Slashdot Mirror
What Will Happen in IT in 2007?
An anonymous reader writes "ZDNet's Paul Murphy has set out his IT predictions for 2007. Featured among the completely predictable, OpenSolaris overtaking Linux is apparently inevitable within one year. From the article: 'By the end of the year the OpenSolaris community will be widely recognized as larger and more active than the Linux community.' Is 2007 the year of the OpenSolaris desktop? Other 'inevitables' include Microsoft's success with Vista, the continuing phase-out of Itanium, and the Cell processor powering most of the world's super-computers."
Specifically Nevada build 54 and Nexenta alpha 5. They have some interesting technologies (specificaly ZFS, which is interesting, and Zones, which is a bit lower overhead than virtualization, but not as flexible (everything still goes through the Solaris kernel)). Nexenta I honestly thought was a cool concept, and executed quite well, Debian package management and GNU software that is clearly better than some of the Sun basic utilities (and much less Java inclined...), offering the benefits that Solaris does have to offer...
Anyway, from my working with it, I know the OpenSolaris is certainly full of themselves, and some denial, but I don't think they can live up to their own expectations. For example, any complaint or bug frequently got met with 'at least you aren't running linux!'. They trashed on lack of documentation in linux while I struggled to find some documentation on their stuff that seemed unwritten. They'd pick up a decade-old howto and say 'this is how linux requires you do it, versus our not-yet released way, see how crappy linux is'. When people talked about how woefully (understandably) incomplete their ACPI and suspend support was, they pointed at linux and said 'linux acpi support hardly works at all, so don't expect too much' despite the reality of 3 out of 3 generic motherboards I've tried worked splendidly with linux acpi. My laptop despite being one of their officially tested still doesn't have clock modulation and their acpi parser barfs on the DSDT that nothing else (not even intel's compiler) even warns on. People discussing panics/hangs are met with 'at least it doesn't crash as much as linux', despite evidence to the contrary. They are used to a closed, proprietary world of a select set of hardware and the open world if they make any headway in is going to give them quite the wake up call. They talk about how much better their driver support is, despite the glaring lack of drivers. Largely their efforts in expanding that involve porting drivers from the BSD projects.
Anyway, their current implementation does admittedly seem adequate for most server type activities if the hardware is supported. I could see a lot of hardware vendors happy about a system with a stable binary interface for drivers that doesn't require rebuilds for every uname -r, but hardware vendors face the market realities and put up with the pain if they want to play in the server space. I understand the hassle, but linux making a PITA for hardware vendors have given us a lot more driver source than we could have hoped for. For the market, probably the single best card they have is ZFS. They have done a good job of consolidating volume management, software raid, filesystem, stuff like snapshots, and paranoia of checksumming everywhere into a single implementation. In doing so they have done things more efficiently (such as RAID format on disk leveraging filesystem layer knowledge for better performance), and trustworthy (a controller failing to report data corruption is detected at a higher level). ZFS is impressive, and that was/is the one thing that makes me really want the rest of the platform to be usable for me day to day.
DTrace is much hyped, and very useful in the hands of good developers and good administrators, but I don't see administrators at large making use of it enough to deliver on the hopes Sun sets up for it.
Zoning is a nice logical extension from simple chrooting which is more comprehensive, and more efficient than the other extreme of virtualization, theoretically. However, with virtualization being ubiquitous and most of the market accepting the ever-reducing overhead for the flexibility, I don't know if Zones are going to excite anyone that much. The BrandZ extension of the metaphor gives it some flexibility, but again their Linux profile still doesn't run linux things just right, and a linux vm with the linux kernel already will do so today.
So you have a platform that probably won't need to be as successful as linux had to be in order for hardware ve
XML is like violence. If it doesn't solve the problem, use more.
Could happen, if Solaris was already massively better than Linux. I don't think that will happen.
/proc was ripped off wholesale from Plan 9. If ZFS is ultimately such a great idea, most of its advantages will be absorbed into Linux, so eventually you'll have OpenSolaris, which implements ZFS perfectly and may have slightly better-looking code, and Linux, which does almost everything you need from ZFS, but also has binary blobs from nVidia and ATI, can run in usermode, has suspend to disk, runs on an iPod, and does many other things that Solaris will still be catching up with.
The simple reason is: Worse is better.
Why do you think absolutely everyone on Linux was using Mozilla? It was the main Gecko program, and your other options kind of sucked. Mozilla got the job done, and everyone was developing for it -- you were guaranteed to have new and interesting stuff (Flash, Java, RSS, tabbed browsing, etc) on Mozilla, either before it was anywhere else, or within a month of it being implemented elsewhere.
Of course, some things never made it into Mozilla -- for instance, Amaya is both a web browser and a WYSIWYG editor, and you can jump into any webpage and edit, and save the new version somewhere -- there may even be a mechanism for re-uploading it. But there must not be that much demand for such features -- after all, most of us either use Notepad (or vim), or we use some nicely-done AJAX WYSIWYG.
You could point to Firefox, but remember: Firefox was originally named "Pheonix", because it rose from the ashes of Mozilla. Had Firefox been written from scratch, it would never have gotten where it is today -- old Mozilla bugs and all.
That is what will happen with Linux and OpenSolaris.
Linux is already much, much more popular than BSD or OpenSolaris -- or, for that matter, Plan 9. So, we take the best ideas from other OSes, so long as we can reasonably implement them, and we also toy with new things of our own. If I remember right,
The only way this picture changes is if Solaris is so ridiculously better than Linux that the few people hacking on it now are enough for it to surpass Linux -- keep in mind, there will be plenty more people hacking on Linux at the same time. This has happened in the past, on a smaller scale, but I just don't think Solaris is better enough -- remember, evangelizing won't work. You won't get me to hack on Solaris till it runs on my Powerbook, at least -- and you need people like me to make it run on that Powerbook. You need it to already be almost as good as Linux, if not significantly better -- and not just in a few areas I don't care about -- in order to get me to hack on it.
If you really want to replace Linux, come up with something that's both better enough that it takes half the time to write it in FooOS than in Linux, and can run a Linux kernel alongside it (do something tricky with UML, or something like what Apple did with Mach/Darwin), so that I can load up my nVidia driver and play Quake 4, and still hack around with something cool like, say, a new cluster filesystem. You have to do it right, though -- I should be able to load my Linux kernel, nVidia driver, and Quake4 binary (and maps) from my own FooOS cluster filesystem.
If you can do that, and provide compelling enough development tools to sway the Linux kernel devs, then we might actually lose the Linux kernel -- slowly -- and replace it with something better. Unless you can do that, Linux will remain the best we've got, now and forever.
Don't thank God, thank a doctor!