Overall I thought the article was well-done. In the PC desktop space it's certainly true that Linux will not displace Windows anytime soon, and IMO it never will due to the entrenced market.
But Microsoft is wrong that Linux has not affected sales of NT. A lot of historical UNIX workstation users such as myself switched to NT in order to take advantage of the cheap PC hardware and easy software availability. I jumped onto NT almost immediately because it was good enough and a lot cheaper than your typical RISC workstation. It took a lot of effort to make a usable development environment, but you could do well enough.
Over the last few months, however, Linux has started to make significant inroads into our engineering department. Today roughly a quarter of all engineers run Linux rather than NT -- up from just two in the spring (that's about 700% growth). Microsoft can't see that in their sales yet because IT is still buying NT licenses whether or not we use them, but the transition is happening remarkably fast -- far faster than the RISC->PC movement that brought in NT. It will not be long before new NT purchases are curtailed to some degree.
There have certainly been some NT sales lost to Linux, however. I recently purchased a new laptop and I chose to go with Linux rather than NT despite running NT exclusively for the last four years. That's a sale NT would have gotten only a few months ago.
Generally speaking Gartner is correct that the people jumping onto Linux are traditional UNIX users. That means that on the desktop Microsoft is pretty safe -- there never were that many UNIX users.
But on the server, well, that's another thing entirely. Microsoft is not particularly well-entrenched; they're competing well against the commercial UNIXen at the low-end but that's largely because the hardware is cheaper. Linux uses the same hardware, costs less, and doesn't have client license issues. That 3.5% market that is dismissed in the article was statistically insignificant only a year ago!
Gartner is right that the traditional UNIX vendors are in the most trouble, but they were already in trouble. Their problem isn't and never was Microsoft, it was the PC. Linux gives those UNIX people a lower-cost solution without giving up UNIX. For small servers it's a very interesting solution to an IT department that is UNIX-savvy.
Microsoft makes great noise about how Linux isn't hurting them in that market, but the truth of the matter is that it's taking sales away from traditional UNIX vendors that Microsoft would otherwise have gotten. It has noticably slowed the impact of NT, and that definitely has their attention. Why else are they running benchmarks all over the place? They have to prove that NT is worth spending the money.
A lot of Linux people believe that Linux will beat NT by outperforming it. That might be true in a couple of years, but it's not true now except on the lowest-end hardware. Linux will beat NT in two respects:
Everything you need to run practically any kind of server is already in the box. That dramatically reduces the up-front cost of the server.
Linux is dramatically more reliable. You set it up and you forget about it. I contrast this with NT, which needs a regular reboot to keep performance up or whenever you install or upgrade an application.
Microsoft is worried about the server market and they should be. Linux is cheap and reliable and if there's any two things that IT departments like in a server those are the things.
Perhaps they should have searched for "Mandrake" rather than "MacMillan". That would have returned tens of hits on Slashdot; perhaps not as much coverage as RedHat, but certainly better than ZD has managed.
"If you want to keep NT secure, you have to apply the HOT-FIXES. There are just as many of those, if not more, than RPMs."
Actually there are only 15 hotfixes for NT4SP5 right now, although there are quite a few additional outstanding "best practices" guidelines associated with closing specific security holes present in the software using the default configuration. I'm not sure where you'd find the whole collection of those things, though. Hell, it even took me some time to track down the whole collection of hotfixes: I couldn't find a link on their support pages so I had to search the Knowledge Base for "hotfix" and find a link in one of the articles back to the ftp server.
I have to contrast this with RedHat's approach of clearly posting errata on every release: you can see every known problem and where to get the fix all in one place.
Simplistically speaking that's right, but practically speaking it's not.
Before applying SPs I wait at least a few weeks to see what people report as breaking under the new SP. There's usually something, and all too frequently (two NT4 SPs out of five!) applying an SP has a detrimental impact on system stability.
On top of that you may have to reapply SPs after installing new packages (particularly those from Microsoft) and you want to create a new emergency repair disk. These things are not necessary under Linux.
IMO, having adminstered both systems (and a bunch of others) for years, I much prefer the small patch approach where I can pick what I want to apply according to my needs: e.g. if I'm not running ftp I don't really need to apply an ftp patch.
But as it turns out there is a way to get all-inclusive patches for Linux. Install a new release. They come out every few months, much more frequently than Microsoft service packs, and generally include all previous patches. The upgrade process is fairly similar in difficulty to applying an NT service pack. Interestingly this isn't mentioned.
Interestingly, ZD says "Imagine the work involved in integrating 21 separate fixes into a change process to be deployed across an enterprise." Actually that doesn't have to be a lot of work. You can set up a master system and use rdist to propagate patched software to everything all at once. This kind of environment is easy to set up (the software is stock) and allows the software to do the grunt work of upgrading systems. You need to buy extra software to do this kind of mass upgrade on NT.
Whoops, nothing like being off by a decimal point. Thanks for the correction. The point still stands though: There's plenty of local bandwidth even though it's shared.
I'm not trying to start a flame war here, but there are a few misconceptions about xDSL that the phone company is promoting for all they're worth. One of them regards the available bandwidth.
Yes, it's true that DSL has (more-or-less)guaranteed bandwidth between your computer and the telco office. But that's where the dedicated bandwidth ends. What the DSL companies don't bother to mention is that all those DSL pipes get merged into a shared pipe, and that pipe will be as small (and cheap) as they think they can get away with.
Cable is effectively the same. Typical local bandwidth is about 10Mbps, or slightly more than two T3s. That bandwidth is shared by neighborhoods, but generally by a lot fewer users than you'll find on a corporate LAN -- and you have six times the bandwidth of a typical corporate LAN in each of those areas (and room for at least a factor of 3 bandwidth growth). There's plenty of bandwidth go go around -- particularly since those big neighborhood pipes get merged into (you guessed it) some smaller pipe at the cable office and everyone uses that to talk to the Internet. They're not going to buy any more bandwidth than they think they can get away with either.
You're in about the same bandwidth boat regardless of which service you use -- lots of local bandwidth, but a pinch point one hop upstream. Side-by-side bandwidth tests can go either way (just ask Nick Petrelas).
The only significant differentiator in my opinion is price. Cable tends to be significantly (20+%) less expensive than even low-end DSL. By the time you're talking about megabit-range DSL that price difference is usually in the hundreds of percent.
Unfortunately most users can't pick and choose between the services yet. Most can't get either. So really you should be happy to get either one.
