AMD has nowhere near the plant capacity to whoop on Intel.
The pro-AMD lobby here on/. is really only focused on a tiny fraction of the market -- the folks who pay a premium to have the very fastest CPU available. It's a dick-sizing contest, that's all. The worst AMD can do cut into Intel's profit margins a bit, and not go out of business themselves (which 2 years ago they were close to doing). Intel's still running at 100% production capacity, and can't build fabs fast enough. A rising tide is going to lift all boats. --
, I think that "law" (and I use that term very loosely) was only created for revenue purposes
You are absolutely correct. Moore's law was not just an academic prediction -- it was a mandate to Intel's engineering department. It's also been used by Intel's marketing department to set prices. (You would think that AMD and RISC vendors wouldn't have any trouble competing with Intel, because Intel's business plan is public and widespread knowledge.)
On paper, Moore's law keeps a stream of revenue coming because it obsoletes hardware every 18 months. In reality, software technology can't keep up with the CPU for most users, so the cycle is quite a bit longer. In fact, processor speed is hardly the real reason for most system upgrades -- the cheap disk and limited memory and the labor costs of upgrading are the real reason people keep coming back for new systems every two years for new systems with fast CPUs and cheap disks and limited memory. --
I think the concept is that the World Wide Web (as an applicaiton) would cause the Apple ][ guy to crap.
Don't forget, back in the early '80s there was lots of hype about personal computers and the on-line future. It was promised back then that you would soon be able to send mail, talk, shop, have sex, do anything on computer networks. Early systems like interactive cable TV, CompuServe, TheSource, Prodigy, and even AOL never quite measured up, and it took 15 more years before the Internet started delivering on the promise. --
It helps the developers, who are trying to build a unix-like system into something which is as transparant as the MacOS.
Not new users trying to learn to edit config files by hand.
You're looking at this from the Unix user's (your) point of view. I can tell you right now that 99% of Mac users will refuse to edit configuration files by hand -- if MacOS X requires it, for any reason, the OS will be considered a failure by the user base.
Note that a structured config format is really nothing strange in the Mac world. The old OS has it's proprietary "resource forks" where code and data and pictures and config settings all live happily in a structured, user-editable format. And, guess what, 99% of the users don't have ResEdit on their computer, and don't care to either. (My understanding is that NextStep had something similar.) --
Of course, if I can get Administrator or root access to one box that is running both TCP/IP and NetBEUI, it's possible I could gateway between them. But by then it's basically too late anyway.
The number one way that Win home boxes are hacked is leaving file sharing enabled on TCP/IP. (Vendors seem to ship this way.) NetBEUI is a great, but not perfect, solution to this problem. --
Yes, but a big chunk of that $39.99 is probably an 'upgrader tax' for a nitch market product, and has nothing to do with the real cost of the chipset. --
Do a little testing with a couple isolated NT boxes. NetBEUI is perceptually quite a bit faster for small file copies and browsing. On our real LAN, we have a old Pentium 60 NT server with an ISA NIC. Large TCP/IP file copies bog down and barely complete. NetBEUI transfers zip.
(Could it be something wrong with NT's SMB/TCPIP instead of something right with SMB/NetBEUI? Got me!) --
I have a Alcatel ADSL bridge, and was concerned about this. According to someone at my ISP, non-IP stuff gets killed right at the CO. With DSL at least, you are clogging your own bandwidth, but you shouldn't be clogging anyone elses or leaking insecure NetBEUI packets.
(I'm aware Cable works differently, but I probably won't care how until they start installing it here in 2003.) --
More history -- Microsoft sold the server version of OS/2 1.3 directly as "Microsoft LAN Manager 2.x". Earlier versions were not sold directly by MS, but were known as 3Com 3+Open and AT&T Somthinerother. These products together accumulated about a 10% market share (largely due to the difficulty of NetBEUI.)
The word "LanMan" is still all over the NT documentation. The "Server" service used to be called "LanMan Server" in NT 3.x, for example. MS claimed that it was based directly on the OS/2 networking code, and therefore bug-for-bug compatible. Microsoft wanted to hang onto that 10% market share when they moved to NT, so the networking was taken wholesale from LAN Manager.(Things started to change a bit around NT4 SP4 -- it was no longer possible to have mixed OS/2 and NT domain controllers, for example.)
