window in the name of convenience or carelessness.
“””
I’m not sure how relevant POSIX requirements are in this case.
Looks to me like GNU utilities *are* the standard today. That’s why Sun, Solaris, and project Indiana are chasing our tail lights^W^W^W^Wlooking to be more similar to GNU utilities today.
Linux and GNU are unifying unix like POSIX and SUS were supposed to do but never did.
Solaris has some really impressive strengths. But get used to playing catch up to Linux and GNU in some areas, because Sun neglected a lot of things while pursuing the lucrative large server market.
The old unix companies were always negotiating for standards, but they could never seem to stick to a deal for long enough to do any good before they started stabbing each other in the backs again.
I welcome Sun’s new initiatives and find the current climate refreshing.
Edited 2007-05-10 01:03
So I have read one of the older posix standard AND the single unix specification.
I’m curious where the GNU utils directly break posix compatibility for the sake of convenience.
While not disagreeing with you, it would be interesting to know.
I can’t tell you where they’ve broken compatibility, but they have added features to their user space for the sake of convenience – take backtrace() for instance, which is a culmination of a couple of other calls put into one to make something commonly used, alot easier.
The problem is not necessary these additions to make life easier, the problem is when programmers start using these and claiming that they are the ‘standard’ over and above those set by multi-party organisations.
That is what we see even today; people Linux-ising their applications rather than aiming for specifications – the net result? we have situations of various operating systems like FreeBSD/OpenSolaris/NetBSD/OpenBSD and numerous others having to go to hell and back to get things compiling and working properly on their said platform, which don’t implement linuxisms.
Edited 2007-05-10 05:27
While you give a good example of lazy programmers who can’t read specs, you didn’t really answer my question
Adding to does *not* break a standard. I was talking about userspace tools in specific, not about the c library.
You shouldn’t say linuxisms, you should say gnuisms. Not that I EVER call Linux GNU/Linux because that is stupid, but because the problems are GNU specific and not Linux specific. A good example of that would be the Linux kernel’s syslog() function and the libc syslog() function which do completely different things.
I’ve had a look – best example of this; GNU diff vs. the diff which is included with Solaris – if you look through the discussion over on ARC at opensolaris:
http://www.opensolaris.org/os/community/arc/caselog/2007/189/
Goes into detail over the differences. Like I said, nothing wrong with extending/building on the functionality – that is the reason why so many specifications are very vague in some areas; to allow vendor differentiation in their implementation as to add value.
The problem starts to occur when the standard isn’t even implemented properly (completely) and instead, a completely new way is done.
Thanks for the link, mod +1
Software from Sun:
Java
Solaris
StarOffice
Software from Linux/GNU:
Python
Gnome – KDE
Kernel 2.6 ported to all arch’s
(KOffice – to be)
Synaptic – Aptitude
There is a reason why they would like it to be more like Linux.
And it’s self explanatory.
Why would one change to Solaris ? Just because it’s Solaris ? Sure, if you are on a supercomputer network but that’s (about) a niche of the market.
(Supercomputer sites powered by Sun systems represent 30 percent of the top 50 sites in the category)
http://www.gridtoday.com/02/0617/100028.html
_____________________________________
Linux and Solaris are cousins that stem from the same Unix heritage, if not from the same source code. <But Linux fans simply have a hard time trying Solaris, Murdock said Tuesday.
Don’t blame the site ! – blame the Sun Representative (Mr. Murdock)
I love it, but I hate it.
Personally, I don’t find Solaris’s patch/package management system to be very “fun”. It works, but it seems very dated. dpkg is a fairly nice tool, and debs are quite easy to work with.
I used to be one of the folks who complained about the old outdated packages available for Solaris – but now I’ve come to realize I much prefer building my own. I just like having nice tools to manage my packages, and I think debs (or some new version) would be a great successor to pkg* tools, in this regard. I hope they hit the patch management tools, too. Maybe kill two birds with one stone.
As to the userland, I’m not a big fan of the Solaris userland. I don’t mind the difference in flags/etc too much (except for the build problems with software such as vlc.) I realize those issues are portability problems with the software itself, but face it – a lot of the industry is moving to GNU userland based OSs.
