Obviously, you can buy Windows if you like, but then you've paid twice for access to the technology (you paid through your taxes, remember).
You forget, some of the taxes that Microsoft paid went into funding that research as well. Also, what happens if the software is a huge hit and makes MS another ton of cash? More taxes to fund more research. And they'll pay a heck of a lot more in taxes than you will. A possible alternative is that MS decides not to include the product and the majority of the world doesn't use it. (Why download something you don't really need? Most people are content with email and surfing now.) How effective were those tax dollars spent on research to provide a benefit to only a small number of people? The network effects on the economy and society of releasing it in a proprietary-friendly format far outweigh any GPL-only distribution. They could release it in multiple formats and please more people though.
Of course, given the amount of power (some of) you Americans waste on air conditioning and your (government's?) refusal to acknowledge global warming is real, it is good news for just about everyone - until the oil companies close it down, of course.
I don't think AC is a waste. When I lived in North Carolina those 100 degree days with high humidity were just a tad bit uncomfortable. And I personally don't put any stock in the predictions of gloom and doom over global warming - there aren't enough facts yet. So, what happens when the ice caps melt? A lot more land area and a greater range of latitude for green plants to grow which suck up the greenhouse gases. And are we in a natural cycle and if so is it stronger or weaker than normal? (And while we're at it, that computer you're sitting at was made with some pretty nasty chemicals, so it'd be best if you just stop using it - you polluter. And paper production causes dioxin (I think) to be released - don't use any paper. And damming up rivers causes high mercury concentrations - don't use hydroelectric power. Don't watch TV - that's an unnecessary drain of electricity. Cows farting releases methane (greenhouse gas) - so don't eat any beef. etc.) I sure as heck am not going to give up my comforts because a bunch of hippies (my term for environmentalists) thinks we ought to hug the planet instead of live. I hugged my last tree long ago (and I hugged lots of 'em when I was young), before I decided that some 'sacrifices' needed to be made for human advancement (preferably the hippies). BTW - there are more acres of forest in the US than there was 100 years ago. Save the forest from those soul-less loggers! Stop supporting them - don't buy lumber, or books, or newspapers, or magazines... The environmentalists have cried wolf far too many times.
Note that I'm not advocating going out and needlessly wasting resources. (Wasting hippies on the other hand...)
Now they say they can't turn over their secret APIs which they denied existed for security reasons?
I didn't see anywhere in the article where it said anything about secret APIs. For all of the conspiracy theories about these so-called secret API's I still haven't seen anybody put forth any credible evidence that they exist. Usually they'll refer to Rtl*, Ke*, or some other kernel functions that are documented in some DDK.
Three arguments against Microsoft's position:
Nimda.
Code Red.
The fact that a virus framework for.Net was released to the wild before the "official".Net specification.
If I installed a Red Hat distro from two years ago and didn't patch it how secure would I be? Code Red didn't happen to boxes that had been patched in the previous couple of months. If it happened on a Linux box the admin would be stupid for not patching and being rooted, but on a Win2k box it's MS's fault for lame users who won't patch. (BTW IIS6 is shipping with most/all unnecessary features turned off by default, so the situation should improve.) As for the.NET exploit - it was a bug that had already been fixed in the next beta by the time the theoretical exploit was announced. I'm sure no OSS code ever had a bug in a beta before.
Oh, and mod this down as flamebait for pointing out the hypocrisy of the typical slashdot crowd.
I took a compiler class from a developer at Cray and anytime I mentioned "cache" he cringed and said it was bad. He also gave us an overview of the supercomputer they were developing. There were no caches - the latency was very low and the memory had some extra bits to indicate whether it was locked (among other things) in order to allow it to be used for synchronization. Very cool stuff. And the MTA processors were pretty cool too:)
Why should the government pay for research and development of software under a license that allows Microsoft to take it, modify it (perhaps trivially, perhaps integrate it into the OS) and then sell it back to the US government and citizens for $big profits?
How about this instead - You don't like that Microsoft takes some publicly funded work and creates jobs and generates taxes off of it? Fine. You and a couple of your friends create a small company, make the equivalent tweaks to the source, then try to sell it to the government. Give MS some competition. It's a free market after all (sort of).