IBM's version was/is known as "LAN Server" -- OS/2 Warp Server 5.0 is actually based on the Warp Client 4.0 codebase, the reason for the version number difference is MS's escalation to 2.0 back in the 1980s. The more things change, the more things seem to stay the same. --
Virtually every PC shipped comes with USB -- you can't get better economies of scale. Ethernet is $20 mature tech, sure, but it's apparently not cheap enough for most vendors to give it away. --
NetBEUI was designed by IBM in around 1983 for old-style single segment LANs. A single segment LAN is what I have in my house. I run NetBEUI between the Windows boxes, and it is F-A-S-T.
Maybe MS's TCP/IP filesharing is just broken. Doesn't matter. If you can get away with using NetBEUI for Doze/Samba stuff, I recommend using it.
(Auto configuration and non-routability is also a plus -- If you have cable/DSL, you can unbind SMB/"WINS Client" from TCP/IP on your NT boxes and use NB for local file+print, and worry less about the evil haxors.) --
Check the MS-DOS Client 3.0 on your NT Server disk -- This does include a DOS TCP/IP implementation, and one that understands MS WINS, DHCP, and Domain security to boot.
It does use lots of memory. It can't be installed directly onto a floppy disk (install to hard disk, selectively copy files to floppy, it will fit.) You need to manually install the DOS NDIS NIC drivers by editing INI files. It should work from a NIC Boot ROM. With some trickery, it should run under Windows 9x.
It's a pain in the ass, but when you get it working, it works pretty good (although much slower than NetBEUI). --
Clarification -- By "college level study skills", I mean that if you have the mental tenacity to cram for the Foo 101 mulitple choice midterm at State U, you can pass the MS tests with prep material alone (and no experience).
Obviously, a good portion of the US populace can pass a college freshman multiple choice test without much effort. There's also a good portion who can't -- all those As, Bs, Cs, and None of the Aboves makes their eyes goggle, even though they might really know their admin stuff.
(This is also the common critique of the SAT and so on -- "test taking skills" often count as much as actually knowing the material.) --
NDS primarily appeals to the operations side of IT organizations. What's more important is that the product is maintained and supported (by the vendor, or a third party, or just a worldwide group of developers and users). In house development is really unlikely to happen in most cases.
Source code access is really not a plus or a minus to most of these folks. They just want something that works and makes their lives easier and budgets smaller. --
BTW, there are more MCSE's because there are more companies that have NT servers
One big reason there's lotso MS Certified goons is because Microsoft flooded the market with easy tests that could be passed by anyone with college-level study skills. I know a couple MCSEs that got their certificate with only a couple months of dinking around with NT on their home computer (= no understanding of the networking part).
Unfortunately for the goons, MS is decertifying everyone at the end of the year. I haven't seen the new NT5 material, but would think it to be quite a bit harder, just to salvage whatever little value the MS Engineer program still has.
Novell, BTW, did the same thing back in the NetWare 3 era. They flooded the market with CNE certs, corporations started requiring a CNE as a hiring checklist item, and the resulting crowd of underskilled but certified Novell engineers really gave the company and the certificate a black eye. --
Just for the record, I'm not any sort of Gnome zealot. My Linux box is a server and usually doesn't run X at all, and when it does, it's WindowMaker. --
Presumably a "Network Object Model Environment" buys you something. I know that MS DCOM buys us something at work, and I have freinds that still tell tales of the wonders of 1980's NeXT systems and their network object RPC software.
Maybe in your case it doesn't. Maybe the software hasn't been written yet to take advantage of it. In that case, you're right -- it's just overhead, don't use it.
(Note that the "useful purpose" tag was aimed at Gnome 1.0's buggy nature, and particularly at the pointlessness of Enlightenment for my purposes. RH 6 was a pretty poor introduction of Gnome for most people.) --
Look at it this way -- you can buy a new P3/K7 system for less than a $1000, a Celeron/K6 system for even less. Every computer you can think of will be 'obsolete' in the Windows world in a year or two, if it isn't already. When Gnome 2.0 ships, my PII-400 will look pretty pathetic next to the 1.2Ghz chips that will be available.
You might disagree with the philosophy, but it makes perfect sence to develop software that takes lots of CPU cycles and puts them to a useful purpose. By the time your thing ships, nobody will tell the difference. If it runs too slow for you, use something else (KDE is fine on my P133, AfterStep flies).
As for driving slow machines users and the third world into Microsofts arms -- Hah! Microsoft doesn't want them either. If anything, Linux is going to grow as fully functional Pentium computers are flushed from corporate desktops to prepare for Windows 2000.