About ISP service: A lot is made of the fact that you "must" use the cable modem's pre-selected ISP. It's true that you'll get their ISP service like-it-or-not, but it's not true that you're forced to use them. You have the whole wide Internet full of ISPs to choose from. Want AOL? Not only is that OK with them, they'll even give you a 60% discount not to call their analog modems. Local ISP? My local ISP is also tickled pink not to have me typing up their modem. Or you can just run your own server and to hell with the lot of them.
High-bandwidth connections are great no matter what flavor. Get one if you can, and if you have the choice then so much the better.
As for AOL and @Home, well, if the quality of service of @Home is anything like MediaOne RoadRunner then there would be a whole lot of people happier with AOL.
Ignoring for a minute that Solaris is already free for a lot of people, it still wouldn't make it.
Why not? It's not a matter of software at all. It's a matter of hardware.
First, my basic assumption: People don't really care which computer system they run so long as it gets the job done. If it works well enough they will decide by price. Sun beat Apollo and DEC not by having better software (even today's Solaris is a pale shadow of Domain in all sorts of ways) but by selling fast machines cheap. SunOS was bare-bones to say the least. But Sun sold their Sun3 boxes at prices the other guys couldn't match, and then they slam-dunked them with the SPARCstation, providing three times the performance at the same price.
The "fast machines cheap" theme runs decades long in the computer world. Remember: DEC beat IBM not by having better stuff but by selling it for less.
We're seeing it happen again. The Intel and the PC manufacturers are building boxes using mainstream hardware that are the equal of some pretty expensive hardware just a couple of years back. Sure, Sun's stuff got faster too but that's not the point: the point is that people who couldn't have dealt with a PC before can now use them to get their job done. So why pay more?
PCs are cheap and up to all but the toughest tasks -- and they're getting faster and more reliable and even cheaper all the time. Lots of people, like myself, who were exclusive SPARC users for years now use PCs because they provide much better bang for the buck. So long as the PC is fast enough to get the job done it doesn't much matter that the SPARC has the performance edge.
In the not-so-distant future the SPARC will become untenable for Sun; the R&D costs will continue to climb while the number of chips they can amortize the cost over will drop due to increasing penetration by Intel-based systems. Sun will get pushed ever upward towards the tip of the pyramid of users, those few who buy the fastest machines, and will have to charge more and more for those machines just to break even.
This theme is familiar too: Sun did it to the supercomputer companies. Thinking Machines? Cray? Gone. All of the specialized supercomputer companies are gone, their markets dwindled to the point where the business was untenable.
I think we can take as a given that whatever wins in the workstation/server space will win on Intel-based hardware. It has such a huge user base that they have the most R&D resources and can spread it amongst the most user. Now the question is: is Solaris good enough on Intel hardware to beat Linux or other contenders?
Today that answer is emphatically "no". Solaris scales better and performs pretty well but it just doesn't support that much hardware. Linux runs on everything while Solaris is, well, picky. You really want top-grade stuff to run Solaris. Linux runs on that piece of junk clone 486 box with the weird CD-ROM. Or any clone box you happen to find in the pages of Computer Shopper.
What that means is that the market for Linux is way, way larger than that for Solaris. Sun could fix that, but it'll take years and cost a fortune. Linux is getting that support for free. And even if Sun makes the investment they can't beat Linux on price.
I've long wondered why Sun keeps bashing on Microsoft. It must just be for the PR. You see, Microsoft is not Sun's problem. Intel is Sun's problem. Linux just makes that problem worse.
So: All else being equal, Linux would probably win because it's faster on common hardware and supports more (read: cheaper) hardware and Sun can't really afford to make Solaris competitive.
But all else is not equal. You see, Sun doesn't make their money on software. Never did; if you needed any proof of that the Solar System fiasco really removed all doubt. Sun makes their money on hardware and they're in trouble because PC hardware is decimating their sales at the low-end and rapidly encroaching on midrange. Over the next few years Intel-based hardware will scale well up into Sun's performance spectrum and will do it at a price Sun cannot afford to match.
Sun can't afford to give away a version of Solaris that might accelerate that. Hell, they can't even afford to SELL a version of Solaris that might accelerate that. Solaris/x86 was supposed to be a hole card if they had to jump off of SPARC. But Linux blindsided them. Now all their competitors have a high-quality UNIX too, and they don't even have to pay anyone else for the right to sell it.
So: Sun can't win this. They can't compete in the PC space against companies that are accustomed to razor-thin margins. They can't give away any kind of seriously supported Solaris on the PC because it'll just chop up their market even faster. And the PC is encroaching, fast. At some point they simply won't be able to afford to do Solaris development anymore, free or otherwise.
Lots of people are claiming that open source will rule the world because it's open. No, that's not it at all. Open source will rule because it's CHEAP. As it turns out "open" and "cheap" are interrelated in this case, but the important point is that the hardware companies don't have to pay a software tax.
For all practical purposes Microsoft is successful because they allow vendors to outsource the OS development, spreading R&D costs out over many vendors. That was true of UNIX, too, once upon a time. Well, Linux is the mother of all outsourcing operations and that will make it a smash hit. It is not only low-cost, it's zero-cost.
Linux gives the PC manufacturers -- particularly Compaq and Dell -- the possibility of competing head-to-head with Sun on functionality but without any of the software R&D costs that Sun has to bear with Solaris. With super-low software costs and the ability to undercut Sun seriously on hardware these PC vendors are going to beat Sun silly.
Some may argue that scalability problems will keep Linux out of the game indefinitely. I don't think so; it'll slow it down in the near term, but not much. Linux' scalability is improving at an unbelievable pace. It did in the last two years what it took Sun five to do with Solaris, and that was without significant vendor support. Linux will likely scale as well as Solaris inside of two years.
And that's why Solaris can't possibly beat Linux. Not now, anyway, and it probably never could have.
But all of this is just a specific case of a more important point that you should keep in mind whenever you're thinking of Linux-versus-whatever. Linux has effectively devalued server OS software to the point where it's not worth spending a lot of money on anymore. That is great for the consumer but it's bad for proprietary server OS vendors. It's certainly going to dent Microsoft's plans, but it's really going to hurt Novell. And you'd have to be blind to miss the fact that it's going to devalue it elsewhere too; the desktop PC is probably immune, but only because it's so late in its life and there's a lot of momentum.
Starting with the Win9x CD you'll have a real hard time beating 20 minutes; that's about the amount of time you spend waiting for it to just copy the software. Typically with probing and whatnot I find that it takes closer to an hour.