If the government pays for research and development of GPL'ed software, they are ensuring that the government, US citizens, and US corporations will always be free to use the fruits of that work, even after it has been extended. That's how I would prefer my tax dollars to be spent, thanks.
Ah, but you'll always be able to enjoy the fruits of that work - the original code will still be available under a less restrictive license. Why should you automatically be given the rights to the work that other folks have done based on private funding and effort? If you write some earth-shattering app in your bedroom and release it under the GPL that's fine. However if everyone collectively contributes funding to some research shouldn't it be available for the widest range of uses? That's how I would prefer my tax dollars spent.
If I'm giving money to the government, and they turn around and spend that money to help develop software, then I sure as hell deserve a piece of it! I helped to pay for it's existence, therefore I believe that I own a piece of it.
I totally agree - if some of my taxes went towards research then I'd like that research to turn around and benefit me economically by contributing a net gain to the economy. It appears that the GPL impairs a company from turning a profit. (Name one large OSS/GPL company that has turned real profit - and none of that "pro forma" crap.) Call me a capitalist if you will, but I think forcing the GPL on a publicly funded research effort is not the most economically beneficial way to go about things. You know, for a group that pays so much lip-service to freedom (as in speech)/.'ers seem pretty quick to condemn wanting to spend (mostly) other people's money on something that would ultimately benefit everyone more than if all code was released under the GPL.
If the government spends the money on a corporation to help them develop software that they are going to turn around and charge me for, then I am being charged twice for the same software!
Wait a minute! Do you mean if some software company came along and packaged up freely available source code into some binaries and then sold it (thus creating a taxable product) that somehow you've been denied access to the original source? How exactly have your rights been violated? If you feel that way, why the hell do you buy the software? Just go and download the source directly. And if said company modifies the source and sells an improved product what's wrong with charging for it? Wouldn't a free market basically decide what that's worth? If you feel so wronged about it, make your own package, distribute it, charge for it, ship the code with it. That's your option.
So, the way I see it - the GPL is more restrictive for businesses trying to make money off of IP and as an American I don't want my taxpayer dollars going towards research for software that will later dictate what I have to do with any extensions to that code if I want to sell it. Giving the source along with the software should be an option, not an obligation.
That is true for certain... I was doing some finite element analysis work several months ago on an PIII/NT4 machine simulating stress/strain effects on a certain kind of iron. And then my friend did the same analysis on a RISC machine running HP-UX. The Unix machine had perhaps 2X the CPU power and RAM, but the analysis was on an order of magnitude faster.
The moral of the story: Real work requires a real computer and a real OS.
Perhaps your moral should be "real work requires a real processor with real FP power", like Athlon and P4 (ugh) in the Intel world. You intimate that it's an MS problem when maybe it truly is a processor problem.
"MS says XP is rock stable, hardly ever crashes, bullshit. The lies in advertising piss me off more than the crashes themselves - false advertising that is something I'd like to see them punished for."
I've had it running on three home-built home machines for months, never had a blue screen, rarely had a problem (flawless installs and automatic updates). Same for work (five machines). Seems rock solid for me. And my home machines are running some cheap hw.
Assuming you've had problems with the stability yourself (otherwise you have no right to be pissed) - Have you analyzed the bluescreens? Chances are it's a driver issue. How about patches - are you up to date? If not then who is really to blame here? Can you say that it's false advertising just because you might have a crappy 3rd party driver that bugs out all the time? That's not MS code, that's someone else's fault. And after a crash do you submit the results back via the online crash analysis dialog that pops up after you reboot (i.e. "report this problem to MS")? If not then it's possible that MS hadn't had a report on that crash yet. If they don't know about it they can't fix it and you don't have any right getting pissed. Have your ruled out faulty HW? You didn't say anything to this effect in your post.
If you haven't analyzed the problem and determined where the bugcheck is coming from then you don't have a right to get pissed at MS, you should only be pissed at yourself for your ignorance. (You start off saying you're a "professional software developer", so I'm assuming you can go research some KB's to learn how to read blue screeens.)
