Fog Creek Software
Discussion Board




Windows Server vs. Linux market share


How well do folks on this forum think Windows will do on the server side vs. Linux over the next five years?

YAMUJOSN
Monday, June 30, 2003

A large portion of .NET is targetted directly at J2EE. Microsoft is trying to leverage it to increase their server market share. How well they do is at least partially tied to how well the .NET adoption goes.

.NET is a very productive development environment, definitely the most productive I've ever used for web development. It's leaps ahead of the competition (not without its flaws, of course). The question is whether or not that that's enough to tip the fence-sitters. I think it might be.

Brad Wilson (dotnetguy.techieswithcats.com)
Monday, June 30, 2003

I think the shining light that is linux will continue to shine in the server space for the next five years and beyond.  After it has eaten up traditional unix share, it might start eating up Windows More Advanced Server 2005 share too.

i like i
Monday, June 30, 2003

If the choice between Linux and Windows boiled down to purely technical measures (e.g. stability, performance, features, hardware support, etc.) then Linux would "win" over time. But MS aren't sitting still and I see evidence of several flanking manoeuvres aimed at keeping Linux out of the game:

1. Lobby law-makers to make reverse-engineering illegal.
2. Keep Apple on life support. Contrary to a recent report (in Wired?), I see Apple hurting Linux adoption rather than vice versa. Once the bulk of the Windows exodus has found a new home with Apple, they can be "attacked" using time-honoured anti-competition.
3. Work to make the PC platform less open. E.g. Palladium. Even better if MS owns the IP.
4. Finally, work to dismember the Internet as a free communications medium. By advocating, oh, say, that the W3C promote patented protocols.

What is interesting is that the Linux community is trying to outcode Microsoft when they should actually be out-lobbying them instead. Unfortunately, this is impossible since political lobbying requires deep pockets and most people running Linux forgot to get rich first (with the notable exception of IBM).

The wildcard in all of this is that MS have stabbed their partners often enough that the necessary industry support may not actually come their way. This could give Linux all the breathing space it needs.

Paul Sharples
Monday, June 30, 2003

"If the choice between Linux and Windows boiled down to purely technical measures (e.g. stability, performance, features, hardware support, etc.) then Linux would "win" over time. "

Why do you believe this to be true? So far they are AFAIK Linux is loosing on all four of those fronts.
Now that people are holding the OSS movement up to the candlelight, the bubble seems to be bursting.

Just me (Sir to you)
Monday, June 30, 2003

Linux stability: windows is closing, but I still need reboots when installing etc

Linux performance: windows is closing, but the big difference is really in hardware

Linux features: server features me thinks..

Linux hardware support: just you install windows on your sun box..

i like i
Monday, June 30, 2003

Linux hardware support -
Let's see... sun box... sun box... Nope. No Sun boxen here. Let me install Linux on the boxes I *do* have. Well, last time I tried I had to recompile the kernel to support USB, which was the end of Linux for me.

Do wireless LAN cards work in Linux yet? When I tried (eight months ago) it was "you can't get there from here"

Oh yes, admittedly since Linux is open source I could've written my own drivers

GET REAL

Let's see - spend month of my own time learning linux source code so I can write drivers for a wlan card OR spend the equivalent of three hours' income for Windows 2000 and plug the thing in.

I think the true benefit of OSS, especially at the enterprise level, is that it's supported by a community. The "if you need something fixed, you can fix it" is a red herring. (BTW, I use the term "supported" loosely - I didn't mean like tech support, since getting help from the much-vaunted Linux community is like pulling teeth)

Philo

Philo
Monday, June 30, 2003

"Linux stability: windows is closing, but I still need reboots when installing etc"

Does it?

John Topley (www.johntopley.com)
Monday, June 30, 2003

OSS, including Linux, will hit an iceberg in a few years once people realise you can't eat and write free software at the same time.

titantic linux
Monday, June 30, 2003

Philo wrote: "I think the true benefit of OSS, especially at the enterprise level, is that it's supported by a community. The "if you need something fixed, you can fix it" is a red herring. (BTW, I use the term "supported" loosely - I didn't mean like tech support, since getting help from the much-vaunted Linux community is like pulling teeth)"

YMMV. In my experience it depends a lot on how you ask the question (show that you have put some efort into finding a solution) and on what subject you ask a question. Example: the JBoss user list is very helpful. Core developers read it and reply, usually within a day or so. These people really know the gritty details of the code and are able to provide in-depth answers.