Is a disk image cheating? No, but to play fair you should realize that installing Linux from images is a snap too. Moreover all the tools necessary to create and install Linux disk partitions are in-the-box; you don't need 3rd-party software like Ghost. That's pretty unusual, though, perhaps because you don't have to reinstall Linux anywhere near as often as Win9x. The only time I've ever had to do a Linux reinstall was when my RH5.0 gateway got cracked and I needed to do a clean install to fix all the backdoors they installed. Not a problem with Windows, but then again Windows couldn't even do the job.
But of course disk image installs are not really an option for most, they start with the CD. I've found that I can do an installation of Red Hat in about 20 minutes if I know all the answers in advance.
But if you're working with new hardware you can easily spend ten times that long hunting down details -- even when you know what you're doing. Dot clocks and RAMDACS? Please. I've been working with X11 since R1 (that's circa 1988 kiddies) and PCs for almost two decades -- and I had no freakin' clue what a dot clock or RAMDAC was before I started working with Linux. I don't even have the paperwork around anymore for most of my monitors to look up things like the vertical sync rate. If there's one place where Linux is too hard to install it's X11. It's worse than "too hard" it's "make-a-smart-guy-feel-stupid hard."
If we expect this stuff to be installed by average users, it *must* improve.
But, honestly, it's not that good even in the Windows Oasis. Ever see a newbie try to install Win98? It's not a lot prettier, other than that it's pretty good at picking out video hardware. And NT is nowhere near as easy as that. Most people get by because they DON'T install Windows -- the manufacturer does, or the IT department does, or the local geek does. And when it fails they go running for help.
Linux could work just fine for those people (modulus available software anyway) if it were pre-installed. The nice thing about Linux is that it's install-and-forget; unless the hardware fails you'll never have to do a thing to it. Things don't just become randomly corrupted and you can't accidentally delete all your shared libraries. (To be fair, NT can be set up like this too, though I'm almost the only person I've ever seen do it, and of course NT is rarer than hen's teeth in the home user market.)
But one place where Linux, or at least Red Hat Linux, beats all variants of Windows hands-down is in upgrades. I stick in the new Red Hat CD and say "upgrade" and 15 minutes later I have an upgraded system. It's so good it's mindless. I wish I wish Windows were that good.
Terabyte databases are eaten up fast whenever high-volume imaging is involved. Somebody mentioned film, but more commonly it's scanned document archives (like checks or credit card receipts).
But interestingly as I read the Oracle challenge it's not for MSQL to beat 1% of the performance of Oracle on the test, it's to beat 51% (or so). Oracle claims to be 100% faster, and the challenge is to beat that "100% faster" mark by 1%.
That was a pretty safe challenge given the limits on hardware that Microsoft has to work with.
My itty bitty 486DX3/100 can (and does) fully load a T-1. CPU should not be an issue -- but RAM is. You want to cache as much of your content as you can.
Note that there's a big difference between a display-only system and an interactive system. The latter needs more firepower, particularly if it's a Perl/CGI-based system (as is Slashdot). I couldn't run much of an interactive site on my 486.
Microsoft's response indicating that the backup is there in case of disaster is simply nonsense.
The first thing that occurred to me (and others in this thread) was that you need only make copies of the key to safeguard against its loss.
Does it really seem likely that Microsoft has only one copy of a key on which their software depends? Not bloody likely. There must be redundant backups. Furthermore the key is probably not stored exclusively in some super-secret place; they need it to generate new builds, a process done on a daily basis. That means that the release engineering team has access to it and you can bet that they're not walking over to some ultra-secret building with the build bits every day.
It makes sense to have a developer key (though it should really only be used in internal builds), but the only way it makes sense to have a second production key is if it belongs to a second party. There is no additional security provided by having a second key that wouldn't be provided by having backup copies of the first key. In fact, it's more secure since two keys gives you twice the targets in a brute-force search for the private key.
So: I think we can take Microsoft's response as being pure bullshit. So why is the key really there?
Consider this new evidence in light of the recent request by the DOJ for the rights to surreptitiously monitor your computer system given a sealed warrant. Well, that key would make it a hell of a lot easier to insert evesdropping hooks, wouldn't it?
Now, aside from not being all that keen on companies selling my personal information all the time I'm not much in the way of a privacy nut. If they want to monitor my system, hey, it's their time and energy to waste. But don't ask me to believe bullshit "backup key" arguments. It ain't so, and you're insulting me by suggesting it is.
That key is there at the request of the US government, you can bet your last dollar on that. It gives them the ability to drop in a bug that can monitor any data manipulated via the crypto API. This is a better technological solution than key-escrow.
Now here's the way you can use this in your favor: build a software package that checks the signature of the crypto API against the different keys. If you have one that verifies against the so-named NSA key then you're not using the stock Microsoft package anymore. And wouldn't that be interesting?
Meanwhile, NT runs on TWO platforms. and 64 bit NT will be a complete rewrite! Dear God man, if you can't even port to a 64 bit CPU, how transportable can your OS be?
By the time NT 3.5 was released it ran or had run on i860, MIPS, x86, Alpha, and POWER. SparcV9 and HPPA ports were in-progress. NT also serves as the basis for WinCE (though this is not widely known) which runs on several more.
He's right, portability is enhanced by targetting more than one architecture from the outset. NT has done quite well in that respect. Where NT has not done so well is in maintaining that support -- perhaps because most of those other architectures are not commercially viable markets for Microsoft (for various reasons).
Now, the fact that Linux did pretty well at that despite its x86-originated design is largely a matter of emulating the design of UNIX -- which had been designed for portability back in about 1970.
Another interesting fact is that Cutler's NT design was strongly (*very* strongly) related to the design of Prism, a VMS follow-on. DEC cancelled Prism so Cutler walked out the door with his whole team and they formed the original NT team.
Now, if we've cut out the manufacuring costs and the middlemen and only have the retailer, why is the price the same?
I asked them about that. They say that the publisher charges them the same for the etext as they would a printed volume.
In effect the publisher is making extra profit on these. Which, in the short term, is probably reasonable since these are low-volume and require unusual distribution mechanisms.
I would expect that things will change as competition increases.
I have a couple of problems with the RocketBook (and all of the other dedicated eBooks).
1) It's way way too expensive for what little it does. For me to consider buying a dedicated digital reader it needs to be under $100 -- and preferably WAY under. That pretty much means that they'll have to be subsidized by the publishers.
2) I can't build up a personal library very easily. With the Pilot I can save 'em on my PC and pick and choose what I want on demand.
3) What happens when I lose the #$@^ thing? Are all my books lost too? This was a primary concern with the Pilot texts too, but they're not locked to the hardware in any way as are some of the other readers (not sure about the RocketBook).