Now, what I would absolutely love to see happen in the Solaris userland is functionality added, even if it has different flags. For instance, I don’t like having to shell script a recursive grep. Why can’t I just grep -R? Various little niggles like this irritate me. That’s not GNU-centric thought, it’s “lack of commonly used functionality” irritation.
With ingenious tools like zfs/zfs cli tools in Solaris, I see no reason why the userland should be full of functionally lacking software. Obviously Sun is capable of producing it. If moving to a GNU userland will expedite adding in such functionality, I’m all for it. POSIX compatibility is nice and all – but face it – GNU userland is becoming the “standard”, quite quickly.
Maybe Sun should investigate adding the POSIX functionality into the GNU toolchain, adding their optimizations from Studio 11 compilers/tools into the GNU compilers/tools, and then blending it into Solaris while tossing out the old cruft one by one. Then they’d have the best of both worlds. Quite a feat, but easier than starting from scratch.
To improve solaris acceptance; sun should allow anyone download their software without registration first of all.
Then they should create a ubuntu like packages installer, that solves dependancies issues, and to simplify installing all free applications on x86 specially.
And to include more advanced management tools and attach shortcuts for them on the start menu; and then to add more network adapters drivers; and to support networked printers properly through cups as my brother printer was installable on all linuxes and it was wireless, with no luck on solaris as of beta 60.
And finally good lcuk for Sun, and linux should learn from sun how fast startup time should be.
They need to register as there are still proprietary code within Solaris which they’ve licenced off third parties – as you can see, you can download, without registration, the source code, binary builds and closed source builds off the dlc.sun.com/osol website.
Yes, I know it is annoying, but at the same time, once you’ve setup a registration, it gives you access to the download, the Solaris forum, and a number of other online Sun services – so please, don’t turn something akin to a mole hill into a mountain.
Pardon? sure, you have to use pkg_add, but at the same time, there is nothing stopping you from writing a gtk front end to pkgtools – its an opensource project, and I’m sure Sun is working on these issues, but at the same time, they have limited resources, and must focus on the things that their customers need now rather than nicesities.
All this is available under Administration Tools in the ‘Java Cup’ menu; again, there is nothing stopping – yes, again, although these are short comings, these are being addressed; like I said earlier, Sun has limited resources, if you so require it within a timely manner, you could take the initiative yourself.
Regarding “brother printer and wireless” – thats all very nice, complain about the lack of hardware support and never actually tell the forum what hardware that they’re having problems. What chipset wireless card? what model and brand of printer? what attempts have you made to fix these issues? in the case of the printer, did you try to compile the latest Guntenprint?
True, but all this takes time – better to take time to think about the things rather than hastfully running into the problem without thinking.
It may result in ‘longer times’ but in the case of Solaris, when things are done, they’re implemented fully – take wireless; WPA has been added to the wireless stack, and now all wireless drivers that are in the distribution have wpa support out of the box.
” what model and brand of printer”
I use MFC-8870DW and I was able to connect it through RHEL4 and 5 by the use of CUPS and postscript driver or CPL3 over TCP/IP address 192.168.x.x. lpd://192.168.x.x/lp
In Solaris I was able to use lp at the command line to activate the use of the printer but when I initiate a print job from firefox or opera or any GUI based application it will not see the printer at all. It is a shortcomming of Solaris GUI print tool.
Going to the support forums I read that networked printers are supported only through manufacturer supplied drivers and nothing else should be attempted!
I understand that there are priorities in a tight budget huge project like solaris but I love solaris and I wish they speed up things to let us use the best UNIX out there to replace Linuxes which till now suffers huge instabilities and ignorace of compatibility.
Again Good Luck for Sun
It doesn’t sound like a short coming of the tool, but a bug in OpenSolaris – have you made a bug submission for this? I hope you did, or otherwise any legitimacy in your argument dies with this post.
What was the wireless card? for the printer, like I said, submit a bug report; if you can print to it via command line by not through a gui application, there is obvious some sort of bug there.
Where could I find a list of _currently_ supported wireless chipsets and wpa/wep support availability ?
Atheos, Intel 2100/2200/2915 and recently added in b64, 3945 – having had a look at the next Solaris Express Developer Edition, it appears it’ll be based on build 64/65; that’ll include basically support for all of Intel’s wireless chipset; 4265 although it has been released and being included with some laptops – personally, I’d stay clear of it – read up on arstechnica about the conjob that is 802.11n, and how it won’t/can’t reach full speed.