So, can you make a solid case against MS about false advertising and show that they've acted in bad faith? You can do a google search and find all sorts of reliability studies showing XP is multiple times more reliable (mostly in MTTF) than 98, 95, and somewhat to a lot more reliable than Win2k. Seems like you're in a minority on this one - most every one else is having improved stability.
First, to nit pick, it's MSCS, not MCS (which is Microsoft Consulting Services).
"It was a total disaster...The SMTP service would hang and the Cluster Service would either ignore it, or hang in the process of failover
I spent my first 3 months on site rebuilding that puppy..(Thanks MS!)"
What could possibly take 3 months to rebuild? And did you contact MS Support? If you're doling out a quarter mil I'm sure you'd have a support contract, so that shouldn't be an issue. At the very least you could spend a couple hundred to get the machine working, right?
"but if you want it to failover when a service hangs, forget it..."
This is the responsibility of the resource dll to detect. If there's a problem with it contact MS PSS and get it fixed. Odd that I've never heard anyone else ever complain about this and that there are thousands of Exchange clusters set up and working just fine...
"We have the same problem with our Webserver on a purpose built Compaq "cluster-in-a-box""
You generally shouldn't be putting web servers on MSCS - WLBS/NLB is much more appropriate for that.
This post - by iankerickson on Thursday February 21, @02:17PM (#3046528) - is such a crock that I have to respond.
First off, it was anything but an "insightful" post.
"MCS works quite well, especially well on Fibre Channel and Brand Name Hardware such as Dells and Compaqs." That's right, anything that's been certified (on the Cluster HCL). Otherwise MS would get all sorts of calls because you're using shitty scsi cards that can't deal with multiple initiators and such. Most cards & drivers were never tested for that and they'll break. Not exactly something you want to find out when your motherboard gets fried and you're trying to failover. You can use non-HCL hw, but it's not supported.
"allows more memory for large applications (if the application is written correctly to use it) and (this is the main feature) allows services to fail-over from one host to the other."
It does nothing for memory. That's in the base OS - Advanced Server & Datacenter. You're now talking out of your ass.
MSCS is a shared-nothing high availability failover solution. Meaning, each resource can only be online on one machine at a time. This includes disks. Just because it doesn't fit you're definition of cluster doesn't mean it sucks. Basically, you put all of your dynamic data on the disks, when the app or service fails the resources move to another node and your service comes back online. This is a solution to maintain HIGH AVAILABILITY. What happens if your motherboard fries? Simple, the second node takes over.
"Visualize how long in your mind that would take, and then double it. If anything goes wrong, like a service won't stop (imagine that) or a service can't start due to a dependancy, it throws a monkey into the whole works." Yes, a manual failover is like doing net stop on one and net start on the other. But that's a lot faster than 'pull the plug and reboot' or 'take a gun to the motherboard'. And having to restart the service is a fact of life when there's a service failure. If the app designers want to do a hot standby model wherein their service is running on the passive node and waiting for the failover in order to start servicing requests, they can do that. And the results of a resource failure depends on the resource dll controlling that service and the cluster restart policy for the resource and it's group. If the service won't stop it's the job of the resource dll to terminate the process. If you're using "Generic Service" and things don't work, that's your own dumb fault. Talk to the service vendor and get them to write a decent one.
"and while it would have been trivial for Microsoft to expose each disk to both hosts always" Okay, so write a small generic app script that will initiate a "net use" on the other nodes and place it in each disk group. Of course, exposing each disk to both hosts doesn't usually make sense in a FAILOVER CLUSTER. Again, your definition of what a cluster should be doesn't match what MSCS is.
"Read the entire manual very carefully and take notes before you even purchase hardware" I.e. Do as I say, don't do as I do.
"MSCS is a different build (compile) of NT" Couldn't be further from the truth. With NT 4 EE it came on the Optional Components CD. With Windows 2000 and later it ships as part of the OS. The only 'different build' was for Terminal Server in NT4 days (that ended with Win2k).
"And then there's the blue screens. And the 7 hour installation procedure" There are two drivers that ship with MSCS - clusreg & clusdisk. There are no known bugcheck problems with either on any version. If you find one please contact Microsoft Support. Chances are you don't know how to read a blue screen and the issue was probably screwed up storage drivers (not MSCS).