This might (or might not) be an isolated example, but I've always had more trouble when trying to get help from a 'professional' paid-for helpdesk where you usually talk to somebody who is able to take a message, mangle it and send it on to someone who might or might not answer.

OSS is not the answer to each and every problem, but in certain areas it really is better than closed source offerings. The usual example is of course Apache. MS has not been able to kill off, Apache, despite including IIS with a lot of their offerings. This might tell you something about the quality of Apache.

The OSS pool has a lot of what i like to call component software: tools that help to build certain applications. The real end user applications are usually not as good as the developer oriented stuff.

Jeroen
Monday, June 30, 2003

A rather interesting point that has not been brought up here is the issue of centralized support.

There are many good aspects of a having a central company to support a product. And, I would freely admit there are also many downsides.

Right now a good number of companies do allow automatic updates to their Virus software. However, for windows updates and security patches it is still a matter of large debate. Some sysadmins do allow auto updates of patches etc to windows. However, a very large number do not.

However, as this technology to “auto update” expands, and the technology gets better and better, then a commercial company that controls those updates tends to have a rather large advantage.

I mean, now windows can auto update your office install (first windows...now office!). 

Clearly, more and more windows products will be able to be auto-updated without user intervention.

I mean, if I have 20 computers, and need to install a new OS patch, why in the world should I have to hire someone to go from PC to PC and install that patch? (or, if I got a good admin, he will write a vbscript and roll out the patch to all the computers via remote scripting).  (actually this problem is that windows scripting can be too complex for many batch type admins). If you can code real kick butt windows scripts...you are probably programming for a living!

If a new Outlook vulnerability is found, why again would I have to go to each PC, and install a update? In this day of cheap communication, it is cave man barbarism to have humans do such lowly work.

I mean, most virus software now does it’s own updates. I mean, before some guy would run around and install the new virus update. How dumb can you get?

Regardless, it makes no sense to have some guy in a T-shirt run around updating and patching computers.

Automation updates and installs of patches etc to software is a large cost to companies. Windows update is capable of reducing these costs by a lot.

This does NOT mean that Linux vendors will not jump on this bandwagon, but it is a area that needs serious attention for Linux.

As a commercial venture, MS is now spending a lot of time on reducing the human cost in running a PC. We don’t need a bunch of geeks coding in C++ and pearl to run a PC today. Heck, we don’t even need knowledge of DOS to run a PC (you certainly did many years ago).

Heck, you also had to be a mechanic to run those old cars too!

Look how many companies have jumped on the automatic payroll deposit for employees. I am now seeing very small companies deposit their employees checks electronically. Why print out all those pay checks etc?

The same concept applies to OS and software updates. It is silly to have people doing this. If all os and software update can be done un-attended, then they should be.

The idea that computers will need a un-shaven person in a t-shirt running around to shove in a floppy disk, or on each pc have to run some udpate is going the way of the do-do bird.

You can even see the new .net 2003 commercials running on TV, and they also stress this reduced human intervention here. We are seeing a new class of systems that can update their software.

Since the issue of TCO is big, those commercial companies are going to want to charge you, but they will now show reduced numbers in people support.

If you think about this for a bit..you will see how people cost in the computer equation will be SUBSTANTIALITY reduced as time passes.


Albert D. Kallal
Edmonton, Alberta Canada
kallal@msn.com
http://www.attcanada.net/~kallal.msn

Albert D. Kallal
Monday, June 30, 2003

"A rather interesting point that has not been brought up here is the issue of centralized support."

I get 99% of my answers from the internet. For both MS and Linux. What's interesting is the difference in info I find. It seems to me that MS community support on the internet usually covers the bottom 75% of "things worth knowing." So it's easy to get *into* a technology, but once you get fairly advanced you're out on your own.

OTOH, it seems like Linux and OSS generally cover the *upper* 75% of "things worth knowing." It's rare to find a very simple step by step "so you've installed [x]? Click this button first" document in the Linux world.

Case in point - though I think someone's finally documented it, for a long time there was no simple explanation of Linux file management (what goes where and why).