In regards to a previous, unrelated post; it isn’t “beta 64” – OpenSolaris isn’t in beta, it is a project in progress; parts are only merged once they’ve been adequately tested – once done, then merged; OpenSolaris is the equivilant to the respins of Red Hat – constantly being updated – a rolling release schedule if you will.
Trying to be a better Linux than Linux will not work, to much of the work to get better than Linux will spill over in Linux and keeping the distance between them more or less constant or at least narrowing very slowly. In the mean time Linux will gain more mindshare.
The only way to get Solaris out of the shadow would be if Sun started to work closer with Apple. They would make a perfect couple, Sun is strong on the heavy technical side and Apple is good at user interfaces and make things just work. Sun is good at server hardware, apple is good at desktops and laptops, and both of them are mostly hardware companies.
While I agree this would be very “cool” (and Apple seems to be “borrowing” from Sun) – I’m not sure a partnership will occur. I sure would love to see one, though. Using Solaris as the underlying OS beneath the OSX GUI would make OSX one heck of a beast. It’d solve all the server-side problems they’re having in terms of performance, that’s for sure.
Unfortunately, I don’t think I’ll live to see the day of “Snapple”.
There’s always Mac OS XI (11). They could go fully 64-bit, adopt full virtualization capability and adopt the open-source Solaris kernel and userland plus a new Aqau desktop interface. They could also probably keep compatibility with at least Mac OS 10.X Intel tech and software or just fully virtualize a Mac OS X environment within Mac OS 11 plus maybe even Mac OS 9 if they really wanted to for some reason.
We can always dream of a Snapple OS:)
you wrote: The interesting thing is that this is the inverse of the balkanization that happened to the UNIX space during the 80’s and early 90’s.
I see your point, but I don’t know if I’d call this the ‘inverse of.. balkanization’. If anything, we’re seeing intense competition for both the server and desktop markets. A notable side-effect (that I’ve mentioned on other threads) is that there are now some amazing, production-quality offerings in both the BSD and GNU/Linux world. Sun perceives that it needs to put out a highly usable OSS OS and start attracting developers for it.
edit: [sorry – that was intended to be a reply to the very first post.]
Edited 2007-05-10 02:02
The real problem is what it takes to get up and running on the crappy Dell/HP laptop I/You/he/she/it develops on. If Sun wants to grab developer mindshare get Solaris to work on the machines they often develop on – laptops. I only seem to get Solaris running on my desktops – mostly because of wireless support and video. While I have wireless issues with Linux, it’s a good bit easier than Solaris. (Although some people are lucky and their model laptop works pretty much out of the box.) I’m not sure that making Solaris’s “ps” work exactly like Linux “ps” is going to send developers swarming into Solaris land. (Although it’s funny to watch a certain Linux guru I know that calls himself a Unix guru flounder helplessly on Solaris).
Wouldn’t it be nice to see a certified lenovo T61 laptop work with solaris 11
I mean right now you can order T60p laptop with certified ubuntu, RHEL and CentOS beside of course XP/Vista on a dual boot out of the box; but sadly you cannot buy it with solaris.
I still know about Tadpole laptops but they are solaris only laptops and I don’t like/hate them too.
I will keep dreaming!
We are moving to a Solaris on one of our server solutions. I was tapped to sys admin it.
I will have to agree with you regarding the package management. pkgadd -d gets a little old.
A `Linux` like package manager would help its adoption a little better.
Yeah, a good shell would be nice, yeah, a good package manager would be nice, still, bring us drivers from the Linux world, that would count more. I’d like to try a Solaris with Linux driver support and KDE4 one day.
Uhmm, lemme ask just one thing. What would really differenciate this new Solaris from let’s say Nexenta ? I’d pick Debian with zfs and dtrace before Solaris, if that would be an option.
From the recent Sun & Solaris stories I’ve understood that one of their main goals now is especially developing the Solaris package management.
Ian Murdock has often mentioned apt-get in Solaris interviews so it is quite possible if Solaris adopted apt-get soon too. The Solaris-based Nexenta already uses it.
