@caitlyn

You wrote:
"If you read the blog headline or the comments this is mostly portrayed as Linux vs. Windows. In reality that just isn’t true."

AAMOF, David Williams makes a similar point as applied to server systems in his ITWire piece 'Why Microsoft will dump their anti-Linux Rhetoric' found at http://www.itwire.com/content/view/17988/1141/
In this piece, Williams specifically reveals how several studies of Microsoft's Get the Facts campaigns are in fact NOT AT ALL true portrayals of Linux vs. Windows on the enterprise systems in question.
Furthermore, this seems to be one of your main points above in the body of your blog regarding Asus EeePC desktop systems with Xandros vs. Windows XP.

Getting back to enterprise server OS deployments, Williams summarizes the faux Linux vs. Microsoft studies at the beginning of the piece here:
-- begin quote ---

I've been combing through Microsoft’s “get the facts” web site this last fortnight. Here Microsoft promise to reveal the “facts” on Windows vs Linux solutions. They cite company after company that abandoned Linux because it was slow and unreliable and generally hopeless, but opted for Microsoft servers and found unsurpassed profits, efficiency and general happiness. Yet, the headlines have little relevance to the case study. Microsoft’s PR department are insulting their audience. They are lacking in integrity. It’s time for the site to be pulled down.
--- end quote ---

What follows here in this is a critical analysis of Microsoft's clearly misleading Linux vs. Windows comparisons.

In a sense, Williams' critique is reminiscent of your own discerning comments in response to those of Rufus and ram in 'Is Linux Really Outgrowing Its Stereotypes? Does It Matter?' found at http://www.oreillynet.com/linux/blog/2008/04/is_linux_really_outgrowing_its.html
One can look forward to reading your own critique(s) to responses given here.
-vv

"There are many situations where Windows or commercial, proprietary UNIX really and truly is the best fit."

Hmmm. OK, Unix maybe, though the situations are increasingly rare when it's worth bypassing Linux, FreeBSD, or OpenSolaris for some expensive commercial Unix. Oh, I'm assuming you mean an expensive commercial unix, which maybe isn't correct.

As for Windows? The only situation I can think of where it rates even the slightest consideration as "best fit" is when it's in a shop where it's already entrenched, and the migration costs are too painful. Which of course is by design, a nice self-reinforcing lock-in cycle. Making matters worse, migration doesn't get any easier or cheaper by waiting. Any cost-benefit analysis needs to include:

-Collateral damage from the worldwide botnet, which most estimates put in the billions of dollars per year
-The direct costs of trying to keep out of the worldwide botnet
-Directly higher hardware costs because Windows costs more to do less
-The costs of license compliance. Microsoft even helpfully offers a licensing server to keep track of everything
-The potential costs of a BSA audit, and you will get reamed
-Getting stuck with DRMed products that get cut off and you're left with nothing (buh bye MSN Music store and all the money you spent)
-The cost of delays when their dommed authentication servers emit false alarms or are not accessible, so you can't even use your own system or complete a new installation
-The indirect costs of a paralyzed, stifled marketplace with few alternatives. If it weren't for FOSS, Microsoft would rule and we would be even worse off
-The direct costs of inflated software prices for guff that should be commoditized

There's probably more. But that's all I feel like typing now :)

@vainrveenr: I actually read David Williams' article for IT Wire shortly before you posted your comment. What I don't get is why Microsoft couldn't just use the Chinese case study in an honest way. That study clearly demonstrated that Windows Server 2008 is more reliable than Windows Server 2003 and that an upgrade may actually save their customers money in the long run and improve the bottom line. It's a compelling case. Instead they tout it as a victory over Linux even though the customer hadn't ever actually deployed Linux or even seriously considered Linux for their hosting environment. Such deceptive advertising begs the obvious question: Why is Microsoft so threatened by Linux? Could it be that Linux is, in most cases, a better server product than Windows?