"I mean, if I have 20 computers, and need to install a new OS patch, why in the world should I have to hire someone to go from PC to PC and install that patch?"

Because you want your employees to be able to work? Autopatching is EVIL. It's 2pm the day before a big demo and MS releases a patch that breaks your dev tools. It doesn't matter whose fault it is - you're screwed.

Large deployments are why SMS was invented (the MS product, not text messaging).

Philo

Philo
Monday, June 30, 2003

>Autopatching is EVIL

Well, that is like saying because we have one bad police officer, we should throw out the whole legal system.

IF (and I stress “IF”) Autopaching is done right...then everyone will go with it.

I mean, many people complained bitterly when the auto industry started removing points in the ignition systems for car. That is because people had more control,  and with knowledge in a pinch could repair the car on their own.
(at least you could if you know something about cars).

However, electronic ignition won the day because it was more reliable, and did not break, or need to be replaced. It also did not need the mechanic either. Of course, the down side was that you COULD NOT fix the unit your self.

The issue here is not Auto patching, but rather  GOOD auto patching VS bad Auto patching.

I am simply saying that the industry better be aware that this technology is coming of age. The use of auto updating is gaining momentum.

This is simply a trend that will only increase in use. The commercial companies realize this, and thus again this a driving force behind this technology.

Albert D. Kallal
Edmonton, Alberta Canada
kallal@msn.com
http://www.attcanada.net/~kallal.msn

Albert D. Kallal
Monday, June 30, 2003

I should add here that it might be 5 PM and everyone goes home. The tech guy stays behind and then decides to update all the pc’s.

Of course, the next morning, the staff comes in to find nothing that works.

The problem here again is not electric updating, but simply that a bad update was done.

Further, perhaps the tech guy burned a cd on Monday, and spent all day Tuesday installing the patch.

When done electronically the first report of a problem can halt the process. Half way through Monday the process might be stopped.

Certainly, the difference between auto updates, and NOT SCHEDULING updates for a important day is a legitimate argument.

However, maybe by auto installing the patch, the system DID NOT go down in the morning do to some new virus or whatever.  This argument goes both ways.

This whole thing revolves around if it makes sense to have a update to occur or not.

Perhaps a company might turn off updates if some special event is to occur the next day...but then again..they might miss out a important update also.

Sure, some mans to schedule, or turn off the updates can be made available......


Albert D. Kallal
Edmonton, Alberta Canada
kallal@msn.com
http://www.attcanada.net/~kallal.msn

Albert D. Kallal
Monday, June 30, 2003


>Let's see - spend month of my own time learning linux source code so I can write drivers for a wlan card

Get Real!!

You expect it for free, you also *EXPECT* somebody else to write the drivers for you? How lazy are you?
You get what you pay for. Don't whine about linux being $whatever, you have the source, you can do it
yourself, you are crying about somebody else not doing it, yet you aren't doing it. So you either code, or shutup.
Use windows.

This is a fairly off topic conversation, you'll always have people on both sides, which is fine, but why fight
over it, it's just a rehashed conversation. Use whatever is best for you, leave others make that choice too.

But..don't complain about linux/*bsd/$whatever not being able to do something, speak with code, not childish crying.

fw
Monday, June 30, 2003

And that, my friends, is why Linux will never beat Windows on the desktop.


Monday, June 30, 2003

Wasnt this supposed to be about servers?


I think we will see alot more linux in the future. There are many reasons, but one that dosnt get alot of attention is that linux is l337. Thousands of little geek-lets around the world use it, and develop for it for no other reason than bragging rights.
Windows may have the desktop, but Linux will have the hearts of the next generation of developers and engineers.

Eric DeBois
Monday, June 30, 2003

fw - the comment was that Linux bested Windows in the hardware arena. I was challenging that assertion.

As for "you get what you pay for" - you just asserted that Linux is crap because it's free. If that's the way you want to go, then fine. *I* think it's a pretty good OS, especially considering its heritage. But I also think many of its selling points are overrated (the "linux community" and "you've got the source, do what you want").

I like Linux because it's a more efficient operating system than Windows and its networking tools are second to none.

