The Economist on The Rise of Linux

nickco3 writes "The Economist is telling the business world that Linux is a worthy adversay to Windows and Unix. It is free, runs on almost any hardware, and generally more secure than Windows As result it is dividing the industry into winners that offer Linux (e.g. IBM and HP), and losers that don't, (e.g. Microsoft). Sun is probably doomed."

Comments

as much as i like the

[minus_273]

economist.. i dont think MS is really a loser. Nor do i believe that anything else will be the dominant desktop OS. As much as we like to bash MS, they do do somethings right. Enough so that they will keep their customers. Heck sometimes they even make technological improvemsnts in their OS that others are just discovering. Look at all the research that went into Async IO in the open source world only to realise MS had done something better for nearly 10 years.

Re:as much as i like the

[hughk]

As you know, the NT kernel is in a way a descendent of Digital's VMS. VMS in turned was based on a 16-bit operating system, RSX-11M. The interesting thing is that in the old days, propietary operating systems weren't particularly closed. RSX-11M was built from source code, so you always got the exec in source. Until VMS 4.5, the source listings were delivered on microfiche with the system. There were some very good books available on the internals of VMS too.

Personally, I think that the NT kernel seems quite ok. The trouble is that it is closed (so I can't easily understand what is going on) and it is surrounded by layers of crap. I agree that the Linux people could have learned a lot from the synchronisation objects in NT, but they could also have leaned a lot from older systems like VMS, which were extremely well documented. However, better asynch support is coming soon.

However, the next big area for Linux is resource locking (needed for clusters). NT does it, but not very well. VMS did it better (and still does, from last time I used it).

Re:as much as i like the

[Ataru]

CreatePipe creates an anonymous pipe, and you can't ask for it to be overlapped. But it just does a CreateNamedPipe with a unique name (and you can do overlapped on named pipes). You could do that yourself, although I admit it's a bit strange the way the API is laid out.
I quote from the documentation for WaitForMultipleObjects:
"The function modifies the state of some types of synchronization objects. Modification occurs only for the object or objects whose signaled state caused the function to return. For example, the count of a semaphore object is decreased by one. When bWaitAll is FALSE, and multiple objects are in the signaled state, the function chooses one of the objects to satisfy the wait; the states of the objects not selected are unaffected." [emphasis mine].
You're not supposed to test objects individually, you just peel one off at a time and go back for another WFMO.
I'm off to read about the UNIX stuff.

Re:as much as i like the

[fitten]

Yeah, in the past, I've basically put a while loop around the WFMO and just loop. Inside the loop is a multiplexor based on the event signaled if you need it. The code is very simple to do this.

I've use the Asynchronous I/O with windows sockets and got very nice results. It was very hard to get nearly the same results on Linux back then (1998-ish) but things may have changed since then on the Linux side. I haven't had to try to get the same levels of performance out of subsystems since those projects because the things I've worked on since then haven't required it.

In my opinion, the POSIX thread/synchronization stuff is bad. It think it's ugly and clunky. I had a friend who was on the POSIX committee for it and I told him the same.

Sun is NOT probably doomed

[Ubergrendle]

Sun is a hardware company; as much as they like to trumpet their 'value add' components and services, their bread winning business is SPARC. Java, a great language, generates little revenue and is a tool that drives need to purchase their hardware. Remember iPlanet? Approaching 0 on the latest web surveys as Apache dominates that space.

IBM and HP are making smart moves adopting Linux business models. As Linux matures and benefits from a gazillion different implementations, AIX and HPUX will begin to look less and less desirable.

Getting back to Sun, Solaris is not a revenue piece for them either. There was alot of complaining in the Slashdot crowd and Sun's commitment to Solaris on Intel has waned, but really, would you like to be running Solaris instead of Linux or Debian? Thought not...

Re:Sun is NOT probably doomed

[Jah-Wren Ryel]

Exactly. The only thing that differentiates Sun hardware is that Solaris runs on it, and thus the multitude of Solaris hosted applications runs on it. Take away Solaris and Sun doesn't lead in anything. Sparc performance has always been and still is a laggard compared to the rest of the industry. Pricing at the low-end is wiped by PCs, and features at the high-end, where the profit margins are still fat, are wiped by IBM and HP in the commercial sector and SGI in the technical markets.