@Carla: Linux still can't compete with commercial UNIX at the high end. It doesn't scale up to big iron well. If you look at successful deployments of Linux on IBM Z series mainframes or P-series RS-6000 servers it's in running multiple virtual servers on one box. When bigger iron is needed for sheer computing power the only way to achieve similar results with Linux is with a sizable cluster. Linux still can't compete well with AIX or Z/OS from IBM, nor is anyone likely to use Linux to replace high end Sun servers or an HP Superdome. Those are examples of where UNIX is still the better fit. I know that's an area Linus & co. are working on.

Windows is usually the best fit in three situations. You hit one squarely on the head: when the migration costs are exorbitant. Training costs, software porting, new licenses for commercial software that can't easily be replaced by FLOSS equivalents, and the cost of testing and deploying new code are all issues that can work in Microsoft's favor. In such situations if a customer is seriously interested in moving away from Windows I recommend doing so when hardware and/or software need to be refreshed anyway to mitigate some of the costs.

The second situation, which applies equally to the server and the desktop, is when a very specialized piece of software is needed for the business and it's only available for Windows. To use a desktop example: there still isn't an adequate replacement for Autocad that runs on Linux. This scenario is becoming less and less common but it still does exist, more so in the business world than on the consumer desktop we discussed in the comments on my previous article. The best use for Windows in the corporate server room, the only really good use I can think of, is as a specialized applications server.

The third scenario is when a business has a customer who demands Windows and not fulfilling that demand will cost the company a valuable customer. In those scenarios a company can try and sell a Linux solution by pointing out advantages over Windows. If the customer isn't buying then you have to weigh the cost of losing that customer. It's often many customers, not just one. In the end the business has no choice but to satisfy customer demand even if the customers are making a poor or uneducated choice. Money talks.

All the issues you raise are real and have to be weighed into any cost/benefit analysis. The truth is that a Windows server or desktop can be secured properly. It is costly and does require serious admin skills. It's part of the cost of running Windows. Authentication servers can be redundant. The cost of having to use larger numbers of servers is part of the cost of using Windows. I won't do work for a customer on a Windows machine that doesn't have all the required licenses. It's my way of insuring I'm not blamed and I don't have any liability due to a customer trying to cut corners. Those licensing costs are also part of what using Windows requires. All of these factors help me sell Linux solutions. There are still times, as I've detailed above, where I just can't replace Windows with Linux as much as I might like to.

Congratulations on writing a fair critique. I wish there were more like it. I am using an eee to write this - bought for its Linux OS and its price; an unequivocal endorsement which trumps my poor linux skills. Any comment I make is in deference to far superior other postings (Hi, Carla!). I also abhor the continual myopic bleatings of the Linux fanboys - which have already cost me much in frustrated expectations that would have sunk a less dedicated explorer. I am unequivecally in the Linux camp, but it is a harder road than I was [mis]led to believe. Linux can stand by its own strengths, it doesn't need fanboy hype. Challenge people to master a new and powerful system, don't be defensive of a system that is too readily compared in terms of it's differences from an inerior one. Just ignore MS-Windows; it is a spent force. Keep to facts, and let others sink by their own misinformation.

Caitlyn Martin says: "Why is Microsoft so threatened by Linux? Could it be that Linux is, in most cases, a better server product than Windows?"

=> From what I understand, it goes a little further than technial merit potential. (technical is one reason, but not the root cause of Microsoft feeling threatened.)

Microsoft's primary business model is of selling software licenses. We can see this quite clearly, as their big money makers are Windows and MS Office.

Linux (of which is well known for being open source), threatens that business model at its roots. The main commercial distros don't charge by licensing software. They charge by support subscriptions. (Red Hat and Novell do this)...Come to think of it, its really the General Public License doing the damage! (Steve Ballmer called it a "cancer", and MS's Open Specification Promise delibrately made things "a grey area" for GPL coders).

Both sides know: it will be an eventuality that Microsoft will find it harder and harder to sell their software solutions as open source become on par in features. (This gets worse as word spreads around the world of open source).

Its safe to assume that Microsoft tried to "nip this in the bud" with guns blazing, (an all-out campaign to sledgehammer while Linux was still young).