The great thing about Linux/Unix is that there a gazillion window managers for all sorts of machines, there's lots of choice, you can even have a fully functional console only environment. Bitching about one particular system's philosophy is pretty pointless. Just run something else, or dive in and see what you can fix. --
I've heard this flame before, and I don't really understand why it's such a big deal to have some unused libs sitting around. Unless it's a diskspace issue, but then maybe RedHat isn't the best choice to begin with. --
It's when some retard builds an 'admin tool'... that Linux will become like Windows.
I think I will take your post as an object example for the point I was trying to make, although I generally agree with the sentiment.
Even so, you contradict yourself. It is not bull that Unix has "tradition of obscure administration practices" -- check the first couple chapters out of any sysadmin book. That's not to say that it's illogical or unnecessary, or that system flexibility is a bad goal (it's not), it's just not the goal of the average midsized shoestring IT operation.
To cater to the file+print+mail crowd requires a philosophical change in how the system comes out-of-box. Folks like you demonstrate that there's lots of resistance to that. --
GUI based admin tools aren't bad in and of themselves...the gui stuff gives you the illusion of being able to competently administer a system without understanding what's going on under the hood
Agreed. If you've ever had to work with the NT4 GUI admin applets, you'd discover how amazingly bad they are. WinNT really only has a couple advantages over unix systems in terms of accessibility --
(1) It's self configuring to a certain extent (at the cost of flexibility). This allows junior admins to go with the defaults and have a functional system. (2) Windows 9x 'power users' can stumble around the Start menu to find the tools they might be looking for. This lowers the entry point for 'administrators'. On the other hand, it's difficult to find what you are looking for on Unix systems, without starting with some documentation.
In reality, the ability to really figure out what's going on 'under the hood' in NT is fairly obscured. The documentation is not that good. The education program (MCSE) isn't really helping. NT's only advantage is really a low entry point and low expectations.
I don't think an all encompasing Admin GUI is really possible or required. Even MS has their command line tools. The only thing that's needed is an entry-point into the system -- things like network config, starting and stopping daemons, basic security settings, printer configuration, user MAC, etc.
The administration issue is still a gaping black hole in the Linux usability picture. Yes, you can make your desktop "beautiful" with themes and whatnot; Yes, the installer is easier than Windows 98's. But, no, nobody has really tackled the administration issue -- things like linuxconf and the other tools you find seem to be shaky or ill supported at best.
The root of the situation seems to be ultimately cultural. Most people use Linux because they want a unix system, and unix systems have a long standing tradition of obscure administration practices. Folks have made your suggestion this board that you can have your cake and eat it too, that admin functionality can be integrated into the pretty GUI without minimizing or removing the unix environment. These suggestions are usually met with disdain -- the community as a whole seems either happy with the way things are now, or afraid of 'windozification'.
For some folks like me, it's part of the fun. But it does bring up the question of whether World Domination is really the goal or not. Widespread deployment of Linux, even on the server level, is going to require the enlistment and education of the current minimally skilled small server admin crowd. It's a tough problem -- the average MS certified goon isn't even qualified to run a mid-sized NT installation (I know, I've interviewed these guys, and apparently MS will certify you even if you don't understand basics like WINS or domain controllers), and now these guys are being asked to wade into quagmire of a unix installation without any tradition or dummies books to guide them.
Small and medium sized businesses are seriously considering Linux as an alternative to Windows, but even at zero cost, it can't do them a lot of good if they can't find/afford anyone to run the system. The barbarians are the gates -- the question is are they going to find a soft landing when they get in, or is the potential configuration nightmare of a unix system going to kill the prospect of widespread Linux deployment? Is World Domination really what people want, or is Linux best suited to the traditional unix niches of academics, ISPs, and glass houses? It's a tough problem, there's no clear answer, and it's something that may well lead to a fork in the community. --
Agreed, it's a difficult scenario because companies like RedHat have no intellectual property, and therefore not much to conceed. The only thing possible is a split-the-market deal, as Microsoft illegally proposed to Netscape.
The point holds that if MS Office shipped for Linux, and Linux started to get a desktop marketshare, Microsoft would think of something to twist the Linux vendors' arms over. It's the nature of the beast.
Slashdot Mirror
AMD has nowhere near the plant capacity to whoop on Intel.
