Subscribe: Comments on: Do Operating Systems Matter? Part 1
Added By: Feedage Forager Feedage Grade B rated
Language: English
distribution  innovation  kernel  linux  much  operating system  operating systems  operating  rel nofollow  stack  system  systems 
Rate this Feed
Rating: 3 starRating: 3 starRating: 3 starRate this feedRate this feed
Rate this feed 1 starRate this feed 2 starRate this feed 3 starRate this feed 4 starRate this feed 5 star

Comments (0)

Feed Details and Statistics Feed Statistics
Preview: Comments on: Do Operating Systems Matter? Part 1

Comments on: Do Operating Systems Matter? Part 1

because technology is just another ecosystem

Last Build Date: Mon, 13 Nov 2017 09:54:21 +0000


By: stephen o'grady

Tue, 21 Nov 2006 21:58:22 +0000

Bryan: count me as somewhat surprised as well. even even virtualized environments need to run on top of something. and frankly, virtualized environments are not perceived by many of the folks i speak with as a first tier production deployment environment. they will be, i think, but aren't there yet. in other words, reports of "death of the os" are slightly exagerrated. James: i'm not sure i buy that. if i try to pitch, say, BSD into many enterprisey clients i'm going to get massive pushback. and windows v linux fights in large enterprise shops are often vicious. OS's may be perceived as commodities, but it's not really playing out that way. Bryan / Mike: maybe you guys can agree to disagree. all i ask is that you respect one another's position. Bert: "I've marketed and supported a LOT of complex technology products and I humbly maintain that the support effort for something like WebShpere or DB2 could be eased by at LEAST an order of magnitude if delivered as an appliance." you think customers are ready to receive either WebSphere or DB2 as an appliance? i don't, not yet anyway. i also think the comparisons between car or cell phone operating systems are a stretch. those are devices with prefined purposes; application servers and databases are general purpose platforms. apples and oranges, in other words.

By: Bert Armijo

Fri, 10 Nov 2006 21:47:14 +0000

Most of the OS's we use in our daily lives we have no interaction with directly. We're sheltered by a user interface (cell phone), or the OS is embedded so deep we never see it (car transmission). This doesn't mean they don't matter, but rather that they can do their job without imposing themselves on us. Conversely, the OS's we typicaly use on our servers have required constant attention. Sys admins seem to be constantly building, installing or patching something on every machine. In this way, these OS's have mattered a great deal - they've driven the cost of operating a data center into the stratosphere. Just because the OS is used in the data center doesn't mean it has to demand so much attention. Consider the number of OS's in the data center with which we have far less interaction. Cisco could probably switch the base OS in their switches and routers and you'd never even know, because your interaction is strictly through the CLI. Likewise F5 or APC. You treat all of these as appliances. Your interaction ends at the UI. Can WebShpere or DB2 be run like an appliance. They can today. And consider the advantages of doing so. The vendor could choose the OS that best meets the needs of the application. They could configure the OS specifically to run the application and know that nothing else would be running. And, possibly most importantly, no one could change the configuration. I've marketed and supported a LOT of complex technology products and I humbly maintain that the support effort for something like WebShpere or DB2 could be eased by at LEAST an order of magnitude if delivered as an appliance. Evidently others agree, because this trend is already taking hold on the desktop for just these reasons.

By: Bryan Cantrill

Fri, 10 Nov 2006 00:00:35 +0000

You write:
A kernel does not equal an OS and 'Linux' as an OS is much more than a kernel.
That's only true if one subscribes to the Larry Ellison definition of Linux (namely, whatever you want it to be). As Linux defines itself, however, it is technically just a kernel. For example, from the gentoo documentation:
What do you think of when you hear the word "Linux"? When I hear it, I typically think of an entire Linux distribution and all the cooperating programs that make the distribution work. However, you may be surprised to find out that, technically, Linux is a kernel, and a kernel only. While the other parts of what we commonly call "Linux" (such as a shell and compiler) are essential parts of a distribution, they are technically separate from Linux (the kernel). While many people use the word "Linux" to mean "Linux-based distribution," everyone can at least agree that the Linux kernel is the heart of every distribution.
This is more than just nomenclature: in order to be able to deliver value up the stack (to get back to the original discussion), one often needs to be able to deliver technology that straddles the user/kernel protection boundary. In the Linux model, this delivery is essentially impossible: one can deliver the kernel portion into Linux, but the user-level portion of any technology must be left to the distributions. It's my opinion that this is the wrong model -- I have a much more expansive view of the operating system than does Linus, because I believe that the operating system can and must deliver value at higher levels of abstraction.

By: Mike Dolan

Thu, 09 Nov 2006 22:03:17 +0000

Ok, since we're pushing corporate agendas now, AIX, APV, z/OS, z/VM, Linux, and i5/OS. OSs, hypervisors, libraries, etc - let's throw them all in with anti-contraction and see what happens. A kernel does not equal an OS and 'Linux' as an OS is much more than a kernel. The above comment belongs on /. - I wouldn't expect to see it here and it's disappointing. Funny... claiming an OS that is trying to be like Linux is more creative than Linux... funny, near Redmond-ish. What do customers think? Last I checked Linux outsold Solaris in 2004... and has been growing 9x faster since. There's a reason.

By: Bryan Cantrill

Thu, 09 Nov 2006 18:21:46 +0000

I guess I would counter that the fact that OS's aren't perceived as having value further up the stack is really a failure on the part of operating systems to realize their potential to add value up the stack. That is, the perceived failure of operating systems is actually due to a lack of imagination (or understanding) on the part of OS implementors more than an abstract limitation of the idea of an operating system. To be honest, Linux has exacerbated this because instead of expanding the definition of the operating system (which must be done to be able to add that higher-level value), Linus and co. have actually contracted it: unlike most operating systems that have come before it, Linux doesn't recognize the system libraries and utilities as being a part of the operating system. (They rely on the distribution for this -- which dramatically limits the value that Linux itself can ever provide up the stack.) For our part in Solaris, we have been busily working to add value higher and higher in the stack. As a concrete example of this, I would point to our recent support for JavaScript in DTrace -- but there is still much that can be done (in many disjoint areas) to allow operating systems to add value high in the stack of abstraction. Point is: the Operating System matters -- even if certain operating systems don't see it that way. ;)

By: James

Thu, 09 Nov 2006 12:03:10 +0000

Operating systems matter and don't at the same time. If you ask me about what operating system matters most in corporate America I would say Solaris as it provides capabilities that Linux doesn't have. Even MS beats Linux is several areas. If you are talking about operating systems to enterprisey folks, they aren't really worth talking about for more than a minute as they are commodities and the value to us is a lot higher up the stack.

By: Bryan Cantrill

Thu, 09 Nov 2006 00:01:11 +0000

I guess I'm a little amazed that this argument is still being had -- if only because we in Solaris have provided so much evidence now that the operating system actually does matter. (Not to toot our own horn, but if OS's don't matter, what the hell is the WSJ's problem giving an operating system its top innovation award? And not just its top software innovation award -- though that too -- but its top innovation award, period.) To me, this is the ultimate vindication that operating systems do matter -- and that innovation in the operating system has the power to provide unique value in information technology. There will of course be laggards that continue to rephrase the well-worn arguments of commoditization in the OS, but they are just that: reflections in the rear view mirror of a zeitgeist that's been left behind.