Is the Linux community afraid of Opensolaris?
Okay,okay … i know the headline is a little bit provoking. But when you think about some comments from Linux proponents you could think so. In the last few weeks i´ve heard one sentence quite often: “Why you you still develop Solaris? You should contribute to Linux!” from people administering Linux systems. And you could read at other places, that Solaris is irrelevant, that there is nothing worth of mentioning it or even for an integration to Linux. Just think about the Zemlin quotations! Or several other comments of proponents of Linux.
This is an interesting development. In the years before, there wasn´t such comments. Solaris was considered as a dead end. But then the game changed. We open-sourced Solaris. The full monty over the time. We open-sourced the cluster framework. And we won´t stop to open source further code until there is no more code to open-source. BTW: I find “Sun should contribute more” really interesting. In the moment you start up your text processor on your favourite Linux distribution you´ve gone through more code contributed by Sun than of anybody else. You´ve already traversed a large amount of code contributed by Sun when you just login into GNOME. This is a fact most people tend to ignore. It´s really interesting, that i find more and more articles that shows Solaris in a positive light. The reports on heise.de and especially the comments in forum are getting more positive for Sun and Solaris. In public forums you find more and more articles regarding Opensolaris. I would think about it as a good indicator for an increased uptake of Opensolaris in the market. Interestingly there are people who think about orchestration, when there is an increasing share of Solaris discussions on other forums, not believing that there is an increased interest in Opensolaris like this article. And this is one of my important question: Why is there an increased amount of articles and comments disavowing the relevance of Solaris? Why are there no comments about BSD? Why are there no comments regarding AIX or HPUX? What´s so special about Solaris? Or to ask the provocative question from the headline: Is the Linux community afraid of Opensolaris? Why is the Linux community afraid of Opensolaris? At first: There is an easy answer to the question asked in the mentioned article regarding Mysql and Opensolaris and this fear of some people regarding Mysql and Linux. It´s a matter of fact, that most Mysql installations run on Linux. There is more money to make with Mysql on Linux with Mysql on Solaris. And Sun has to make money. Just tell me one reason, why we should limit our money making oportunities by limiting Mysql to Solaris. And when there is a large effort for Mysql on Opensolaris, it´s just because of leveling the playfield, as the development of Mysql was a linux centric one in the past (e.g. the scalability of Mysql is somewhat limited to the scalability of Linux respectively to the common sizes of systems used for Linux) Of course ZFS is important for Solaris to stay relevant. Dtrace is important to stay relevant. An operating environment needs posterboy features to attract new users. An unix is an unix is an unix. Just doing another flavour of Unix isn´t an incentive for someone to change to another unix. Of course other features are interesting as well, but they won´t attract users. I once coined the phrase in a customer presentation:”There are cool features, and there are important features. There are features to attact new users. And there are features to keep users on your plattform”. Binary compatibility is a cool feature, but it´s a feature you learn to love at the next migration to a new mayor release of your operating system. This is a feature that keeps you at Solaris, but it won´t really attract you to change the plattform. By the way: Don´t underestimate the importance of Dtrace. Of course it´s not a feature for the every day user of Solaris. It doesn´t have to be one. I find the lack of fantasy in the Linux community in regard the role of Dtrace a little bit astounding. When a developer recognizes the usefulness of dtrace, she or he will use Solaris, when she or he uses Solaris, more software will optimized or developed at all for Solaris (it begins with decent makefiles, goes to support for compiles other than gcc and ends with the usage of nice features of Solaris making the developers life easier). And more software running well on Solaris leads to a bigger user community and bigger mind share. So, why does Sun doesn´t contribute in large scale to the Linux kernel (we contribute a lot of code to Linux as an operating environment as i have stated before)? I´m sure it would help Linux. The Solaris engineering is one of the best or the best in the world of operating systems. But such a move wouldn´t really help Sun. Linux and Solaris have pretty much different design principles. Linux and Solaris have pretty much different targets of systems. And the design prinicples of Solaris are imporant to Sun: Binary Compatibility. Scalability - a quadsocket Victoria Falls systems is essentially an 256-way SMP problem. Maintainablility. How to update a running system while retaining a running version of a system. You can´t sell high-end highavailable everything-redundant servers, without having an operating system that is capable of certain mechanisms that allow