On the other hand, another Solaris variant, Belenix prefers pkg-src from NetBSD which is a very good package management framework too, and in some ways more advanced than the older apt-get: http://www.netbsd.org/gallery/advocacy/sschumacher/a4-pkgsrc-en.pdf (pdf)
What ever they choose, I hope they don’t just copy others but rather learn from others and then try to do things even better (usability, security, stability, flexibility etc.). That way, by doing things even better, they might also win over more Linux users.
Edited 2007-05-10 09:16
This is a kind of thing that I hate in unix world: “windows-friendly”, “linux-like”, “whatever-like”.
Why this ? The target of all unix projects is look-like something else ? Make your target become something better.
And that’s where theses shortcomings stem from. Thank God SUN has now admitted that there are indeed shortcomings. It’s certainly not about the kernel (I wish Linux had a kernel such as Solaris), but about:
– a modern GNU userland (or comparable to GNU)
– a modern desktop
– a modern package system
…
Solaris has a lot of good and reliable things. And they’re short of other things, and so it’s a good thing they’re able to admit the latter and are open to learn from its younger cousin.
Is it so much that Solaris is hard to use because it is hard to use (which I don’t agree with), or because Linux users do not want to change what they are doing and expect everyone else to cater to their specific needs and wants?
Does changing the package management tools for example actually improve the package management process or does it just make it easier for Linux users because they are familiar with tools like apt-get?
If Sun is going to make “improvements” to Solaris, it should be the result of a process where there is demonstrated ease of administration or use benefits and not just adding Linux tools simply because Linux users want them.
If Sun is going to make “improvements” to Solaris, it should be the result of a process where there is demonstrated ease of administration or use benefits and not just adding Linux tools simply because Linux users want them.
Unless you’re trying to cater to Linux users, which is what Sun is trying to do. Sun isn’t trying to grab AIX users, Windows Server users, or even Free/Open/NetBSD users. They’re trying to get Linux users to convert to Solaris.
If making the Solaris environment easier to transition to, then Sun is smart to do it. And if it improves the Solaris environment, then good for Solaris.
I have read a number of comments about Linux users who find Solaris hard to use simply because they do not want to change what they are doing. For Sun to make those kind of changes so that Linux users don’t have to change in my opinion is wrong.
It would be no different than me asking RedHat or Novell to make Linux more Solaris like because I don’t want to change how I do things. Would those changes benefit RedHat and Novell users or just make it easier for me? I think it is a case of more the latter than the former.
Every operating system is different, and as somone who has used AIX, HP-UX, Linux and Solaris, I do not expect the OS vendor to make changes to major functionality just because a handful of users for whatever reason can’t (or won’t) use the tools provided by the vendor. It is a choice and I really don’t think that turning Solaris into another Linux distro is going to improve either its functionality or make Linux users embrace it much more than they already have or have not.
I have read a number of comments about Linux users who find Solaris hard to use simply because they do not want to change what they are doing. For Sun to make those kind of changes so that Linux users don’t have to change in my opinion is wrong.
I’m have no doubt that is part of the motivation, to make things easier for Linux users.
But why is that wrong? Sun is actively and fanatically courting Linux users. Why wouldn’t they expend some effort to cater their product to their intended demographic? If they want more users, they’re going to have to take them from Linux.
I have no doubt either, that Sun is doing this to just make Solaris better. I like Solaris too, but it is a bit crufty. Just because the ideas didn’t originate from Sun itself, doesn’t mean it’s not a good idea.
It doesn’t matter where the ideas come from, I am all for improving Solaris. But I am not convinced that changing package management and adding the GNU userland is going to make that much of a difference to a lot of people. Will these changes make it easier for me to:
1. Administer multiple systems
2. Build and package software for distribution
3. Consolidate logs and audit trails from multiple machines
4. Provision systems faster and provide more flexible building options
Most of what I just mentioned is either already a part of Solaris or is in active development. I use Solaris as both a server and a desktop and for the vast majority of the tasks I am expected to perform, Solaris does it without much in the way of additional software.
Most of the driving force behind the Linux feature push is desktop Linux users who feel that Solaris is to use your phrase “crufty”. I somehow don’t think that Sun is going to get a whole lot of mileage out of people who want everything for free and do not want to contribute in any way to the effort.
What part of Solaris is crufty?