/. is really only focused on a tiny fraction of the market -- the folks who pay a premium to have the very fastest CPU available. It's a dick-sizing contest, that's all. The worst AMD can do cut into Intel's profit margins a bit, and not go out of business themselves (which 2 years ago they were close to doing). Intel's still running at 100% production capacity, and can't build fabs fast enough. A rising tide is going to lift all boats.
The pro-AMD lobby here on
--
, I think that "law" (and I use that term very loosely) was only created for revenue purposes
You are absolutely correct. Moore's law was not just an academic prediction -- it was a mandate to Intel's engineering department. It's also been used by Intel's marketing department to set prices. (You would think that AMD and RISC vendors wouldn't have any trouble competing with Intel, because Intel's business plan is public and widespread knowledge.)
On paper, Moore's law keeps a stream of revenue coming because it obsoletes hardware every 18 months. In reality, software technology can't keep up with the CPU for most users, so the cycle is quite a bit longer. In fact, processor speed is hardly the real reason for most system upgrades -- the cheap disk and limited memory and the labor costs of upgrading are the real reason people keep coming back for new systems every two years for new systems with fast CPUs and cheap disks and limited memory.
--
I think the concept is that the World Wide Web (as an applicaiton) would cause the Apple ][ guy to crap.
Don't forget, back in the early '80s there was lots of hype about personal computers and the on-line future. It was promised back then that you would soon be able to send mail, talk, shop, have sex, do anything on computer networks. Early systems like interactive cable TV, CompuServe, TheSource, Prodigy, and even AOL never quite measured up, and it took 15 more years before the Internet started delivering on the promise.
--
It helps the developers, who are trying to build a unix-like system into something which is as transparant as the MacOS.
Not new users trying to learn to edit config files by hand.
You're looking at this from the Unix user's (your) point of view. I can tell you right now that 99% of Mac users will refuse to edit configuration files by hand -- if MacOS X requires it, for any reason, the OS will be considered a failure by the user base.
Note that a structured config format is really nothing strange in the Mac world. The old OS has it's proprietary "resource forks" where code and data and pictures and config settings all live happily in a structured, user-editable format. And, guess what, 99% of the users don't have ResEdit on their computer, and don't care to either. (My understanding is that NextStep had something similar.)
--
Of course, if I can get Administrator or root access to one box that is running both TCP/IP and NetBEUI, it's possible I could gateway between them. But by then it's basically too late anyway.
The number one way that Win home boxes are hacked is leaving file sharing enabled on TCP/IP. (Vendors seem to ship this way.) NetBEUI is a great, but not perfect, solution to this problem.
--
Yes, but a big chunk of that $39.99 is probably an 'upgrader tax' for a nitch market product, and has nothing to do with the real cost of the chipset.
--
Do a little testing with a couple isolated NT boxes. NetBEUI is perceptually quite a bit faster for small file copies and browsing. On our real LAN, we have a old Pentium 60 NT server with an ISA NIC. Large TCP/IP file copies bog down and barely complete. NetBEUI transfers zip.
(Could it be something wrong with NT's SMB/TCPIP instead of something right with SMB/NetBEUI? Got me!)
--
You might have reached aesthetic perfection by squashing NetBEUI, but your network is much slower for it.
--
Can Linux print to a NetWare print queue? (Or, how about an oldskool HP DLC printer?)
AppleTalk+PostScript might be a better bet.
--
I have a Alcatel ADSL bridge, and was concerned about this. According to someone at my ISP, non-IP stuff gets killed right at the CO. With DSL at least, you are clogging your own bandwidth, but you shouldn't be clogging anyone elses or leaking insecure NetBEUI packets.
(I'm aware Cable works differently, but I probably won't care how until they start installing it here in 2003.)
--
More history -- Microsoft sold the server version of OS/2 1.3 directly as "Microsoft LAN Manager 2.x". Earlier versions were not sold directly by MS, but were known as 3Com 3+Open and AT&T Somthinerother. These products together accumulated about a 10% market share (largely due to the difficulty of NetBEUI.)
The word "LanMan" is still all over the NT documentation. The "Server" service used to be called "LanMan Server" in NT 3.x, for example. MS claimed that it was based directly on the OS/2 networking code, and therefore bug-for-bug compatible. Microsoft wanted to hang onto that 10% market share when they moved to NT, so the networking was taken wholesale from LAN Manager.(Things started to change a bit around NT4 SP4 -- it was no longer possible to have mixed OS/2 and NT domain controllers, for example.)