"deleting or changing some MSCS settings " Just the storage devices because usually they're controlled by 3rd party sw (e.g. IPSHA from IBM). In NT4 the storage story was pretty touchy (1.0 product), but it's much better with Win2k and.NET Server is even better (the setup wizard absolutely rocks). But, you're probably talking about the actual apps that run on top of cluster. Again, how well they play on cluster is not controlled by the MSCS folks.
"However, for just plain applications, it's OK. Anything you can run from the command line proper can be put in the cluster and will fail over." Yeah, stuff like SQL Server, Exchange, Oracle, MSMQ (terrible on NT4, much better on Win2k) work just fine.
"But you kinda imply that MSCS is worth the exorbitant price tag,... DON'T BUY MSCS -- IT SUCKS. IF THEY GIVE IT TO YOU FOR FREE, SEND IT BACK" Amazing. It never cost a dime to buy. You see, it came with Option Pack (which is free) and is on the disk with Windows 2000 and later. You obviously have no clue what you're talking about.
"Microsoft Message Queue and the Microsoft Transaction Server. Those are more BackOffice-variety PHB-entitled products that really don't do much except provide an API for sending guaranteed IPC and doing transactions" Yeah, don't use them. Anyone can write their own two-phase commit and guaranteed messaging systems, so why bother using something that is proven to work and has been debugged? I mean, it's not like anyone does anything with distributed systems these days anyways. Just break out your MIDL compiler and write the RPC libraries yourself.
"Free with the option pack." Oh, so you have heard of it. Take a look under the "\MSCS". Couldn't be more obvious.
"So the sales rep probably doesn't have a clue what your cluster really does" Probably more so than you, but that's not hard.
"and you'll end up beta-testing a 0.1a version of some bizarro-beowulf for MS." There have been some oil & gas / seismic surveying companies deploying beowulf on Windows recently (can't find links - sorry). But basically Beowulf isn't about modifying the OS. If I understand it correctly the most you'll need is some drivers, the MPI, PVM, etc. libraries, some job schedulers, and maybe a couple of other tools. There are companies that specialize in this stuff right now. So I doubt it would be any more "bizarro" than a Linux beowulf cluster.
In conclusion, it's obvious you didn't know anything about MSCS, little about the Windows NT line, and practically nothing about clustering. If you had known anything about clustering you would have realized that "cluster" is very ambiguous and means many different things. Read Gregory Pfister's "In Search of Clusters" before you post your next cluster comment.
Slashdot Mirror
You forget, some of the taxes that Microsoft paid went into funding that research as well. Also, what happens if the software is a huge hit and makes MS another ton of cash? More taxes to fund more research. And they'll pay a heck of a lot more in taxes than you will. A possible alternative is that MS decides not to include the product and the majority of the world doesn't use it. (Why download something you don't really need? Most people are content with email and surfing now.) How effective were those tax dollars spent on research to provide a benefit to only a small number of people? The network effects on the economy and society of releasing it in a proprietary-friendly format far outweigh any GPL-only distribution. They could release it in multiple formats and please more people though.
I don't think AC is a waste. When I lived in North Carolina those 100 degree days with high humidity were just a tad bit uncomfortable. And I personally don't put any stock in the predictions of gloom and doom over global warming - there aren't enough facts yet. So, what happens when the ice caps melt? A lot more land area and a greater range of latitude for green plants to grow which suck up the greenhouse gases. And are we in a natural cycle and if so is it stronger or weaker than normal? (And while we're at it, that computer you're sitting at was made with some pretty nasty chemicals, so it'd be best if you just stop using it - you polluter. And paper production causes dioxin (I think) to be released - don't use any paper. And damming up rivers causes high mercury concentrations - don't use hydroelectric power. Don't watch TV - that's an unnecessary drain of electricity. Cows farting releases methane (greenhouse gas) - so don't eat any beef. etc.) I sure as heck am not going to give up my comforts because a bunch of hippies (my term for environmentalists) thinks we ought to hug the planet instead of live. I hugged my last tree long ago (and I hugged lots of 'em when I was young), before I decided that some 'sacrifices' needed to be made for human advancement (preferably the hippies). BTW - there are more acres of forest in the US than there was 100 years ago. Save the forest from those soul-less loggers! Stop supporting them - don't buy lumber, or books, or newspapers, or magazines... The environmentalists have cried wolf far too many times.