What they fail (and still fail to achieve) is a good, solid, honest reason as to why one shouldn't at least give open source a try. ("Get the Facts" failed because it was simply a bunch of manipulated half-truths to make Microsoft solutions look better).

The fact is, in terms of human relations, you can't lie forever. When you lie and deceive, it takes MORE effort than it does to be honest and open about your deficiencies. Admitting your weaknesses upfront gives you an opportunity to solve them, so they never bother you again. It takes a load off your shoulders. (This is what I like about open source...Nothing to hide).

This simple principle not only applies to computing, but to life! Think about it: When someone cheats, someone runs from the law, someone lies to make a monthly quota, etc. The stress gets so high, that something or someone is bound to crack. It hurts more when things are found out.

The key thing that Microsoft is failing at, is integrity. Integrity leads to a great reputation. When you are known for having a great reputation, people are more likely to work with you and buy your products. (They won't have doubts about their decision because they know you can be trusted).

So far, they have sacrificed integrity and their reputation for massive profits. Such behaviour is unsustainable. Eventually, things will start to crack, then chip off, then crumble...The situation will head to a point of "critical mass", and if nothing is seriously done about it, the whole infrastructure will collapse.

Absolutely true! This also applies to the eternal Mac comparisons at well. I do have to say that in the past few years as we've gotten more Macs (at an .edu) I've had significantly fewer hardware problems, so hardware quality is also something important but impossible to "measure".

@Al: There is always a learning curve when changing operating systems since things are done in a very different way. What you have gone through is perfectly normal and applies equally to a MacOS user moving to Windows, for example. (I actually think moving to Windows is more painful.) Linus Torvalds, the creator of Linux, once compared changing operating systems to "performing brain surgery on yourself" and I wrote a piece here expanding on that idea. Trust me, in six months you will wonder how you ever got by with Windows.

@josuhuadfranklin: Of course, you didn't say which brand(s) of PCs were replaced by the more reliable Macs at your university. I'm sure we all know that you didn't mean to imply that Mac hardware is more reliable than every other PC manufacturers hardware on the planet since you can't possibly have tested them all. :)

Whenever a comparison is made the parameters of the testing have to be laid out or the comparison becomes meaningless. Thanks for reminding us of that.

>Those with an agenda, either agenda, will, >I’m sure, attack what I have to say.

Talk about being arrogant: anyone who doesnt agree with you has an agenda.

Ego much?

>What usually makes Linux painful or >impossible for the average Joe or Jane user >is the task of installation and >configuration.

Does the average Joe install Windows?
I have to believe that of my non-tech relate friends, most people I know would never install Windows or know what Device Manager is.
I've been building my own PC's for years and I have no problem on either Mac,Linux or Win but having installed a lot of dual boots over the past year (4 of our own computers and another dozen for family and games are a must) or so, Ive done the Linux side easier and quicker.

I wouldnt let my mother install either but for her first computer at age 72, I gave her PCLinusOS. My dad who is older is a Windows vet, having used a computer for at least 4 years is on kUbuntu.

@zeke: Of course installing Windows isn't easier than installing Linux. The point, which you seem to have utterly missed, is that Windows comes preinstalled on most computers sold retail so it doesn't have to be installed by Joe or Jane user. Prior to the EeePC Linux didn't and most of the time still doesn't.

Also, please don't put words in my mouth. I NEVER said that anyone who disagrees with me has an agenda. What I have said is that fanatics on either side of the issue are likely to attack me, something which has happened repeatedly in the two years I've written this blog for O'Reilly. It has nothing to do with ego or arrogance except in your hasty judgment of me.