4) Not much text available. It's getting better, I know, but damn -- $500 is a lot to spend for a pretty limited book selection.
What I like is the big screen. It's a much better form-factor than the Pilot.
eBooks will be a big thing, but the economics and text availability has to be worked on. If they're going to succeed with dedicated readers the publishers ought to sell them way below cost and make it up in the document sales (ie the cellphone model). That's the only way to ramp up the volume of the devices fast.
I don't expect this to happen any time soon; I think it much more likely that we'll see palmtop computers develop the market to the point where publishers are willing to subsidize dedicated readers.
Unfortunately most or all of the current eBook companies will be dead by then.
Keep in mind: I have tried eBooks on my Palm and am a huge fan of the idea. Even so I'm not willing to spend $500 on a single-function device when I can get the feature on a $200 device (albeit with a smaller screen) that does a lot of other things too.
I had trouble reading long texts on my old Palm 5000, but after I got a Palm V -- well, the display is hell-and-gone better. It's really not much different than reading a small-type paperback.
I find it near perfect for the train -- its small size works well in crowds and easy to find my place again when moving between trains and buses and whatnot. In addition to carrying a book around I suck down the NY Times front page with AvantGo (*way* more convenient than a full-size newspaper on the train!).
I would be happier if the display were larger but that's mostly a matter of how often I have to page it.
This was quite a surprise, I thought it would give me headaches. Nope, works great.
This is what convinced me that digital books will succeed. They just have to get the readers out there for cheap.
So lots of people noticed the strong Delphi bias. I was more than surprised by that; it has to have been a seriously skewed demographic.
I'm actually a pretty big fan of Delphi (nee Object Pascal, nee Turbo Pascal, all of which I've used) for many of the same reasons I'm a fan of Java; it's a lot easier to write correct code in it than it is in C or C++, and that's the name of the game to me. I wrote a hell of a lot of code in Object Pascal back in the 80s and much of it is still in use.
But c'mon, you can't ask me to believe that half of the Linux crowd is interested in it. That defies belief: if so many people wanted it GNU Pascal would be out there and in frequent use. It ain't so: Pascal is basically dead outside of Delphi, and Delphi isn't particularly popular.
Anybody basing the decision to build a product based on this survey is getting the wool pulled over their eyes.
Yea, I could go for some modern development tools for Linux -- I think most of us could. But Delphi is pretty far down on the scale of languages I'd be interested in.
Now, give me a native code compiler for Java on Linux (or NT, or Solaris) and I could sink my teeth into that.
There are more than a few things wrong with the lightning story.
1) If lightning took out the installation, why did the DNS entry disappear during the downtime? That's on a disparate system and it had to have been removed manually. The only reason for doing that is that you don't want people hitting the machine (which would have been impossible anyway if the router were the problem). So, ask yourself, why not?
2) If lightning took out the router, why was the router responding but not the machine (according to reports during the downtime). This is in direct conflict with Microsoft's explanation.
3) Isn't it unusual that after an external network failure they found it necessary to reconfigure the machine? (http://www.news.com/News/Item/0,4,40185,00.html, last paragraph)
4) PC Week reports (http://www.zdnet.com/pcweek/stories/news/0,4153,1 015849,00.html) that the machine was up, but operating strangely and that it had to be rebooted a couple of times. This was before the "lightning" episode, and the long-duration outage had two different explanations out of Microsoft (lightning was one, a "known bug" was the other).
5) News stories (see previous CNet link) claim that an application (guest book) on the server had been changed. Microsoft brushes that off as "that's an application, not Windows 2000." Maybe so, but it sounds like a security compromise to me.
I think the lightning story is bullshit. I think the server went down almost immediately after people started pushing it, *and* that people got in and screwed with it. The story is there to make it look like an act of God, not an embarrassing failure.
Seriously, doesn't it seem awfully coincidental that lightning took out a critical system (but NOT their actual server -- according to Microsoft) within hours of the challenge? I mean, what are the odds of that? Keep in mind that this router had to have been inside a datacenter, and typically those datacenters are usually well protected against that kind of thing.
The story seems fishy to me, like they're lying to us because their bluff got called.
Ok, so eBay goes down and everyone gets all up in arms about it because this stuff is not 100% reliable.
Hey, we knew that. Even the best systems out there are expected to be down a few minutes a year, and most of 'em (including those "super-reliable" Suns) are on the order of a couple of *days* a year. Throw a relational database into the equation and, well, reliability ain't so hot.
There are ways to deal with that, and eBay didn't do ANY of them.
At a minimum they should have had a hot backup available, PARTICULARLY for the single point of failure -- the database. With a hot backup they could have been back online in a matter of a couple of minutes. It was insane to bet their business on a single Sun/Oracle box! Whoever made that decision should be out on the street.
But they can do a lot better than that with a little middleware infrastructure. There's no reason they can't replicate transactions to multiple databases -- or even split their databases up so they have lots of little ones handling part of the load rather than One Big Server.
Of course that will take some technology that is a bit beyond the duct-tape-and-bailing-wire stuff they're using. It's not rocket science but it's gonna be a bitch to do with CGI.
What it all comes down to is that they bet on an infastructure design that had a single point of failure and were screwed when it failed. That could have been -- SHOULD have been -- foreseen and protected against.
I could maybe see that being OK in a startup that didn't have the cash for duplicate hardware outlay, but eBay has the cash in spades and they STILL didn't do it. There's a certain level of stupidity at work here.
Regarding installation difficulty, at best they are both roughly the same -- but NT does a much better job of working with unusual or new hardware. Linux, particularly RH Linux, beats the tar out of NT when it comes to disk management though. Unfortunately both suck rocks at PnP support, can't give either one the advantage here.
Regarding "user-friendliness", I find the opinion that X is more friendly to be hard to support (and I've been using X since before Windows 3.x even hit the streets). Let me explain.
The variety of window managers and desktops makes X a lot harder to deal with for the beginner; most of the time you don't just get a working desktop when you log in and start it up, you have to tweak the hell out of it. Red Hat and Caldera have made great strides here, but the fact that there's no standard sure makes things tough.
That assumes that the server even starts up. One huge, huge problem with the design of today's X servers (at least the ones I've used) is that they tend to be monolithic (server tuned for a particular video card or class of cards). They should be redesigned with a loadable driver system instead, and a hell of a lot more work needs to go into getting them to properly identify display systems.
Don't get me wrong, I'm not particularly fond of NT, but in terms of end-user usability it creams any UNIX variant I've ever used (even NeXT, although that had a few high points).