Note that I'm not advocating going out and needlessly wasting resources. (Wasting hippies on the other hand...)
http://www.snpp.com/guides/prof.frink.html
(Search for "At a Xena fan meeting". Note the Raimi connection.)
I didn't see anywhere in the article where it said anything about secret APIs. For all of the conspiracy theories about these so-called secret API's I still haven't seen anybody put forth any credible evidence that they exist. Usually they'll refer to Rtl*, Ke*, or some other kernel functions that are documented in some DDK.
If I installed a Red Hat distro from two years ago and didn't patch it how secure would I be? Code Red didn't happen to boxes that had been patched in the previous couple of months. If it happened on a Linux box the admin would be stupid for not patching and being rooted, but on a Win2k box it's MS's fault for lame users who won't patch. (BTW IIS6 is shipping with most/all unnecessary features turned off by default, so the situation should improve.) As for the .NET exploit - it was a bug that had already been fixed in the next beta by the time the theoretical exploit was announced. I'm sure no OSS code ever had a bug in a beta before.
Oh, and mod this down as flamebait for pointing out the hypocrisy of the typical slashdot crowd.
http://www.cray.com/products/systems/mta/
How about this instead - You don't like that Microsoft takes some publicly funded work and creates jobs and generates taxes off of it? Fine. You and a couple of your friends create a small company, make the equivalent tweaks to the source, then try to sell it to the government. Give MS some competition. It's a free market after all (sort of).
If the government pays for research and development of GPL'ed software, they are ensuring that the government, US citizens, and US corporations will always be free to use the fruits of that work, even after it has been extended. That's how I would prefer my tax dollars to be spent, thanks.
Ah, but you'll always be able to enjoy the fruits of that work - the original code will still be available under a less restrictive license. Why should you automatically be given the rights to the work that other folks have done based on private funding and effort? If you write some earth-shattering app in your bedroom and release it under the GPL that's fine. However if everyone collectively contributes funding to some research shouldn't it be available for the widest range of uses? That's how I would prefer my tax dollars spent.
I totally agree - if some of my taxes went towards research then I'd like that research to turn around and benefit me economically by contributing a net gain to the economy. It appears that the GPL impairs a company from turning a profit. (Name one large OSS/GPL company that has turned real profit - and none of that "pro forma" crap.) Call me a capitalist if you will, but I think forcing the GPL on a publicly funded research effort is not the most economically beneficial way to go about things. You know, for a group that pays so much lip-service to freedom (as in speech) /.'ers seem pretty quick to condemn wanting to spend (mostly) other people's money on something that would ultimately benefit everyone more than if all code was released under the GPL.
If the government spends the money on a corporation to help them develop software that they are going to turn around and charge me for, then I am being charged twice for the same software!
Wait a minute! Do you mean if some software company came along and packaged up freely available source code into some binaries and then sold it (thus creating a taxable product) that somehow you've been denied access to the original source? How exactly have your rights been violated? If you feel that way, why the hell do you buy the software? Just go and download the source directly. And if said company modifies the source and sells an improved product what's wrong with charging for it? Wouldn't a free market basically decide what that's worth? If you feel so wronged about it, make your own package, distribute it, charge for it, ship the code with it. That's your option.
So, the way I see it - the GPL is more restrictive for businesses trying to make money off of IP and as an American I don't want my taxpayer dollars going towards research for software that will later dictate what I have to do with any extensions to that code if I want to sell it. Giving the source along with the software should be an option, not an obligation.
That is true for certain ... I was doing some finite element analysis work several months ago on an PIII/NT4 machine simulating stress/strain effects on a certain kind of iron. And then my friend did the same analysis on a RISC machine running HP-UX. The Unix machine had perhaps 2X the CPU power and RAM, but the analysis was on an order of magnitude faster.
