English | 中文 | Русский | Français

2,595 Posts served

8,341 Conversations started

Why do we need another OS?

By David Stewart (Intel) (138 posts) on April 10, 2008 at 3:03 pm

If you have Windows and Linux, why do you need anything else? Oh sure, there are always special-purpose OS's, which address a particular niche or usage. But in terms of general-purpose OS's, you have two basic models, one of them open source the other closed. Both have a broad and lively development community and broad hardware support. Both can be adapted to scale from cell phones to mainframes.

I sometimes measure my progression along my career road by how often I hear the same concept repeated. As I approach 25 years out of university, this OS meme has made the rounds many times and in many forms. And from what I hear from the true graybeards, everything we deal with today started out in the mainframe days.

The last time I experienced it was in 1994. At that time, UNIX was fragmenting, Microsoft was trumpeting the scalability and power of Windows NT. I believed them. I concluded that OS work was pretty much going to be a Microsoft-only thing. Outside of Redmond, there would eventually be no OS jobs. Well, maybe there would be a couple of boutique OS shops, but the reality is that NT was going to take over, because they have the power, support and resources of Microsoft and their developer community to take over.

At the time, I transitioned my career to focus on the application level, and didn't look back.

All of this sounds really stupid today, of course, because we all know that the power of the Internet and the availability of cheap hardware and free information enabled Linus Torvalds and company to launch Linux, and fight The Man. But at the time it seemed like we could "close the patent office" so to speak when it comes to operating systems.

So when it comes to today, I think there absolutely is space for another general purpose OS. Why?

• There is now a lot of fantastic OS intellectual value out there which is still closed source. Now that Sun has shown us how to open source most of its crown jewels and use it for competitive advantage, there is now no reason why a lot of other companies couldn't do exactly the same. Read this month's (April 2008) Harvard Business Review for the business case.
• Not all open source is created equal. There are actually different models for licensing, innovation and control, each with its pros and cons.
• There is a constant hunger in the technology world for what can speed up innovation and return more return on investment. There is an unplumbed depth of desire, mostly in The Generation, for what is new, cool, free and powerful. It's like the world youth culture is on a constantly moving treadmill - if you don't keep yourself moving, you will find that you are dumped off of the end of the belt.

OpenSolaris may very well become the new and cool general purpose OS to take its place beside Windows and Linux. And gee, we didn't even mention Apple OS X!

(Apropos of that last analogy, did I ever tell you the story of how I got my leg caught in a conveyor belt when I was a kid?)

Categories: Open Source, Software Engineering

Comments (4)

April 11, 2008 1:10 AM PDT vermaden

And gee, you didn't even mention FreeBSD either! FreeBSD is a big competitor to Linux and (open or not) Solaris, it already implemented ZFS, they are implementing DTrace currently, and if you are not interested in Xen like virtualization, then you will find FreeBSD much cleaner, more logical, and more production ready. Also lots of FreeBSD code is used in Mac OS X. Check more here: http://www.freebsd.org/features.html

April 11, 2008 12:54 PM PDT Jamie Fenton

I believe a new operating system will become absolutely necessary, for two reasons. 1) Unix, Windows, etc., emerged a human generation ago. There have been a number of fundamental changes in how software is engineered which are supported by these systems as after-thoughts, if at all. The implication of this is API bloat - eventually you have several (if not more) ways of essentially doing the same thing. Alan Kay describes the life cycle of an OS something like this: A new system launches with high hopes like a sleek ship and then starts growing barnacles relentlessly until the hull scrapes along the bottom of the ocean and the only way to clean them off is to build a new vessel - and the process repeats. Vista and other recent Microsoft stuff are deeply encrusted. Linux was born wearing 3 generations - but Linux fans seem less troubled by them. So we need to refactor an OS to where a mortal programmer can remember the APIs. Reason 2 is what will actually cause this to happen, which is PARALLELISM. While Windows and Linux work OK with 2-8 CPUs, we will have thousands, perhaps millions of these things. Neither OS seems to fit well with this "one CPU per pixel" inevitable future. Unfortunately there is not a clear answer yet to the question of "How can we project the problem space of parallel computation onto the existing capacities of human metaphoric understanding in productive way?" While I can list a number of possibilities here, this is a blog comment and not a PhD thesis. I will note that there are a lot of recent developments reminiscent of the 1960s explosion in domain-specific programming languages. This suggests that the phase after will bring new general tools (like C was to the 1970s). Part of the process will be improving both the machines and the humans to the point that this meet-up is possible. A quick digression to Intel's Research group shows a number of initiatives in this direction. http://techresearch.intel.com/articles/Tera-Scale/1421.htm

April 11, 2008 10:51 PM PDT Michael Shadle (Intel) Total Points: 373 Status Points: 323 Green Belt

FreeBSD is cool, but ZFS is not "already implemented" - I wouldn't consider it practical until it is considered "stable" ZFS is "implemented" in FUSE on Linux too... but hardly worth a mention. Not too long ago I heard rumors about FreeBSD and it's following slowly dying off, updates taking too long, etc. It was probably all speculation, since I believe it was relating to the upcoming 7.0 at the time. You wouldn't hear about that from Microsoft (due to it being commercial and probably their largest source of revenue) or Linux (the community is just too big, with plenty of people who would step in and take over control, and a few commercial companies like Red Hat would still compete with it probably...) It makes me nervous to hear chatter like that, since it could very well be true. The FreeBSD committer group is only what, 7 people? The community is not as large as Linux, and there's no popular commercial interests in it other than OS X (which has probably forked off enough code now to be on its own...) Either way, I have not really hard much chatter recently about FreeBSD losing steam. I would love to see ZFS hit production (even if it's not feature complete) and I do consider it an option for my own servers still. Relating to this article though, I believe it would suffer the same fate as Linux - it designed for x86 architecture, it will scale a certain amount of cores, but not thousands properly. On, and while I'm on my soapbox: it's also disappointing about the different licensing out there... it's like the open source version of software patents to me, sometimes stifling some interoperability and innovations...

April 12, 2008 1:24 PM PDT David Stewart (Intel) Total Points: 9,435 Status Points: 9,435 Black Belt

To Jamie Fenton, I love this comment: "So we need to refactor an OS to where a mortal programmer can remember the APIs." It used to be that people loved UNIX v6 because you could fit the source code listing of the whole OS in your briefcase.

Trackbacks (0)

Leave a comment