updates in the shortest possible time (and no, i don´t consider the in-situ updating process of Linux distributions as a good way to update packages. I really want Live Upgrade or Snap Upgrade before using it on a system in the M4000 and beyond range). Even the development process is differently enough to justify the development of Solaris. It´s the concept of the Architecture Review Comitees, the existence of a Solaris Sustaining Engineering. It´s important to have such entities. Sun needs such entities to satisfy the needs of our customers. The customers expect patches for older releases, people just goaled to fix bugs and not to develop new features. It´s one of the non-technical features people like. And we can´t simply discard it, just to start developing at Linux. And i´m sure Linus and Linux doesn´t want to take over the development concepts of Solaris. You don´t think, that´s such an big issue: Even top contributors to Linux like Andrew Morton thinks, that Linux has it´s own share of problems: (here(2005), here(2006),here(2007) and here(2008)). It´s a reoccuring theme in every year. Problems that may lead to the need to restart the Linux franchise in the distant future. But that´s a different story. In my opinion, Linux is a good-enough Unix as x86 is a good-enough processor architecture. You can solve a good amount of problems. At some customers 100% (when you have hundred thousands of nodes, even good-enough is more than enough), at other the share-of-wallet for Linux is 0%. There is a reason, why we sell M9000/32 systems in good amounts despite of all sayings that you could solve all problems with a bunch of x86 servers … harharhar. You shouldn´t discard this need of computing in the area of “better than good enough” as a small niche. You have to consider, that both environments have pretty much different customers at their respective extremes. There are people using Linux on WLAN routers and there are people running SAP central instances on a full-blown SPARC Enterprise M9000. Albeit the desktop experience of Solaris got better in the last versions, but it´s a server operating system by it´s DNA. Linux is something between a desktop and a small server operating system. Okay … i´m sure somebody will say “Mainframe Linux! SGI Altix!” but you should really look at the architecture of this systems before taking them as an example for scalability. Linux must decide in the near future, what it wants to be. Without focus to the desktop, it will not be a valid alternative to Mac OS X. Without an focus to server operations it will never really get out of the small server sector. I tend to look at the increased comments that we should stop to develop Solaris and start to contribute at Linux as a sign of people being nervous. They don´t know, which role Solaris will play in the future. They assumed a waning relevance similar to other closed source operating systems like HPUX or AIX. But the opensourcing changed the rules of the game. When we announced Opensolaris, they said “Show us the code”. We showed them the code. When we showed them the code, they said “Show us the community”. We start to show them the community. Now they say “Is this community real?”. There is a real community, there is real interest. And now we will see where this will end. I´m sure that this end wont be irrelevance. And you should consider: When there were sure of their dominance, the Linux community wouldn´t react in this way. They would ignore Solaris as they ignore *BSD. When did you heard the last comment like “The BSD community should give up it´s niche operating system and help us with Linux!” (Okay … one reason for desisting from such a demand could be the consideration that one community is too small for two egos like Linus and Theo ;) ) After all i find this comments “Hey, help us, don´t develop an own system. Help us.” from some parts of the OSS community a little bit strange. It´s not the first time: When we announced the availability of an in-kernel CIFS stack, the same words came from the Samba community. Having alternatives is a good thing, at the end it´s one of the reasons, why Linux got such an important building block of modern IT infrastructures. At which date “Having a choice” got a bad name ? I´m not ridiclious. I do not believe that we will the a market share of Opensolaris larger than the one of Linux anytime soon. But i strongly believe, that we will see a market share that enables a vivid community around it. We already see the first startups using Opensolaris as the foundation of their appliances. More important: Linux needs Opensolaris. Do you really think, that there would be any development fo5 brtfs without ZFS? I´m observing the development of Linux for quite a while (i´m using Linux longer than Solaris), but from my perspective the speed of the kernel development slowed down in the recent years. The Linux community would celebrate ext7 in 2015 (some new features per release, but essentially the same stuff). Linux needs a strong competitor for its further development as Solaris needed the wake-up call from the Linux community. The big step in functionality from Solaris 9 to Solaris 10 is in a part an answer to Linux. And to comment the opening statement of this article at the end of my article: Not the purchase of Mysql will rescue the Solaris franchise, it was Solaris 10 and Opensolaris that already rescued it.