The moral of the story: Real work requires a real computer and a real OS.
Perhaps your moral should be "real work requires a real processor with real FP power", like Athlon and P4 (ugh) in the Intel world. You intimate that it's an MS problem when maybe it truly is a processor problem.
"MS says XP is rock stable, hardly ever crashes, bullshit. The lies in advertising piss me off more than the crashes themselves - false advertising that is something I'd like to see them punished for."
I've had it running on three home-built home machines for months, never had a blue screen, rarely had a problem (flawless installs and automatic updates). Same for work (five machines). Seems rock solid for me. And my home machines are running some cheap hw.
Assuming you've had problems with the stability yourself (otherwise you have no right to be pissed) - Have you analyzed the bluescreens? Chances are it's a driver issue. How about patches - are you up to date? If not then who is really to blame here? Can you say that it's false advertising just because you might have a crappy 3rd party driver that bugs out all the time? That's not MS code, that's someone else's fault. And after a crash do you submit the results back via the online crash analysis dialog that pops up after you reboot (i.e. "report this problem to MS")? If not then it's possible that MS hadn't had a report on that crash yet. If they don't know about it they can't fix it and you don't have any right getting pissed. Have your ruled out faulty HW? You didn't say anything to this effect in your post.
If you haven't analyzed the problem and determined where the bugcheck is coming from then you don't have a right to get pissed at MS, you should only be pissed at yourself for your ignorance. (You start off saying you're a "professional software developer", so I'm assuming you can go research some KB's to learn how to read blue screeens.)
So, can you make a solid case against MS about false advertising and show that they've acted in bad faith? You can do a google search and find all sorts of reliability studies showing XP is multiple times more reliable (mostly in MTTF) than 98, 95, and somewhat to a lot more reliable than Win2k. Seems like you're in a minority on this one - most every one else is having improved stability.
First, to nit pick, it's MSCS, not MCS (which is Microsoft Consulting Services). "It was a total disaster...The SMTP service would hang and the Cluster Service would either ignore it, or hang in the process of failover I spent my first 3 months on site rebuilding that puppy..(Thanks MS!)" What could possibly take 3 months to rebuild? And did you contact MS Support? If you're doling out a quarter mil I'm sure you'd have a support contract, so that shouldn't be an issue. At the very least you could spend a couple hundred to get the machine working, right? "but if you want it to failover when a service hangs, forget it..." This is the responsibility of the resource dll to detect. If there's a problem with it contact MS PSS and get it fixed. Odd that I've never heard anyone else ever complain about this and that there are thousands of Exchange clusters set up and working just fine... "We have the same problem with our Webserver on a purpose built Compaq "cluster-in-a-box"" You generally shouldn't be putting web servers on MSCS - WLBS/NLB is much more appropriate for that.
This post -
.NET Server is even better (the setup wizard absolutely rocks). But, you're probably talking about the actual apps that run on top of cluster. Again, how well they play on cluster is not controlled by the MSCS folks.
by iankerickson on Thursday February 21, @02:17PM (#3046528) - is such a crock that I have to respond.
First off, it was anything but an "insightful" post.
"MCS works quite well, especially well on Fibre Channel and Brand Name Hardware such as Dells and Compaqs."
That's right, anything that's been certified (on the Cluster HCL). Otherwise MS would get all sorts of calls because you're using shitty scsi cards that can't deal with multiple initiators and such. Most cards & drivers were never tested for that and they'll break. Not exactly something you want to find out when your motherboard gets fried and you're trying to failover. You can use non-HCL hw, but it's not supported.
"allows more memory for large applications (if the application is written correctly to use it) and (this is the main feature) allows services to fail-over from one host to the other."
It does nothing for memory. That's in the base OS - Advanced Server & Datacenter. You're now talking out of your ass.
MSCS is a shared-nothing high availability failover solution. Meaning, each resource can only be online on one machine at a time. This includes disks. Just because it doesn't fit you're definition of cluster doesn't mean it sucks. Basically, you put all of your dynamic data on the disks, when the app or service fails the resources move to another node and your service comes back online. This is a solution to maintain HIGH AVAILABILITY. What happens if your motherboard fries? Simple, the second node takes over.