IBM's version was/is known as "LAN Server" -- OS/2 Warp Server 5.0 is actually based on the Warp Client 4.0 codebase, the reason for the version number difference is MS's escalation to 2.0 back in the 1980s. The more things change, the more things seem to stay the same.
--
Virtually every PC shipped comes with USB -- you can't get better economies of scale. Ethernet is $20 mature tech, sure, but it's apparently not cheap enough for most vendors to give it away.
--
NetBEUI was designed by IBM in around 1983 for old-style single segment LANs. A single segment LAN is what I have in my house. I run NetBEUI between the Windows boxes, and it is F-A-S-T.
Maybe MS's TCP/IP filesharing is just broken. Doesn't matter. If you can get away with using NetBEUI for Doze/Samba stuff, I recommend using it.
(Auto configuration and non-routability is also a plus -- If you have cable/DSL, you can unbind SMB/"WINS Client" from TCP/IP on your NT boxes and use NB for local file+print, and worry less about the evil haxors.)
--
Check the MS-DOS Client 3.0 on your NT Server disk -- This does include a DOS TCP/IP implementation, and one that understands MS WINS, DHCP, and Domain security to boot.
It does use lots of memory. It can't be installed directly onto a floppy disk (install to hard disk, selectively copy files to floppy, it will fit.) You need to manually install the DOS NDIS NIC drivers by editing INI files. It should work from a NIC Boot ROM. With some trickery, it should run under Windows 9x.
It's a pain in the ass, but when you get it working, it works pretty good (although much slower than NetBEUI).
--
Clarification -- By "college level study skills", I mean that if you have the mental tenacity to cram for the Foo 101 mulitple choice midterm at State U, you can pass the MS tests with prep material alone (and no experience).
Obviously, a good portion of the US populace can pass a college freshman multiple choice test without much effort. There's also a good portion who can't -- all those As, Bs, Cs, and None of the Aboves makes their eyes goggle, even though they might really know their admin stuff.
(This is also the common critique of the SAT and so on -- "test taking skills" often count as much as actually knowing the material.)
--
NDS primarily appeals to the operations side of IT organizations. What's more important is that the product is maintained and supported (by the vendor, or a third party, or just a worldwide group of developers and users). In house development is really unlikely to happen in most cases.
Source code access is really not a plus or a minus to most of these folks. They just want something that works and makes their lives easier and budgets smaller.
--
BTW, there are more MCSE's because there are more companies that have NT servers
One big reason there's lotso MS Certified goons is because Microsoft flooded the market with easy tests that could be passed by anyone with college-level study skills. I know a couple MCSEs that got their certificate with only a couple months of dinking around with NT on their home computer (= no understanding of the networking part).
Unfortunately for the goons, MS is decertifying everyone at the end of the year. I haven't seen the new NT5 material, but would think it to be quite a bit harder, just to salvage whatever little value the MS Engineer program still has.
Novell, BTW, did the same thing back in the NetWare 3 era. They flooded the market with CNE certs, corporations started requiring a CNE as a hiring checklist item, and the resulting crowd of underskilled but certified Novell engineers really gave the company and the certificate a black eye.
--
Just for the record, I'm not any sort of Gnome zealot. My Linux box is a server and usually doesn't run X at all, and when it does, it's WindowMaker.
--
Presumably a "Network Object Model Environment" buys you something. I know that MS DCOM buys us something at work, and I have freinds that still tell tales of the wonders of 1980's NeXT systems and their network object RPC software.
Maybe in your case it doesn't. Maybe the software hasn't been written yet to take advantage of it. In that case, you're right -- it's just overhead, don't use it.
(Note that the "useful purpose" tag was aimed at Gnome 1.0's buggy nature, and particularly at the pointlessness of Enlightenment for my purposes. RH 6 was a pretty poor introduction of Gnome for most people.)
--
Look at it this way -- you can buy a new P3/K7 system for less than a $1000, a Celeron/K6 system for even less. Every computer you can think of will be 'obsolete' in the Windows world in a year or two, if it isn't already. When Gnome 2.0 ships, my PII-400 will look pretty pathetic next to the 1.2Ghz chips that will be available.
You might disagree with the philosophy, but it makes perfect sence to develop software that takes lots of CPU cycles and puts them to a useful purpose. By the time your thing ships, nobody will tell the difference. If it runs too slow for you, use something else (KDE is fine on my P133, AfterStep flies).