Don't even get me going on support for peripherals such as printers. They scale well, but they can be a freakin' nightmare to configure even for an experienced UNIX hand.
If you're not focused on UI issues, though, Linux generally whips NT's ass. Gateway services, email, web services, file sharing... not much of a contest, particularly given the price differential.
Slashdot Mirror
But Microsoft is wrong that Linux has not affected sales of NT. A lot of historical UNIX workstation users such as myself switched to NT in order to take advantage of the cheap PC hardware and easy software availability. I jumped onto NT almost immediately because it was good enough and a lot cheaper than your typical RISC workstation. It took a lot of effort to make a usable development environment, but you could do well enough.
Over the last few months, however, Linux has started to make significant inroads into our engineering department. Today roughly a quarter of all engineers run Linux rather than NT -- up from just two in the spring (that's about 700% growth). Microsoft can't see that in their sales yet because IT is still buying NT licenses whether or not we use them, but the transition is happening remarkably fast -- far faster than the RISC->PC movement that brought in NT. It will not be long before new NT purchases are curtailed to some degree.
There have certainly been some NT sales lost to Linux, however. I recently purchased a new laptop and I chose to go with Linux rather than NT despite running NT exclusively for the last four years. That's a sale NT would have gotten only a few months ago.
Generally speaking Gartner is correct that the people jumping onto Linux are traditional UNIX users. That means that on the desktop Microsoft is pretty safe -- there never were that many UNIX users.
But on the server, well, that's another thing entirely. Microsoft is not particularly well-entrenched; they're competing well against the commercial UNIXen at the low-end but that's largely because the hardware is cheaper. Linux uses the same hardware, costs less, and doesn't have client license issues. That 3.5% market that is dismissed in the article was statistically insignificant only a year ago!
Gartner is right that the traditional UNIX vendors are in the most trouble, but they were already in trouble. Their problem isn't and never was Microsoft, it was the PC. Linux gives those UNIX people a lower-cost solution without giving up UNIX. For small servers it's a very interesting solution to an IT department that is UNIX-savvy.
Microsoft makes great noise about how Linux isn't hurting them in that market, but the truth of the matter is that it's taking sales away from traditional UNIX vendors that Microsoft would otherwise have gotten. It has noticably slowed the impact of NT, and that definitely has their attention. Why else are they running benchmarks all over the place? They have to prove that NT is worth spending the money.
A lot of Linux people believe that Linux will beat NT by outperforming it. That might be true in a couple of years, but it's not true now except on the lowest-end hardware. Linux will beat NT in two respects:
Microsoft is worried about the server market and they should be. Linux is cheap and reliable and if there's any two things that IT departments like in a server those are the things.
jim frost
jimf@frostbytes.com
Perhaps they should have searched for "Mandrake" rather than "MacMillan". That would have returned tens of hits on Slashdot; perhaps not as much coverage as RedHat, but certainly better than ZD has managed.
Actually there are only 15 hotfixes for NT4SP5 right now, although there are quite a few additional outstanding "best practices" guidelines associated with closing specific security holes present in the software using the default configuration. I'm not sure where you'd find the whole collection of those things, though. Hell, it even took me some time to track down the whole collection of hotfixes: I couldn't find a link on their support pages so I had to search the Knowledge Base for "hotfix" and find a link in one of the articles back to the ftp server.
I have to contrast this with RedHat's approach of clearly posting errata on every release: you can see every known problem and where to get the fix all in one place.
Before applying SPs I wait at least a few weeks to see what people report as breaking under the new SP. There's usually something, and all too frequently (two NT4 SPs out of five!) applying an SP has a detrimental impact on system stability.
On top of that you may have to reapply SPs after installing new packages (particularly those from Microsoft) and you want to create a new emergency repair disk. These things are not necessary under Linux.
IMO, having adminstered both systems (and a bunch of others) for years, I much prefer the small patch approach where I can pick what I want to apply according to my needs: e.g. if I'm not running ftp I don't really need to apply an ftp patch.
But as it turns out there is a way to get all-inclusive patches for Linux. Install a new release. They come out every few months, much more frequently than Microsoft service packs, and generally include all previous patches. The upgrade process is fairly similar in difficulty to applying an NT service pack. Interestingly this isn't mentioned.
Interestingly, ZD says "Imagine the work involved in integrating 21 separate fixes into a change process to be deployed across an enterprise." Actually that doesn't have to be a lot of work. You can set up a master system and use rdist to propagate patched software to everything all at once. This kind of environment is easy to set up (the software is stock) and allows the software to do the grunt work of upgrading systems. You need to buy extra software to do this kind of mass upgrade on NT.
Whoops, nothing like being off by a decimal point. Thanks for the correction. The point still stands though: There's plenty of local bandwidth even though it's shared.
jim frost
jimf@frostbytes.com
I'm not trying to start a flame war here, but there are a few misconceptions about xDSL that the phone company is promoting for all they're worth. One of them regards the available bandwidth.
Yes, it's true that DSL has (more-or-less)guaranteed bandwidth between your computer and the telco office. But that's where the dedicated bandwidth ends. What the DSL companies don't bother to mention is that all those DSL pipes get merged into a shared pipe, and that pipe will be as small (and cheap) as they think they can get away with.
Cable is effectively the same. Typical local bandwidth is about 10Mbps, or slightly more than two T3s. That bandwidth is shared by neighborhoods, but generally by a lot fewer users than you'll find on a corporate LAN -- and you have six times the bandwidth of a typical corporate LAN in each of those areas (and room for at least a factor of 3 bandwidth growth). There's plenty of bandwidth go go around -- particularly since those big neighborhood pipes get merged into (you guessed it) some smaller pipe at the cable office and everyone uses that to talk to the Internet. They're not going to buy any more bandwidth than they think they can get away with either.
You're in about the same bandwidth boat regardless of which service you use -- lots of local bandwidth, but a pinch point one hop upstream. Side-by-side bandwidth tests can go either way (just ask Nick Petrelas).
The only significant differentiator in my opinion is price. Cable tends to be significantly (20+%) less expensive than even low-end DSL. By the time you're talking about megabit-range DSL that price difference is usually in the hundreds of percent.
Unfortunately most users can't pick and choose between the services yet. Most can't get either. So really you should be happy to get either one.