"Visualize how long in your mind that would take, and then double it. If anything goes wrong, like a service won't stop (imagine that) or a service can't start due to a dependancy, it throws a monkey into the whole works."
Yes, a manual failover is like doing net stop on one and net start on the other. But that's a lot faster than 'pull the plug and reboot' or 'take a gun to the motherboard'. And having to restart the service is a fact of life when there's a service failure. If the app designers want to do a hot standby model wherein their service is running on the passive node and waiting for the failover in order to start servicing requests, they can do that. And the results of a resource failure depends on the resource dll controlling that service and the cluster restart policy for the resource and it's group. If the service won't stop it's the job of the resource dll to terminate the process. If you're using "Generic Service" and things don't work, that's your own dumb fault. Talk to the service vendor and get them to write a decent one.
"and while it would have been trivial for Microsoft to expose each disk to both hosts always"
Okay, so write a small generic app script that will initiate a "net use" on the other nodes and place it in each disk group. Of course, exposing each disk to both hosts doesn't usually make sense in a FAILOVER CLUSTER. Again, your definition of what a cluster should be doesn't match what MSCS is.
"Read the entire manual very carefully and take notes before you even purchase hardware"
I.e. Do as I say, don't do as I do.
"MSCS is a different build (compile) of NT"
Couldn't be further from the truth. With NT 4 EE it came on the Optional Components CD. With Windows 2000 and later it ships as part of the OS. The only 'different build' was for Terminal Server in NT4 days (that ended with Win2k).
"And then there's the blue screens. And the 7 hour installation procedure"
There are two drivers that ship with MSCS - clusreg & clusdisk. There are no known bugcheck problems with either on any version. If you find one please contact Microsoft Support. Chances are you don't know how to read a blue screen and the issue was probably screwed up storage drivers (not MSCS).
"deleting or changing some MSCS settings "
Just the storage devices because usually they're controlled by 3rd party sw (e.g. IPSHA from IBM). In NT4 the storage story was pretty touchy (1.0 product), but it's much better with Win2k and
"However, for just plain applications, it's OK. Anything you can run from the command line proper can be put in the cluster and will fail over."
Yeah, stuff like SQL Server, Exchange, Oracle, MSMQ (terrible on NT4, much better on Win2k) work just fine.
"But you kinda imply that MSCS is worth the exorbitant price tag,... DON'T BUY MSCS -- IT SUCKS. IF THEY GIVE IT TO YOU FOR FREE, SEND IT BACK"
Amazing. It never cost a dime to buy. You see, it came with Option Pack (which is free) and is on the disk with Windows 2000 and later. You obviously have no clue what you're talking about.
"Microsoft Message Queue and the Microsoft Transaction Server. Those are more BackOffice-variety PHB-entitled products that really don't do much except provide an API for sending guaranteed IPC and doing transactions"
Yeah, don't use them. Anyone can write their own two-phase commit and guaranteed messaging systems, so why bother using something that is proven to work and has been debugged? I mean, it's not like anyone does anything with distributed systems these days anyways. Just break out your MIDL compiler and write the RPC libraries yourself.
"Free with the option pack."
Oh, so you have heard of it. Take a look under the "\MSCS". Couldn't be more obvious.
"So the sales rep probably doesn't have a clue what your cluster really does"
Probably more so than you, but that's not hard.
"and you'll end up beta-testing a 0.1a version of some bizarro-beowulf for MS."
There have been some oil & gas / seismic surveying companies deploying beowulf on Windows recently (can't find links - sorry). But basically Beowulf isn't about modifying the OS. If I understand it correctly the most you'll need is some drivers, the MPI, PVM, etc. libraries, some job schedulers, and maybe a couple of other tools. There are companies that specialize in this stuff right now. So I doubt it would be any more "bizarro" than a Linux beowulf cluster.
In conclusion, it's obvious you didn't know anything about MSCS, little about the Windows NT line, and practically nothing about clustering. If you had known anything about clustering you would have realized that "cluster" is very ambiguous and means many different things. Read Gregory Pfister's "In Search of Clusters" before you post your next cluster comment.