Unless Sun figures out a way to turbo-charge the Sparc architecture, they won't have any competitive advantage in the near future.

Sun isn't doomed

[Sir_Bill_William_Jen]

Sun will be around for quite a some time, because they offer specific product for one market on which they have quite a monopoly.

Linux will probably take some share from Microsoft but It won't be so huge that it will be used more than Windows. Microsoft is too smart to let Linux take hold of them. They will do everything to stop Linux from taking their share, whether it be through Advertizement, Deals with companies, or Palladium.

However Unix might fade away, BSD and Linux provide great alternatives for Unix... But they won't be completely gone.

Sun increase market share against x86 last year

[ChrisRijk]

In the "low-end" of the market (systems costing under $100,000) Sun increased their revenue market share last year (2002 over 2001). Ie they grew faster than the industry average. Almost no Intel, Windows or Linux based systems are sold for more than $100,000.

So if Sun was being hurt by Linux (or x86 based systems in general) why did their market share increase?

It wasn't anything to do with the LX50 (dual P3 Solaris x86 / Linux) systems they launched late last year - they only shipped a bit over $1m worth by the end of the year.

"Open Systems for Open Minds" ...

[Big Jojo]

Sun's just being forced to go back to its roots, which were running a commodity Open Source OS (SunOS started as BSD) on commodity hardware (m68k at first, then SPARC except for a couple years when they also sold 80386 hardware, as they started Solaris). That should be a healthy thing, long term, though they have to get rid of a lot of closed-system attitudes. Like the ones that have crippled so much Free Java work.

If Sun had kept true to their roots, they'd have been running Linux on x86 from day one ... instead, they wanted to keep one founder (Andy Bechtolstein), who wanted to design a RISC chip (became SPARC). So Sun sold out SunOS in favor of Solaris/SVr4, so they could switch to non-commodity hardware. Well I've got news for you: Andy's long gone, and SPARC was never that hot. And the customer lock-in is going away ... customers always wanted the open systems approach, even when Scott McNealy refused to play that game.

Re:"Open Systems for Open Minds" ...

[Anonymous Coward]

Sigh...

False. Sun didn't have the money to invest in SPARC engineering ("productizing the RISC machine", if you will) until it got a boatload of $$$ from ATT as part of its work to merge SunOS features into SVR4.

You sound quite authoritative on this, something that to be verified would require being awfully tightly into the company's balance sheets and strategies. Can you explain your sources?

Given that AT&T owned a serious chunk of Sun at that time, was this the only way they could drive this deal? Seems unlikely.

If ATT hadn't wanted to kill of the main BSD variant, Sun would likely still have been an "open systems" company. And if you don't recall that, you weren't at the company when that was all happening, or in the industry watching the dynamics of SVr4 and the "Oppose Sun Forever/1" OS.

Happens that I was.

One of the strings on that contract was that all new hardware had to ship with SunOS 4 (the SVr4 based code) starting at some particular date.

That would be "SunOS 5," but I get the idea. Once again, your source on this?

The Sun386i predated the SPARC1 you mentioned, and it was one of the last boxes that could ship with 3.x (the "old" BSD codebase). Sun wasn't allowed to continue trying to scale the BSD code -- the ATT contract said so. Those Sun4 machines couldn't (legally) be sold with the SunOS 3.x kernels.

Sun4 machines could run SunOS 4.x, which was BSD-based. The first machines that truly could not run SunOS 4.x were the Ultra 1 systems, which started shipping in late 1995. Once again, source on the AT&T contract having this restriction, which seems to go against future events, would help here.

As has been mentioned before, lots of machines that could run BSD-based SunOS came out after the 386i: all the SPARCstations, for example, and the entire Sun-4 product line, as well as the last Sun-3 Moto-based systems, which came out after the 386i.

As for performance ... odd, the Sun486i was faster than the Sun 4/110, not that it ever shipped.

How odd was that, really? The Sun-4/110 was the slowest SPARC machine ever released, 7 MIPS with no integrated FP; I got to have one on my desk for a while, lucky me. The 486i by the time it was released would have been a contemporary of the next generation of faster SPARC machines, the SPARCstation product line.