1. Administer multiple systems
2. Build and package software for distribution
3. Consolidate logs and audit trails from multiple machines
4. Provision systems faster and provide more flexible building options
Those are all things you and I both do with Solaris systems. That’s a traditional, orthodox workload for Solaris. Not that there’s anything wrong with it. It’s a proud tradition hailing back to the days when you’d punch a baby seal in the face to get a Sun Ultra 5 workstation.
But that’s not all that Sun is going for. They want more of the market, they want to be on the desktop, they want to be in way more servers than they are now, they want people to build appliances based on Solaris, and hobbiests to setup Solaris workstations instead of Linux workstations. Anywhere you see Linux in a headline, press release, or “based on” label, they want it to say Solaris/OpenSolaris.
What’s crufty? Package management. Adequate, but not as convenient as say apt-get. What about a Sun maintained supported repository, where you just typed “apt-get” to install all the latest patches, or to install a piece here or there.
Like many Solaris users, I’m very particular about my work environment. The first thing I do when I install Solaris is to start replacing pieces. I put in a decent shell, I put in GCC, OpenSSH, either compiled myself or more likely one downloaded from sunfreeware.
The desktop. If you’re doing server administration, all you need is olwm and some xterminals. But most people do a lot more, and enjoy a nice looking desktop with apps that aren’t a few years old. You can add them by hand of course, but that’s a hassle.
It’s all adequate, but there’s a better way to do it. I like convenience. There was a time when I loved working in the crevices and crannies of Solaris and other operating systems, doing hand compiles of all my applications, but now I’d rather just have a smooth running system.
You may not want Linux interlopers in your Solaris environment, but Sun sure does.
What I don’t understand is where everybody gets that Sun wants to be in the desktop market? A corporate desktop yes, an individual user’s desktop, I doubt it.
My desktop on my Sun Ray is JDS 3 using Solaris 10 6/06, it looks fine and allows me to administer Solaris and Windows (through uttsc). I am limited more by the Sun Ray than Solaris as to what I can or cannot do.
And what is wrong with adding software yourself? What is the point of using an “advanced” OS if you are not going to do “advanced” things like compile software you want? If I want the latest version of KDE, I have no problems in pulling down the source and compiling it myself. When I used Linux “back in the day” the only way you could add software was to compile it yourself, this is where I learned how to use a compiler and associated tools. I just don’t understand why so many people depend on repositories of software and want the latest GNU userland if they are not going to use it!
And what is wrong with adding software yourself? What is the point of using an “advanced” OS if you are not going to do “advanced” things like compile software you want? If I want the latest version of KDE, I have no problems in pulling down the source and compiling it myself. When I used Linux “back in the day” the only way you could add software was to compile it yourself, this is where I learned how to use a compiler and associated tools. I just don’t understand why so many people depend on repositories of software and want the latest GNU userland if they are not going to use it!
There’s nothing wrong with it, but there’s nothing wrong with leveraging what the vendor does so you don’t have to. Could I download KDE and compile it myself? Sure. Do I want to? No. There was a time when I did, and it was useful in the knowledge I gained from it (getting KDE 1.0 to compile on SunOS 4.1.3 was quite, er, enjoyable), but I no longer care to do it anymore. If the end result is getting KDE, and spending either 5 minutes to do the apt-get install, or longer to wait for it to compile and possibly hit Google up when the compilation breaks for whatever reason, I’ll take the option that requires the least time and energy.
The operating system in an enterprise environment is there to serve as a means to an end. If it’s for a hobby, then yeah, I’ll compile my own stuff. But if it’s for work, then I’ll spend the least time required to fully accomplish the task, so that my time and energies are better spent elsewhere, on other projects, more servers, other crisis.
There’s nothing wrong with compiling it yourself, like there’s nothing wrong with using sh instead of bash, or vi instead of vim. It’s just preference and time savings.
In my case I have an unfair advantage compared to most system administrators. Since we are replacing most of our equipment I have the opportunity to choose the OS we install. After some testing we chose Solaris 10 6/06 as the base OS for our machines. There is very little in the way of additional software we had to add to compile most of the software we use daily, there are notable exceptions like Fedora Directory Server, but for the most part Solaris 10 has what we need.
We also have test machines which gives us the opportunity to “blow things up” without affecting production machines. And again, Solaris 10 delivers what we need based on a Full Distribution installation.