Philo

Philo
Monday, June 30, 2003

Albert:
Good Autopatching: SMS
Bad Autopatching: Virus-update-style auto update with no controls

I wasn't aware that Windows XP allowed sysadmins to remotely schedule the patch updates - I thought it was either a) automatically patch when the patch is available, or b) the user patches when they're ready, from their desktop

Philo

Philo
Monday, June 30, 2003

>You expect it for free, you also *EXPECT* somebody else to write the drivers for you? How lazy are you?
You get what you pay for.

You're right - I mean I could spend $110 on Windows 2000/XP Pro and get it to work in under 3 hours as opposed to d*cking around with Linux for 2 days to get something trivial to work.  Linux is slowly getting there, but Windows is already there.  That's why Microsoft isn't going out of business anytime soon.


Monday, June 30, 2003

WTF is l337?!

John Topley (www.johntopley.com)
Monday, June 30, 2003

It's "leet" as in "elite".

BC
Monday, June 30, 2003

Yeah, the kids use it alot.

My favorite word fron the new psudo-hacker-kid-lingo is PWNS which somehow evolved out of owns. In counter strike and other online games, someone might say "I PWNS Y000!" when they shoot someone.

Another good one is fuxxored which hardly needs an explanation. "Windose is fuxxored, linux pwns windose"

It dosnt really work when I do it though. I guess Im too old. :D

Eric DeBois
Monday, June 30, 2003

"WTF is l337?! "


Its the reason the children of today perform so well in school....

apw
Monday, June 30, 2003

Whew, where to begin.

Brad: ASP.NET is not a Windows-specific technology. The Mono project has a version that runs under Apache. It's already decent and getting better. The Mono C# compiler surpasses Microsoft's in some areas (example: support for generics or "templates", which was added to the C# spec but Microsoft has not yet implemented). http://go-mono.com/asp-net.html

Albert: Commercial Linux distributions do have auto-update features, similar to Windows Update, and which can be run in unattended mode if desired. Also, an administrator of a Linux network would not have to run around to every computer to install the update: it could be done remotely since Unix has always had strong remote administration features. But if you prefer auto-updating a-la antivirus programs, you can do that too. http://www.redhat.com/software/whichnetwork/

Philo: "Recompiling the kernel" sounds scary but is not that difficult. When I recompiled a FreeBSD kernel to support a wireless LAN card (almost 2 years ago -- I'd be surprised if Linux doesn't have robust WLAN support) it was as simple as uncommenting a line in a configuration file, running a make command, and rebooting. Not as easy as Windows driver installation, but not hard either.

fw: You're an idiot.

And finally, to be on topic: I think servers will continue to specialize. For example, NAS (appliance fileservers) are taking over the low end of the market. Who cares if the underlying operating system is Linux, Windows, or something else.

Network printers used to need dedicated print server computers; now that's built into your laser printer. If you need a web server, you buy one that supports whatever platform your developing for. Support for the platform (J2EE, ASP.NET, whatever) is the deciding factor, the OS choice is secondary, or is determined by your platform.

I don't think you can expect future networks to be heterogeneous Microsoft-only, or Linux-only, things. And I think servers will be less visible as they mutate into appliances and turnkey systems.

Nate Silva
Monday, June 30, 2003

"homogenous" not "heterogeneous"

Nate Silva
Monday, June 30, 2003

yeah...what nate said.

I would like to add though that I haven't seen such a stereotypical linux geek as fw in an awfully long time. 
<g> if I believed in conspiracy theories Id suggest he was a creation of the other side...

FullNameRequired
Monday, June 30, 2003

- There are enough companies that provide commercial support for linux (HP and IBM for example). Just buy it.

- just try to get Microsoft to fix bugs for you in Windows or one of their products (e.g. VC++ had very annoying bugs for years). The support situation for specific problems is worse for windows than it is for linux. And there's only one company that has the source...

- about automatic updates: Debian has had them for years (apt-get). The updates are fine-grained, you have multiple distributions ("stable" for servers with conservative update policies, "testing" and "unstable" for less critical systems). Updates can be retrieved from any number of Debian package servers, for updating all machines on a LAN, tools are provided for mirroring Debian package servers locally and redirecting clients to the mirror. The free BSD systems have similarly powerful package systems. I dare to predict that Windows will *never* have anything that comes even close to this infrastructure.