About ISP service: A lot is made of the fact that you "must" use the cable modem's pre-selected ISP. It's true that you'll get their ISP service like-it-or-not, but it's not true that you're forced to use them. You have the whole wide Internet full of ISPs to choose from. Want AOL? Not only is that OK with them, they'll even give you a 60% discount not to call their analog modems. Local ISP? My local ISP is also tickled pink not to have me typing up their modem. Or you can just run your own server and to hell with the lot of them.
High-bandwidth connections are great no matter what flavor. Get one if you can, and if you have the choice then so much the better.
As for AOL and @Home, well, if the quality of service of @Home is anything like MediaOne RoadRunner then there would be a whole lot of people happier with AOL.
jim frost
jimf@frostbytes.com
Why not? It's not a matter of software at all. It's a matter of hardware.
First, my basic assumption: People don't really care which computer system they run so long as it gets the job done. If it works well enough they will decide by price. Sun beat Apollo and DEC not by having better software (even today's Solaris is a pale shadow of Domain in all sorts of ways) but by selling fast machines cheap. SunOS was bare-bones to say the least. But Sun sold their Sun3 boxes at prices the other guys couldn't match, and then they slam-dunked them with the SPARCstation, providing three times the performance at the same price.
The "fast machines cheap" theme runs decades long in the computer world. Remember: DEC beat IBM not by having better stuff but by selling it for less.
We're seeing it happen again. The Intel and the PC manufacturers are building boxes using mainstream hardware that are the equal of some pretty expensive hardware just a couple of years back. Sure, Sun's stuff got faster too but that's not the point: the point is that people who couldn't have dealt with a PC before can now use them to get their job done. So why pay more?
PCs are cheap and up to all but the toughest tasks -- and they're getting faster and more reliable and even cheaper all the time. Lots of people, like myself, who were exclusive SPARC users for years now use PCs because they provide much better bang for the buck. So long as the PC is fast enough to get the job done it doesn't much matter that the SPARC has the performance edge.
In the not-so-distant future the SPARC will become untenable for Sun; the R&D costs will continue to climb while the number of chips they can amortize the cost over will drop due to increasing penetration by Intel-based systems. Sun will get pushed ever upward towards the tip of the pyramid of users, those few who buy the fastest machines, and will have to charge more and more for those machines just to break even.
This theme is familiar too: Sun did it to the supercomputer companies. Thinking Machines? Cray? Gone. All of the specialized supercomputer companies are gone, their markets dwindled to the point where the business was untenable.
I think we can take as a given that whatever wins in the workstation/server space will win on Intel-based hardware. It has such a huge user base that they have the most R&D resources and can spread it amongst the most user. Now the question is: is Solaris good enough on Intel hardware to beat Linux or other contenders?
Today that answer is emphatically "no". Solaris scales better and performs pretty well but it just doesn't support that much hardware. Linux runs on everything while Solaris is, well, picky. You really want top-grade stuff to run Solaris. Linux runs on that piece of junk clone 486 box with the weird CD-ROM. Or any clone box you happen to find in the pages of Computer Shopper.
What that means is that the market for Linux is way, way larger than that for Solaris. Sun could fix that, but it'll take years and cost a fortune. Linux is getting that support for free. And even if Sun makes the investment they can't beat Linux on price.
I've long wondered why Sun keeps bashing on Microsoft. It must just be for the PR. You see, Microsoft is not Sun's problem. Intel is Sun's problem. Linux just makes that problem worse.
So: All else being equal, Linux would probably win because it's faster on common hardware and supports more (read: cheaper) hardware and Sun can't really afford to make Solaris competitive.
But all else is not equal. You see, Sun doesn't make their money on software. Never did; if you needed any proof of that the Solar System fiasco really removed all doubt. Sun makes their money on hardware and they're in trouble because PC hardware is decimating their sales at the low-end and rapidly encroaching on midrange. Over the next few years Intel-based hardware will scale well up into Sun's performance spectrum and will do it at a price Sun cannot afford to match.
Sun can't afford to give away a version of Solaris that might accelerate that. Hell, they can't even afford to SELL a version of Solaris that might accelerate that. Solaris/x86 was supposed to be a hole card if they had to jump off of SPARC. But Linux blindsided them. Now all their competitors have a high-quality UNIX too, and they don't even have to pay anyone else for the right to sell it.
So: Sun can't win this. They can't compete in the PC space against companies that are accustomed to razor-thin margins. They can't give away any kind of seriously supported Solaris on the PC because it'll just chop up their market even faster. And the PC is encroaching, fast. At some point they simply won't be able to afford to do Solaris development anymore, free or otherwise.
Lots of people are claiming that open source will rule the world because it's open. No, that's not it at all. Open source will rule because it's CHEAP. As it turns out "open" and "cheap" are interrelated in this case, but the important point is that the hardware companies don't have to pay a software tax.
For all practical purposes Microsoft is successful because they allow vendors to outsource the OS development, spreading R&D costs out over many vendors. That was true of UNIX, too, once upon a time. Well, Linux is the mother of all outsourcing operations and that will make it a smash hit. It is not only low-cost, it's zero-cost.
Linux gives the PC manufacturers -- particularly Compaq and Dell -- the possibility of competing head-to-head with Sun on functionality but without any of the software R&D costs that Sun has to bear with Solaris. With super-low software costs and the ability to undercut Sun seriously on hardware these PC vendors are going to beat Sun silly.
Some may argue that scalability problems will keep Linux out of the game indefinitely. I don't think so; it'll slow it down in the near term, but not much. Linux' scalability is improving at an unbelievable pace. It did in the last two years what it took Sun five to do with Solaris, and that was without significant vendor support. Linux will likely scale as well as Solaris inside of two years.
And that's why Solaris can't possibly beat Linux. Not now, anyway, and it probably never could have.
But all of this is just a specific case of a more important point that you should keep in mind whenever you're thinking of Linux-versus-whatever. Linux has effectively devalued server OS software to the point where it's not worth spending a lot of money on anymore. That is great for the consumer but it's bad for proprietary server OS vendors. It's certainly going to dent Microsoft's plans, but it's really going to hurt Novell. And you'd have to be blind to miss the fact that it's going to devalue it elsewhere too; the desktop PC is probably immune, but only because it's so late in its life and there's a lot of momentum.
Food for thought.
jim frost
jimf@frostbytes.com
Starting with the Win9x CD you'll have a real hard time beating 20 minutes; that's about the amount of time you spend waiting for it to just copy the software. Typically with probing and whatnot I find that it takes closer to an hour.