The fanatics always make things harder, and now we have a small war betwen distributions. I belive that anyone should use what thry want and we just remaind them that are alternatives and they should try those. I want a more pozitive energy in the linux comunity, we should not fight against someone just ask for fairness(i just want drivers and Microsoft to play fair,in Europe they have to pay for theyir attitude and this should happen evrywhere)

Good article and straight to the point cuts through the BS. Linux scales down good is modular as the writer points out and you can find many flavors ie (Distributions) that are small and work good on old machines while still getting support, also virus and spyware proof so you don't have to bloat the system with fat spyware & virus programs wich can and most often slows a system down. Linux needs none of these adding to it's speed. I would not choose a OS just for speed but for stability, security, the ability to run the apps that I use and yes performance is a plus . For me the clear winner is Linux

Re: Lin vs Win Metrics: How incredibly RARE and refreshing: an evenhanded, objective article that looks at the issue realistically, cutting through agendas and biases to give an accurate and reliable description of the topic. I'm beside myself!

Unfortunately, to paraphrase a phrase, 'well behaved writers never make history'. Maybe if you could communicate the same message in l33t speak or IT Marketdroid jargon there'd be some hope for fame and lasting influence.

Thanks for the article- it was good to see at least ONE reputable writer out there has a grip.

-Arawn

"Windows is usually the best fit in three situations."

Let's analyze these three and see where considerations might be missed.

"...when the migration costs are exorbitant. Training costs, software porting, new licenses for commercial software that can't easily be replaced by FLOSS equivalents, and the cost of testing and deploying new code are all issues that can work in Microsoft's favor. In such situations if a customer is seriously interested in moving away from Windows I recommend doing so when hardware and/or software need to be refreshed anyway to mitigate some of the costs."

Your recommendation is good, but there are other factors. The problem with treating the negative factors as overwhelming is in ignoring the opportunities costs. Business people have a very tough time doing this, so they do this all the time.

The reality is that most businesses aren't going anywhere (absent the circumstance where they can expect to fail.) Thus, the issues of training, software porting, and replacement development can be extended out over time. Many organizations seem to feel that they need to have everything converted over and done with in ninety days. This is pointless.

The other reality is that the longer you wait to migrate, the more it's going to cost you NOT to migrate, both in the direct costs you're absorbing now to deal with Windows headaches, and in the opportunity for lower ongoing costs in the future.

No matter how much it costs to migrate, those costs will be absorbed and amortized over time. Not so with the perpetual costs of licensing, security, reliability, and support costs which can only go up with any proprietary solution.

Training costs are also frequently overrated. Many companies have switched to Linux without extensive training expenses. It all depends on the quality of the staff, and the method of migration, and the uses to which the systems are being put.

The key is to MAKE A PLAN. Start now to plan how the migration will go. What needs to be done, what are the lowest cost solutions, how can the organization minimize the need for training and porting. There are solutions to all those issues if the company looks for them. If the company can't find them on its own, hire a consultant. While it's hard to find a consultant who's good at digging up solutions even he doesn't know about at the moment, when you get someone like that, they can save you big bucks.

"The second situation, which applies equally to the server and the desktop, is when a very specialized piece of software is needed for the business and it's only available for Windows. To use a desktop example: there still isn't an adequate replacement for Autocad that runs on Linux. This scenario is becoming less and less common but it still does exist, more so in the business world than on the consumer desktop we discussed in the comments on my previous article. The best use for Windows in the corporate server room, the only really good use I can think of, is as a specialized applications server."

And yet, there are solutions for that, too. The question boils down to: how many people and how many licenses for the specialized app do you need and what is the cost factor in comparison with the OTHER costs of running Windows?

I have a client in exactly this position. They do digital conversion of consumer media - still, film, and video. They use Windows and Adobe products exclusively because their staff is trained on them, and there are no equivalent products in the Linux space for things like Adobe Premiere. The owner would switch to Linux in a heartbeat if there were, because he knows Linux is more reliable.

There is no Linux solution to their situation at the moment. Even Cinelerra couldn't handle their needs, particularly since their film capture methods are Windows specific. There are also specific hardware metrics that have to be met in their workflow processes, which eliminate using virtualization of Windows on top of Linux as a solution.

So now they're opening a new store in another city - and they are testing a scenario of switching to Macs. This is likely to raise their hardware costs, but significantly improve their production work flow, with less Windows crashes and less Adobe crashes (Adobe software sucks), which will enable them to make more money.