So what this all boils down to a conspiracy theory that AT&T demanded the death of BSD, which somehow tied into a plot to kill everything not SPARC. If the latter were true, please explain the Sun-4/80 and Sun-4/370, both of which feature Moto 68040 and were competitive with SPARC performance in the repective niche, and were on Sun's price list into the '90s.

I think it's more likely that the 486i was killed because demand for the 386i had tapered off; it wasn't well-integrated into Sun's product plans, and was more a victim of East Coast/West Coast internecine battles than anything you've proposed so far (it's a good day when I can work "internecine" into a conversation :-). But the easier to swallow explanation was Scott's famous "all the wood behind one arrow" edict, which made good sense for a $1B company to follow, and served Sun and its customers well for over a decade. This is the one part of your story that makes sense, that Sun would want to go with one chip architecture rather than the way things were being done at that time, which was to have three separate chip architectures. It also made sense that that architecture be SPARC, for many, many reasons, none of them having to do with AT&T that I am aware of. If you can substantiate this, I'd be interested in hearing about it.

A Lesser Form of Unix

[MBCook]

"But for many applications, Solaris is overkill, and Linux, a less capable flavour of Unix, is good enough."

I don't know much about Solaris, so I'd like to ask you guys out there. What makes Linux less capable? What does Solaris do that Linux can't do (at least well enough)? Just wondering.

Re:A Lesser Form of Unix

[jc42]

While it's not nearly the whole story, an interesting benchmark is the top 500 supercomputers list. Sun and linux are both on the list. Sun seems to be the main supplier of supercomputers these days. But they first appear on the list as number 156. There are only two linux systems, but they are numbers 5 and 45.

I guess that doesn't tell you all that much about the top end, other than that linux and Sun are both quite capable of supporting raw, top-end gigaflops power. The actual computing is done by processes, of course, so this benchmark really just tells you that neither OS has any showstoppers. Neither interferes with the ability of a process to crunch bits and bytes.

Now on to the other benchmarks ...

Sun doesn't support Linux?

[Biolo]

Ok, Sun is not 100% behind Linux (yet), but that's because Linux isn't ready for the high end (yet). By high end I'm talking about F15
and F12K
servers. It is pretty close to having the capabilities to run on the Sun midframe stuff, for example I'm sure it would run fine on a 3800,
maybe even the 4800, but you start to reach its current limit with a fully stuffed 6800 system.

Now, step back for a minute and think why Suns UltraSparc and Solaris solution is so strong. Simple, at the risk of repeating the marketing guys the lure is that you can give your development and deployment guys a bunch of cheap Sunblade 150s or some cheap UltraSparc blades and whatever they come up with can be moved straight onto anything up to and including an F15K without recompiling. Put yourself in the place of a big corporation. Your putting together a new system, you have no idea just how big a load it will eventually have to take (say in 5 years). Today, sure you could run it on high end Linux box, but what happens if 6 months in the system needs a bigger box? If you chose Sun in the first place you simply buy the bigger box and move over. No porting, no redevelopment, and you know there is always a bigger, faster system you could move to. It buys you severe scalability that Linux isn't placed today to provide.

Now, about not supporting Linux, what about the LX50, the Sun Open Desktop that is coming soon, the Lintel blades (Coming Soon(TM)) the fact that the entire Sun One stack (web, directory, identity, etc, etc, etc) is either available now for Linux or coming soon, not to mention Star/OpenOffice.

So what is the perceived issue? I think people don't see Sun offering Linux on the UltraSparc range and thing they don't get it. Sun does get it, but look at their selling point for the last 10 years, total scalability. Linux doesn't provide this yet so they can't buy into it. What they are doing is making Solaris as compatible with Linux as possible, whilst at the same time helping Linux by providing software (openoffice, SunOne and much more) and I believe some kernel code too.

Believe me, when Linux is ready for the F15K class systems Sun will be ready for Linux to be there.

Disclaimer - I work for Sun, but nothing I have said here is not already public information.

Re:Because the article submitter is full of BS

[TKinias]

scripsit 0x0d0a:

Okay, I'll certainly grant integration and easier to use for unskilled staff, but where does the Economist get more "bells and whistles" from? When I think "bells and whistles", I generally don't think of Windows...