Is a disk image cheating? No, but to play fair you should realize that installing Linux from images is a snap too. Moreover all the tools necessary to create and install Linux disk partitions are in-the-box; you don't need 3rd-party software like Ghost. That's pretty unusual, though, perhaps because you don't have to reinstall Linux anywhere near as often as Win9x. The only time I've ever had to do a Linux reinstall was when my RH5.0 gateway got cracked and I needed to do a clean install to fix all the backdoors they installed. Not a problem with Windows, but then again Windows couldn't even do the job.
But of course disk image installs are not really an option for most, they start with the CD. I've found that I can do an installation of Red Hat in about 20 minutes if I know all the answers in advance.
But if you're working with new hardware you can easily spend ten times that long hunting down details -- even when you know what you're doing. Dot clocks and RAMDACS? Please. I've been working with X11 since R1 (that's circa 1988 kiddies) and PCs for almost two decades -- and I had no freakin' clue what a dot clock or RAMDAC was before I started working with Linux. I don't even have the paperwork around anymore for most of my monitors to look up things like the vertical sync rate. If there's one place where Linux is too hard to install it's X11. It's worse than "too hard" it's "make-a-smart-guy-feel-stupid hard."
If we expect this stuff to be installed by average users, it *must* improve.
But, honestly, it's not that good even in the Windows Oasis. Ever see a newbie try to install Win98? It's not a lot prettier, other than that it's pretty good at picking out video hardware. And NT is nowhere near as easy as that. Most people get by because they DON'T install Windows -- the manufacturer does, or the IT department does, or the local geek does. And when it fails they go running for help.
Linux could work just fine for those people (modulus available software anyway) if it were pre-installed. The nice thing about Linux is that it's install-and-forget; unless the hardware fails you'll never have to do a thing to it. Things don't just become randomly corrupted and you can't accidentally delete all your shared libraries. (To be fair, NT can be set up like this too, though I'm almost the only person I've ever seen do it, and of course NT is rarer than hen's teeth in the home user market.)
But one place where Linux, or at least Red Hat Linux, beats all variants of Windows hands-down is in upgrades. I stick in the new Red Hat CD and say "upgrade" and 15 minutes later I have an upgraded system. It's so good it's mindless. I wish I wish Windows were that good.
Terabyte databases are eaten up fast whenever high-volume imaging is involved. Somebody mentioned film, but more commonly it's scanned document archives (like checks or credit card receipts).
But interestingly as I read the Oracle challenge it's not for MSQL to beat 1% of the performance of Oracle on the test, it's to beat 51% (or so). Oracle claims to be 100% faster, and the challenge is to beat that "100% faster" mark by 1%.
That was a pretty safe challenge given the limits on hardware that Microsoft has to work with.
My itty bitty 486DX3/100 can (and does) fully load a T-1. CPU should not be an issue -- but RAM is. You want to cache as much of your content as you can.
Note that there's a big difference between a display-only system and an interactive system. The latter needs more firepower, particularly if it's a Perl/CGI-based system (as is Slashdot). I couldn't run much of an interactive site on my 486.
The first thing that occurred to me (and others in this thread) was that you need only make copies of the key to safeguard against its loss.
Does it really seem likely that Microsoft has only one copy of a key on which their software depends? Not bloody likely. There must be redundant backups. Furthermore the key is probably not stored exclusively in some super-secret place; they need it to generate new builds, a process done on a daily basis. That means that the release engineering team has access to it and you can bet that they're not walking over to some ultra-secret building with the build bits every day.
It makes sense to have a developer key (though it should really only be used in internal builds), but the only way it makes sense to have a second production key is if it belongs to a second party. There is no additional security provided by having a second key that wouldn't be provided by having backup copies of the first key. In fact, it's more secure since two keys gives you twice the targets in a brute-force search for the private key.
So: I think we can take Microsoft's response as being pure bullshit. So why is the key really there?
Consider this new evidence in light of the recent request by the DOJ for the rights to surreptitiously monitor your computer system given a sealed warrant. Well, that key would make it a hell of a lot easier to insert evesdropping hooks, wouldn't it?
Now, aside from not being all that keen on companies selling my personal information all the time I'm not much in the way of a privacy nut. If they want to monitor my system, hey, it's their time and energy to waste. But don't ask me to believe bullshit "backup key" arguments. It ain't so, and you're insulting me by suggesting it is.
That key is there at the request of the US government, you can bet your last dollar on that. It gives them the ability to drop in a bug that can monitor any data manipulated via the crypto API. This is a better technological solution than key-escrow.
Now here's the way you can use this in your favor: build a software package that checks the signature of the crypto API against the different keys. If you have one that verifies against the so-named NSA key then you're not using the stock Microsoft package anymore. And wouldn't that be interesting?
He's right, portability is enhanced by targetting more than one architecture from the outset. NT has done quite well in that respect. Where NT has not done so well is in maintaining that support -- perhaps because most of those other architectures are not commercially viable markets for Microsoft (for various reasons).
Now, the fact that Linux did pretty well at that despite its x86-originated design is largely a matter of emulating the design of UNIX -- which had been designed for portability back in about 1970.
Another interesting fact is that Cutler's NT design was strongly (*very* strongly) related to the design of Prism, a VMS follow-on. DEC cancelled Prism so Cutler walked out the door with his whole team and they formed the original NT team.
In effect the publisher is making extra profit on these. Which, in the short term, is probably reasonable since these are low-volume and require unusual distribution mechanisms.
I would expect that things will change as competition increases.
I have a couple of problems with the RocketBook (and all of the other dedicated eBooks).
1) It's way way too expensive for what little it does. For me to consider buying a dedicated digital reader it needs to be under $100 -- and preferably WAY under. That pretty much means that they'll have to be subsidized by the publishers.
2) I can't build up a personal library very easily. With the Pilot I can save 'em on my PC and pick and choose what I want on demand.
3) What happens when I lose the #$@^ thing? Are all my books lost too? This was a primary concern with the Pilot texts too, but they're not locked to the hardware in any way as are some of the other readers (not sure about the RocketBook).
4) Not much text available. It's getting better, I know, but damn -- $500 is a lot to spend for a pretty limited book selection.
What I like is the big screen. It's a much better form-factor than the Pilot.
eBooks will be a big thing, but the economics and text availability has to be worked on. If they're going to succeed with dedicated readers the publishers ought to sell them way below cost and make it up in the document sales (ie the cellphone model). That's the only way to ramp up the volume of the devices fast.
I don't expect this to happen any time soon; I think it much more likely that we'll see palmtop computers develop the market to the point where publishers are willing to subsidize dedicated readers.
Unfortunately most or all of the current eBook companies will be dead by then.