But many companies may not be in so tight a situation that solutions such as virtualizing Windows on Linux (to improve reliability and security), or the use of application servers and/or thin clients couldn't allow access to Windows specific mission critical software sufficiently to allow the replacement of their other systems with Linux.

"The third scenario is when a business has a customer who demands Windows and not fulfilling that demand will cost the company a valuable customer. In those scenarios a company can try and sell a Linux solution by pointing out advantages over Windows. If the customer isn't buying then you have to weigh the cost of losing that customer. It's often many customers, not just one. In the end the business has no choice but to satisfy customer demand even if the customers are making a poor or uneducated choice. Money talks."

This really isn't much of an argument against switching to Linux because the same argument could be made for a company demanding a switch from Windows or proprietary UNIX to Linux. That the customer isn't always right, but he's still the customer is certainly true, but it's not an argument. If you're a Linux consultant and the customer wants Windows, walk away from the assignment. The same is true for a Windows consultant. If you're both, it's not an issue.

But before walking away, determine whether the customer has a legitimate argument, or is just working on hearsay, as most management does. Just as most doctors get their medical advances information from drug company salesmen, most IT management get their information from IT company sales pitches. Microsoft is known for this sort of thing. Their whole "Get the Facts" campaign is based on that knowledge.

What most consultants know is that if you provide a seriously larger ROI or smaller cost of implementation to a customer, many times that customer will not care what the solution entails. And as many of the Linux migration case studies have shown, a company can save really massive amounts of money - even up front money, let alone opportunities costs - by switching to Linux. What matters is how you make the case.

Of course, there will always be Windows fanatics, just as there have always been IBM fanatics in the old days.

In short, the situations where Linux cannot be used in place of Windows are very narrow and getting narrower all the time. It's almost always just a question of how - and that question usually has answers if you look hard enough.

@Richard: Your arguments refuting migration costs make sense but many business (most businesses?) are bottom line oriented IN THE SHORT TERM. Any expense that can be postponed is postponed, particularly in a recessionary economy. IT departments are short staffed DELIBERATELY as short sighted management doesn't see IT as a profit center. I can argue the advantages of a Linux migration until I'm blue in the face but if the up front cost is significant most businesses where I live just say "no". You also have a lot of IT management that are sold on Microsoft and it is very hard to convince them to consider an alternative in the first place. You start talking about needing more develoopers and admins, even as contractors for a relatively short period, and the wall is up.

In regard to your refuting my second argument: If a company is dependent on a Windows-based application with no alternative you are going to have Windows in the shop, period. My goal in such shops is to make sure Windows doesn't creep into areas where it isn't needed and where Linux is a better fit.

In regard to the third argument: Again, many business managers are sold on Microsoft and some small time consulting company isn't going to convince them otherwise. The cost of losing customer is very real. There is little or no reasonable argument you can make here that changes reakity.

I would dearly love to be an exclusively Linux/UNIX consultant. In the market where I am located the large companies go with large IT outsourcing firms. IBM Global Services has a huge presence here. My opportunities are mainly in small and medium sized businesses which are either heterogeneous shops or strictly Microsoft. I have yet to run into even one business that didn't have Windows somewhere. Not one. If I am unwilling to support their current infrastructure, at least through a migration period, then I won't get the business. If I insist on Linux only my business would fold very quickly. That's reality.

There are many ways to compare operating systems, but one no-BS-way is to simply install them on low-end hardware. One of the town halls I'm working in - in a remote place in South France - had a series of battered but solid PIII-500 Compaqs with 128 MB RAM. They were told before that they would have to replace them, and they didn't know when and how, because budgets were tight.

In a second hand computer shop, I found some addition RAM, so they now have between 192 and 256 MB RAM. And actually they're running a nicely configured Arch Linux with XFCE and a reasonable choice of recent and even bleeding edge software: Firefox3b5, Thunderbird, Sunbird, OpenOffice.org, AMSN, ... OK, OpenOffice.org takes some time to startup, but once it's there, you can work with it.