That's interesting. I think you may understand the term ``bells and whistles'' differently than I, and apparently the author, do. Would it make more sense if he had written, in lieu of ``bells and whistles,'' ``cute and briefly entertaining but eventually tedious and ultimately useless misfeatures''?

Clippy is ``bells and whistles.''

Correction Mr. Mundie

[bluGill]

Sorry Mr. Mundie, but Apache is NOT a clone of comercial software as you tried to claim. Apache is a fork (clone) of the Origional NCSA web server, which was NOT commercial. (open source, but I'm not sure exactly what license was used so it might not meet the exact legal definition of Open Source). IIS, and the other comercial servers are clones of an open source webserver.

Of course this is all an accidemic exercise, but don't try to claim some high ground where Apachee has it.

Apache a CLONE??

[cmacb]

"According to Craig Mundie, Microsoft's chief technical officer, as the open-source movement grows, it will get better at producing free clones of commercial software."

I give up. What is Apache a clone of? I wish they had included an actual quote on this. Maybe the journalist was "interpreting".

Most of the article is on target though. The easy way to evaluate the strengths of the companies mentioned is to look at how diversified they are (or aren't).

IBM is no longer primarily a hardware company. They have a strong consulting division, they do fundamental research and grab pattents on REAL things (rather than new parsing algorithms like some companies I can think of), they have a very strong software development component, they farm out hardware manufacturing that is no longer profitable (disk drives) while hanging on to things that they do best and can make money on (chip fabrication).

Sun is primarily a hardware company. Their operating systems are (almost) exclusively sold to customers who use their hardware. Java and Star Office are far from being cash cows. Their weakness is that as Intel, AMD, etc, chips get cheaper there is less incentive to use Sun's higher priced hardware components. Supporting Linux helps them a bit, but it is the price of their hardware that puts them at a disadvantage.

Microsoft is a software company. They are trying real hard to become something else too, but like Sun they are having a heck of a time making anything else work. They don't really do fundamental research, but instead try and grab patents on programming concepts so that they can bully other companies in court when it suits them. They don't really make any hardware, but instead stamp their logo on a few things to make it seem that they do. They do select good subcontractors for mice and keyboards, I'll grant them that. Everything they do except Windows and Office lose money. Prospects for either of those (because they are already so successful) can only go down. They currently have a scatter shot approach to the "next big thing" which consists of trying everything at once and seeing if any of it takes off. Few companies have the money to do this. But they will bleed themselves dry rather quickly if they are not carefull. Something tells me they are not going to be carefull.

Apple is trying to diversify too. Since they are starting small the only way they have to go is up. It would be nice to see them further popularize the power-pc server. My personal experience with OS X is that they are rushing versions of it out the door too fast. I've decided to wait for XI (or whatever they call it) and switched to running Linux on my iBook. Those gel buttons are cute though.

I'm curious.

[Kaki Nix Sain]

My background: been using Linux for a good while now (current=gentoo), so I'm not a newbie from the config-it, use-it, script-it angle. However, I've never had anything but short, infrequent inclinations to dig down and learn about the mechanics of how the fundimentals of the OS work. So from that angle, I'm total newbie.

Now to the question. You say "...that's because Linux isn't ready for the high end (yet). By high end I'm talking about F15 and F12K servers." Which gives me a jolt of curiosity. What are some parts of the mechanics in Linux that prevent its move to high end systems? Are there any fundamental issues that will prevent Linux from 'growing up' and eventually running on those systems? Must some parts of the Linux mechanics be scrapped/overhauled? Or is it more a matter of time and will to add/extend pieces that will make the move possible?

Basic economics..

[segfault_0]

Microsoft and Sun will both be big losers in my opinion, for reasons of simple economics. Commoditization of software has been going on for a long time. This is why you see Microsoft constantly screaming from the rooftops about their new features, which 90% of people dont care about. Its the only thing that makes them different from your Linux's and OpenOffice's.

There will come a point when it really just wont functionally matter which os or productivity suite you use. At that point in time, the cheapest alternative always wins. So unless Microsoft plans on paying us to use their software - time and open source software will inevitably have their way with them.

Re:Basic economics..

[Anonymous Coward]

Hey Batman, riddle me this: If commoditization has been going on for a long time then why didn't Sun die 10 or 15 years ago?