- USB -- well, the last time I tried to use a USB memory stick on a Windows XP machine, it just *didn't work* (yellow quesion marks in the device manager). It *just worked* on another box (and under Linux on the same box). Given the intransparent nature of Windows, you have a hard time fixing such problems (I gave up after a short time). Under Linux, use a kernel wit USB support, and you're done (if not, you can reproduce and track down problems).

multi_io
Monday, June 30, 2003

Nate - I've recompiled the kernel before. What I hate about Linux (and what Linux folks hate about Windows) is the learning curve issue - unless you're really lucky, every step of recompiling the kernel means learning another chapter of material.

However, as I mentioned before - Linux seems particularly averse to helping newbies learn to walk. Sure if you're walking and you want to run then they're everywhere, but while trying to crawl and learning to walk I sure seemed to get a lot of
a) "If you can't be bothered to learn the basics..." (note - that's why I was asking, dipstick)
b) "Well excuse me, if you can't handle Linux maybe you should go back to Windows"

I eventually followed the advice given in (b).

Philo

Philo
Monday, June 30, 2003

Hi Philo,

"What I hate about Linux (and what Linux folks hate about Windows) is the learning curve issue"

totally agree, it sucks.  IMO no user except those with _very_ advanced or unusual requirements should ever have to look at a line of code or fight their way through the complicated bits of Linux.
Ultimately I think that companies with their own distro (redhat etc) should be handling this kind of stuff themselves, working as a kind of bridge between the raw linux coders and the end users, smoothing over the edges.
They all do this to a certain extend, but mostly still do not go far enough IMO....OTOH they are all improving over time.


"Linux seems particularly averse to helping newbies learn to walk"

depends a lot on where you go, which mailing list were you asking for help from?

The kernel mailing list is best to avoid unless you are asking a sufficiently advanced question, IIRC there are a number of newbie type mailing lists around which would probably have been able (and a lot more willing ;)  to help you.

FullNameRequired
Monday, June 30, 2003

Linux does have a shallow learning curve (I used to say "steep learning curve" but apparently that is backwards). It's not easy to get started but once you get past a certain stage things become much easier. Having managed both Windows and Unix servers for several years I can say that my time is more productive with Unix/Linux/FreeBSD.  And I am by no means a Unix guru. But the barrier to "try out" Linux is still too high.*

* Except in the case of "appliance" devices, which was my prediction for the future: more appliance and turnkey servers that require no direct management. The OS matters less. I'm talking servers, not desktops.

Nate Silva
Monday, June 30, 2003

Philo said "I eventually followed the advice given in (b)."

I concur.  I am fast reaching that point myself.  *nix does seem to be a bit of patronizing from a crusty old grandpa.
"Why in my day, we had to make our own zeros and ones, you didn't have the computer do it for you... And we like it that way." 

Mike
Monday, June 30, 2003

"I'm talking servers, not desktops."

totally agree, the server market is where Linux _currently_ has it all over Windows.  Windows still wins in the desktop arena.

Personally I dont care much which OS comes out in front in either the desktop or server arenas. 

<g> the bottom line is that Ill _still_ have to fix the things when they break.

FullNameRequired
Monday, June 30, 2003

"What is interesting is that the Linux community is trying to outcode Microsoft when they should actually be out-lobbying them instead."

I'd like to quote here the cyborg from T3 "Desire is irrelevant. I am a machine".

Lobbying is irrelevant. Since Linux is easy to introduce in a step by step fasion without upfront monetary costs (discarding learning expenses), it will surely make its way into shops of all kinds. There is no turning back for this kind of process.

The situation is likely to reach a stable split similar to Apache/IIS. A certain fraction will be running Linux while the rest will run whatever version Windows. Windows users will gain a lot since Linux should keep MS in check.

Mr Curiousity
Tuesday, July 01, 2003

I agree with Nate. Most servers are going to morph into appliances and you will be buying them to serve a specific function.

I do think that the appliance market will be dominated by Linux simply due to cost. When you deal with units, cost per unit becomes a driving force for almost every choice.

That said, I use Windows. I took the (b) pill like Philo.

Marc
Tuesday, July 01, 2003

I expect Linux to peak in two years and then start a slow decline as Unix currently is.  Why?  Because Linux is just Unix redux, and Unix was already dying except at the high end in the data center.