As for driving slow machines users and the third world into Microsofts arms -- Hah! Microsoft doesn't want them either. If anything, Linux is going to grow as fully functional Pentium computers are flushed from corporate desktops to prepare for Windows 2000.
The great thing about Linux/Unix is that there a gazillion window managers for all sorts of machines, there's lots of choice, you can even have a fully functional console only environment. Bitching about one particular system's philosophy is pretty pointless. Just run something else, or dive in and see what you can fix.
--
I've heard this flame before, and I don't really understand why it's such a big deal to have some unused libs sitting around. Unless it's a diskspace issue, but then maybe RedHat isn't the best choice to begin with.
--
It's when some retard builds an 'admin tool' ... that Linux will become like Windows.
I think I will take your post as an object example for the point I was trying to make, although I generally agree with the sentiment.
Even so, you contradict yourself. It is not bull that Unix has "tradition of obscure administration practices" -- check the first couple chapters out of any sysadmin book. That's not to say that it's illogical or unnecessary, or that system flexibility is a bad goal (it's not), it's just not the goal of the average midsized shoestring IT operation.
To cater to the file+print+mail crowd requires a philosophical change in how the system comes out-of-box. Folks like you demonstrate that there's lots of resistance to that.
--
GUI based admin tools aren't bad in and of themselves...the gui stuff gives you the illusion of being able to competently administer a system without understanding what's going on under the hood
Agreed. If you've ever had to work with the NT4 GUI admin applets, you'd discover how amazingly bad they are. WinNT really only has a couple advantages over unix systems in terms of accessibility --
(1) It's self configuring to a certain extent (at the cost of flexibility). This allows junior admins to go with the defaults and have a functional system.
(2) Windows 9x 'power users' can stumble around the Start menu to find the tools they might be looking for. This lowers the entry point for 'administrators'. On the other hand, it's difficult to find what you are looking for on Unix systems, without starting with some documentation.
In reality, the ability to really figure out what's going on 'under the hood' in NT is fairly obscured. The documentation is not that good. The education program (MCSE) isn't really helping. NT's only advantage is really a low entry point and low expectations.
I don't think an all encompasing Admin GUI is really possible or required. Even MS has their command line tools. The only thing that's needed is an entry-point into the system -- things like network config, starting and stopping daemons, basic security settings, printer configuration, user MAC, etc.
--
The administration issue is still a gaping black hole in the Linux usability picture. Yes, you can make your desktop "beautiful" with themes and whatnot; Yes, the installer is easier than Windows 98's. But, no, nobody has really tackled the administration issue -- things like linuxconf and the other tools you find seem to be shaky or ill supported at best.
The root of the situation seems to be ultimately cultural. Most people use Linux because they want a unix system, and unix systems have a long standing tradition of obscure administration practices. Folks have made your suggestion this board that you can have your cake and eat it too, that admin functionality can be integrated into the pretty GUI without minimizing or removing the unix environment. These suggestions are usually met with disdain -- the community as a whole seems either happy with the way things are now, or afraid of 'windozification'.
For some folks like me, it's part of the fun. But it does bring up the question of whether World Domination is really the goal or not. Widespread deployment of Linux, even on the server level, is going to require the enlistment and education of the current minimally skilled small server admin crowd. It's a tough problem -- the average MS certified goon isn't even qualified to run a mid-sized NT installation (I know, I've interviewed these guys, and apparently MS will certify you even if you don't understand basics like WINS or domain controllers), and now these guys are being asked to wade into quagmire of a unix installation without any tradition or dummies books to guide them.
Small and medium sized businesses are seriously considering Linux as an alternative to Windows, but even at zero cost, it can't do them a lot of good if they can't find/afford anyone to run the system. The barbarians are the gates -- the question is are they going to find a soft landing when they get in, or is the potential configuration nightmare of a unix system going to kill the prospect of widespread Linux deployment? Is World Domination really what people want, or is Linux best suited to the traditional unix niches of academics, ISPs, and glass houses? It's a tough problem, there's no clear answer, and it's something that may well lead to a fork in the community.
--
Agreed, it's a difficult scenario because companies like RedHat have no intellectual property, and therefore not much to conceed. The only thing possible is a split-the-market deal, as Microsoft illegally proposed to Netscape.
The point holds that if MS Office shipped for Linux, and Linux started to get a desktop marketshare, Microsoft would think of something to twist the Linux vendors' arms over. It's the nature of the beast.
--