I missed the original article and subsequent flamewar but I agree with what your saying, benchmarking Windows vs Linux is a very tricky thing as there are so many different applications to compare. Do you go Pidgin Vs MSN Messenger or something more lightweight in the Windows camp? The only real way I could think of to do it would be attack the underlying OS itself which falls down to boot speed, file write speed, shutdown time and perhaps a somewhat generic 'time to open app X'.

It's not surprising the Linux build on an EEE won that sort of contest, by all accounts Xandros has been heavily tweaked for the diminutive machine whereas Windows XP hasn't or at least not to the same level.

"It's not surprising the Linux build on an EEE won that sort of contest, by all accounts Xandros has been heavily tweaked for the diminutive machine whereas Windows XP hasn't or at least not to the same level."

Another way to look at it is Linux is infinitely tweakable, and it's not very hard- anyone willing to study a bit (start with Linux From Scratch) can do it. Can't do that with poor old Windows. It needs help. Lots and lots and lots and lots and lots of help. And nobody is allowed to even try except designated persons from the mother ship.

At work I use Windows XP, with MS Office and OpenOffice. I find MS Office to be much faster, and especially documents with many images are handled slowly by OO Writer. On my EEE, with Xandros, OpenOffice is a new experience. It feels like it should. I bought the EEE because I wanted a small laptop, and I see it as an opportunity to learn to use Linux. The small screen is on the one hand a problem, on the other a challenge to find ways to use the screen and keyboard in an optimal way.

In the end, thát will make my use of the system much faster and easier. I learn how to use keyboard shortcuts, and now I'm trying to configure the system in such a way that I can handle most tasks without using the mouse or touchpad.

Could I do this with XP on the EEE? Probably! But I don't want to. I want to learn to use Linux and this is an ideal opportunity. Furthermore, now I see the power of OpenOffice for the first time. I removed the button bar, and am busy configuring keyboard shortcuts for those buttons. It looks more like Wordperfect 5.1 now. I wouldn't have done this on my Windows machine.

So for me it's not at all about which system is faster. It's about how I can learn to work faster, or differently. And it's about learning to use another OS. I've used Linux before, but never for a long time for real use. Those are the wins. It will take considerable time to get where I want to be, but then I can make a real judgement.

@Sarah & @Carla: I don't believe you are quite correct. I believe XP has been tweaked as far as XP can be tweaked on the Eee PC. I don't think Xandros has an advantage there other than, as Carla pointed out, the fact that Linux can be fine tuned to a much greater extent than Windows can be.

@Niki Kovacs: A test on legacy hardware, like the test on the Eee PC, has limited value. It says a lot about the ability to scale down and work on older hardware. It says nothing about performance in other environments, for example running 64-bit code on a high end system. This is my problem with all metrics that proclaim one OS faster than another -- the results are limited to the specific environment in which the tests are conducted and they cannot be extrapolated to cover the full range of systems out there.

@Roger: In my experience once people live in Linux, as in use it as their primary operating system for an extended time, they end up reaching a point where they can't imagine how they ever got by with Windows. The Eee PC, as nice as it is, isn't giving you the full experience you'd have on a higher end machine. Still, it's getting you comfortable with the OS and serving a purpose for you and that's a great start.

I own an Eeepc 701 4GB computer that shipped with the Xandros Desktop pre-installed. It has an Intel 900MHz processor that clocks at 630MHz with a 70MHz FSB. It is my understanding from frequenting forum.eeeuser.com that other Xandros equipped models clock at around 570MHz. The version of the 701 4GB Eeepc that ships with Windows XP pre-installed features a 400MHz FSB and clocks at 900MHz.

Comparing the original Xandros equipped 701 to the Windows XP enhanced version of the 701 seems irrelevant due to the differences in hardware.

@James: Actually, you've shown that the metrics are more relevant that I would have guessed. Xandros, running on inferior hardware, is faster than Windows running on superior hardware for a given, admittedly limited set of tasks.

Another interesting issue: It would be great to get a Windows EeePC, load a fast Linux distro on it (i.e.: Vector Linux Light), tweak it and optimize it to the max, and then run some metrics. That woudl be the way to get the fastest possible EeePC. Hmmm...