Mike
Tuesday, July 01, 2003

Hi Mike,

"I expect Linux to peak in two years and then start a slow decline"

are you interested in having a wee side bet?  if so, let me know and we can figure out a good way of exchanging emails :)

anyone else want to make the same bet?
Ill take any bets up to a reasonable total...

FullNameRequired
Tuesday, July 01, 2003

Mike, I think your analysis is missing some important points.

Yes Linux is a flavor of Unix, but it has significant differences.  The first and possibly most important of these is that it runs on cheap PC hardware.  How much of Unix's market decline is because of the OS itself, and how much is because it usually comes on proprietary hardware?  And truthfully, no Solaris machine I have ever used comes close to the speed of a modern PC.  Now maybe that is just a poor generalization based on using older Sun boxes, but the difference is there.  Now with Linux you can get the best of both worlds if you want it.  A Unix-like system, running on cheap and fast PC hardware.

There is also a lot of appeal to the intangibles of Linux.  Many people (especially outside of the US) are attracted to Linux because it is not controlled by a US company.

Back to a previous topic now:  one big problem I've always had with Linux is that it doesn't separate *administering* the system from *using* the system.  When used as a desktop, the user must also be the administrator (not in terms of accounts, but in terms of the actual person using the machine).  You can't get up and running and use a Linux desktop without being a system admin.  Now, it could be that things have improved since I last set up Linux, but I think this really contributes to the problem of Linux being unfriendly to new users.  From a strict *usability* standpoint, I think Linux isn't bad.  I use it every day at work and for the most part find it quite good.  But in order to get to the point where you can do that you have to do a lot of low level configuration.

This same idea permeates through Linux, it isn't just at setup time.  Installing programs, adding hardware and upgrading drivers, etc all suffer from the same problem.  Sure it is nice to have some of those options if you're a power user, but it just raises the barrier to entry for new users.  The majority of desktop users want something that just plain *works*.  They don't want to have to go through a huge configuration hassle to install a program.  They don't want to have to recompile their kernel to add USB support.  They don't want to have to search newsgroups for hours trying to find what strange incantation to use to get their network card working.  Windows is far from perfect in this respect (though with 2000 and XP it is very good, in my opinion), but it is still far better than any Linux system I've used.

Mike McNertney
Tuesday, July 01, 2003

What, IMO really kills linux from and end user perspective is the relatvely steep learning curve. Sometimes jst getting linux installed and running can be a real hassle, esp. for someone who is computer illiterate, aka the end user (employees, etc).  Case in point: I installed Redhat 9 jsut the other day, and to my surprise, it did not support the **generic** 2 button mouse I was using, or the 2 + wheel mouse I also had. Now, for me this is not that serious. However, for an end user this is disastrous, because it also caused keyboard conflict, meaning neither the mouse nor the keyboard would work in X.
This is the primary fault of linux, as stated above, ahrdware support is lacking, therefore making it hard for end users to be able to use the system on their own, esp with new devices.
What **would** definately help linux is if end user installs such as plug on and drivers came in a more friendly package that automatically detected settings and installed when downloading, something similar to Red Hat's Network's auto update function. This would definately help to make it a viable alternative to windows in the desktop market. I also believe that when you install linux distro's and create users as admin, they should be fairly locked out and should a different working interface which would more cater to their needs.
As for the server market, linux is much MORE appealing, becasue of it's general versatily. If a network admin (me) demands unmatched performance for web applications or serving, he/she can recompile the  kernel taking out everything that isn't necessary for that specific situation, and definately achieving a level of performance which is nearly impossible to achieve on any other system. Case in point, I have recompiled the 2.4 kernel and apache 2.0 along with MySql and a couple of other apps to under 50 megs, and the performance is ASTOUNDING. This saves me from having to have the Army spend a lot of money on hardware, and on client liscensing fees (where Microsoft really screws you). I can run a server with much cheaper hardware and still achieve better performance.
Still, as of right now Windows is Definately a better interface for the computer weak of heart. For the rest of us, who don't mind WORKING for our money, linux can definately have it's advantages from a business perspective..

ArmyAdministrator
Wednesday, March 17, 2004

*  Recent Topics

*  Fog Creek Home