There is a lot of discussion in the OpenSolaris forums about package management, and one idea that has surfaced that shows potential is using pkg-get from Blastwave in conjunction with a GUI and a Sun approved repository. Whether this happens or not is yet to be seen but the approach is sound. Instead of tossing in a bunch of tools many people might not be familiar with and increase the learning curve, install tools that people know and use and extend their capabilities. I for one don’t use Debian, and while it might be easy to use their package tools, I cannot speak for everyone and their skill and comfort levels.
Each situation is different and calls for different tools, in my case Solaris 10 “out of the box” delivers what I need. Your case is obviously different.
Using SunOS, that brings back some bad memories!
I think this really shows the whole Unix is Unix comment!
The GNU stuff is what I used years ago on BeOS, and its OK.
The Solaris stuff is fine, I used it 12 years ago and still do.
I’m learning AIX stuff too.
Sure its annoying doing ps -ef or ps -A to try and get the same thing, but no more annoying than switching between ksh and bash. These are all uncomfortable things that are part of life.
Sure, I’d love AIX to have df -h, but it doesn’t. I’d love Solaris network configuration to be easier (its coming).
Most of the arguments, not just on this site but on /. and the rest of net, seem to be people (read linux users) moaning about Solaris being a crap desktop – which it is, but thats because its a server OS you moron. Although thats not half as dumb as the calls to merge Solaris onto the Linux kernel. Its the linux kernel thats the problem. Its too ‘generalised’. From a BeOS angle we poked fun because its latency was poor, and from a server angle we poked fun at it as it was too fickle to be truly performant. Theres been a lot of work in 2.6 since then, and things are getting better, but a kernel cannot be everything at once. Linux simply is not as scalable as Solaris, or any BSD based system as they’ve been scaling from before Linux even compiled. Any sys admin who tarts around the *nix scene knows this.
Sure, Red Hat, Unisys, and SGI did some great work, but hey, Linux 2006 is Sequent 1998. For those who don’t know about IBM’s lobotomy of genius go here http://en.wikipedia.org/wiki/Sequent_Computer_Systems
Linux makes a reasonable desktop these days, and a passable server. BSD/AIX/Solaris are pants desktop OS’s, but great servers. See where I’m going?
Perhaps we will see a desktop edition (Open?)Solaris, with ‘userland’ personalities coming. Theres a blog where I think J"org Schilling does a simple one liner to switch from GNU userland to Solaris and back. Superb idea. As most people have already said, its already there as Nexenta.
Sorry, that should read Sun Sparc 5, not Sun Ultra 5. No one would punch anything in the face for an Ultra 5.
I had a SPARC 5 at home, along with a 20 with Ross HyperSPARC processors. The 5 was OK, but I had real problems driving a HME card and a 24-bit frame buffer on it. Dumped them both when Sun introduced Software Express (now Solaris 10).
I remember that it cheaper to buy a SPARC 5 on eBay than it was to buy an Ultra 5. The horsepower gain wasn’t that great for the money people were spending.
While I dont want solaris to become linux I do want the tools with the same name to do the same thing.
bash#fdisk -l
No sorry I dont know how to list disks…
Small things like that annoy me. format is a useless tool when trying to find out what disk slices you have.
ZFS DTRACE… on linux so solaris 10 is going to be yald (yet another linux distro)
Personaly why isnt there a java like platform independent layer for driver developers. To add extra functionality OS makers can plug code under it like dladm (nemo project) etc..
I have a idea why not just make solaris 10 public domain and have 100000000 solaris variants pop up over night? So every os has DTRACE even microsoft…
SUN is a public company and has extensive IP rights…
Linux took many years to get to where it is now do sun realy think they can build a big community again overnight just with GPL???
I do however feel sun need to rejuvinate the geeks and release a Netra X1 style sparc as we all know there is only one linux distribution that can run on sparc and even that is lacking support. (excluding T1000’s)
How many people would buy a sub 1k sparc?
I’d love to try it out, but AMD/ATI has decided I shouldn’t…
To be exact, Solaris is Unix. It is one of the few that can say that and be correct.
The only problem with Solaris is that the common home users or tinkerer has not tried it lately. Download the newest release and learn up a bit on the few oddities and there is little reason to consider a linux.
POSIXLY_CORRECT=1 ; export POSIXLY_CORRECT
And voilà!! You can make the GNU tools POSIX compliant quite quickly.