Keep in mind: I have tried eBooks on my Palm and am a huge fan of the idea. Even so I'm not willing to spend $500 on a single-function device when I can get the feature on a $200 device (albeit with a smaller screen) that does a lot of other things too.
I had trouble reading long texts on my old Palm 5000, but after I got a Palm V -- well, the display is hell-and-gone better. It's really not much different than reading a small-type paperback.
I find it near perfect for the train -- its small size works well in crowds and easy to find my place again when moving between trains and buses and whatnot. In addition to carrying a book around I suck down the NY Times front page with AvantGo (*way* more convenient than a full-size newspaper on the train!).
I would be happier if the display were larger but that's mostly a matter of how often I have to page it.
This was quite a surprise, I thought it would give me headaches. Nope, works great.
This is what convinced me that digital books will succeed. They just have to get the readers out there for cheap.
So lots of people noticed the strong Delphi bias. I was more than surprised by that; it has to have been a seriously skewed demographic.
I'm actually a pretty big fan of Delphi (nee Object Pascal, nee Turbo Pascal, all of which I've used) for many of the same reasons I'm a fan of Java; it's a lot easier to write correct code in it than it is in C or C++, and that's the name of the game to me. I wrote a hell of a lot of code in Object Pascal back in the 80s and much of it is still in use.
But c'mon, you can't ask me to believe that half of the Linux crowd is interested in it. That defies belief: if so many people wanted it GNU Pascal would be out there and in frequent use. It ain't so: Pascal is basically dead outside of Delphi, and Delphi isn't particularly popular.
Anybody basing the decision to build a product based on this survey is getting the wool pulled over their eyes.
Yea, I could go for some modern development tools for Linux -- I think most of us could. But Delphi is pretty far down on the scale of languages I'd be interested in.
Now, give me a native code compiler for Java on Linux (or NT, or Solaris) and I could sink my teeth into that.
...but eOne doesn't look much like an iMac at all. It's way more boxy, different lines. I can't see myself confusing them.
... innovate don't litigate."
I guess Apple feels that if you design an all-in-one computer with a translucent case you're infringing on their rights.
I can't say it ever really appealed to me, but my pager was done in that style years ago so it's not like Apple invented it.
Maybe it's time to do another march like we did years back: "Lotus, Apple, Ashton-Tate
There are more than a few things wrong with the lightning story.
1 015849,00.html) that the machine was up, but operating strangely and that it had to be rebooted a couple of times. This was before the "lightning" episode, and the long-duration outage had two different explanations out of Microsoft (lightning was one, a "known bug" was the other).
1) If lightning took out the installation, why did the DNS entry disappear during the downtime? That's on a disparate system and it had to have been removed manually. The only reason for doing that is that you don't want people hitting the machine (which would have been impossible anyway if the router were the problem). So, ask yourself, why not?
2) If lightning took out the router, why was the router responding but not the machine (according to reports during the downtime). This is in direct conflict with Microsoft's explanation.
3) Isn't it unusual that after an external network failure they found it necessary to reconfigure the machine? (http://www.news.com/News/Item/0,4,40185,00.html, last paragraph)
4) PC Week reports (http://www.zdnet.com/pcweek/stories/news/0,4153,
5) News stories (see previous CNet link) claim that an application (guest book) on the server had been changed. Microsoft brushes that off as "that's an application, not Windows 2000." Maybe so, but it sounds like a security compromise to me.
I think the lightning story is bullshit. I think the server went down almost immediately after people started pushing it, *and* that people got in and screwed with it. The story is there to make it look like an act of God, not an embarrassing failure.
Seriously, doesn't it seem awfully coincidental that lightning took out a critical system (but NOT their actual server -- according to Microsoft) within hours of the challenge? I mean, what are the odds of that? Keep in mind that this router had to have been inside a datacenter, and typically those datacenters are usually well protected against that kind of thing.
The story seems fishy to me, like they're lying to us because their bluff got called.
Ok, so eBay goes down and everyone gets all up in arms about it because this stuff is not 100% reliable.
Hey, we knew that. Even the best systems out there are expected to be down a few minutes a year, and most of 'em (including those "super-reliable" Suns) are on the order of a couple of *days* a year. Throw a relational database into the equation and, well, reliability ain't so hot.
There are ways to deal with that, and eBay didn't do ANY of them.
At a minimum they should have had a hot backup available, PARTICULARLY for the single point of failure -- the database. With a hot backup they could have been back online in a matter of a couple of minutes. It was insane to bet their business on a single Sun/Oracle box! Whoever made that decision should be out on the street.
But they can do a lot better than that with a little middleware infrastructure. There's no reason they can't replicate transactions to multiple databases -- or even split their databases up so they have lots of little ones handling part of the load rather than One Big Server.
Of course that will take some technology that is a bit beyond the duct-tape-and-bailing-wire stuff they're using. It's not rocket science but it's gonna be a bitch to do with CGI.
What it all comes down to is that they bet on an infastructure design that had a single point of failure and were screwed when it failed. That could have been -- SHOULD have been -- foreseen and protected against.
I could maybe see that being OK in a startup that didn't have the cash for duplicate hardware outlay, but eBay has the cash in spades and they STILL didn't do it. There's a certain level of stupidity at work here.
Regarding installation difficulty, at best they are both roughly the same -- but NT does a much better job of working with unusual or new hardware. Linux, particularly RH Linux, beats the tar out of NT when it comes to disk management though. Unfortunately both suck rocks at PnP support, can't give either one the advantage here.
... not much of a contest, particularly given the price differential.
Regarding "user-friendliness", I find the opinion that X is more friendly to be hard to support (and I've been using X since before Windows 3.x even hit the streets). Let me explain.
The variety of window managers and desktops makes X a lot harder to deal with for the beginner; most of the time you don't just get a working desktop when you log in and start it up, you have to tweak the hell out of it. Red Hat and Caldera have made great strides here, but the fact that there's no standard sure makes things tough.
That assumes that the server even starts up. One huge, huge problem with the design of today's X servers (at least the ones I've used) is that they tend to be monolithic (server tuned for a particular video card or class of cards). They should be redesigned with a loadable driver system instead, and a hell of a lot more work needs to go into getting them to properly identify display systems.
Don't get me wrong, I'm not particularly fond of NT, but in terms of end-user usability it creams any UNIX variant I've ever used (even NeXT, although that had a few high points).
Don't even get me going on support for peripherals such as printers. They scale well, but they can be a freakin' nightmare to configure even for an experienced UNIX hand.
If you're not focused on UI issues, though, Linux generally whips NT's ass. Gateway services, email, web services, file sharing