Actually, there are tons of us out there that have run decent-size windows networks with no unexpected problems. If you do it right, ie RTFM (so to speak) and manage your systems correctly, things tend to 'just work'.
There are lots of us out there that have been there, done that, and bought the t-shirt.
So to get to your situation specifically, what caused the crashes? Were the machines getting a blue-screen, or were the apps crashing? If the former, what did the dump logs say was the cause of the crash? If it was a third-party software or driver, did you contact that vendor with the situation and what was their response?
The bottom line is, if your systems are just spontaneously rebooting, then its:
1. Bad Hardware. 2. Bad Power. 3. Bad Drivers.
Dont get me wrong, Windows is far from perfect, but I get so sick of people blaming their organizational incompetence on the vendor. The information is widely and readily available on how to run windows securely, reliably, and consistently. And what you'll find is that its damn near identical to how you you run Unix systems securely, reliably, and consistently. But so often, people cant be bothered to institute some operational excellence, and 'do things right'. So they do a crappy job on managing the systems, and then yell and scream when things dont work well.
It's the equivalent of never changing the oil on your car (or doing any sort of maintenance) and then crying to Honda when the engine fails at 30,000 miles.
I've seen half a dozen buy into Exchange as their primary mail server, then have to buy front-end Sendmail based servers to handle all their incoming email from outside their company because Exchange just couldn't deal with it. Other than the MTAs using Sendmail, thats a best-practices environment for very high volume exchange organizations. Why didnt they just put some extra Exchange servers running as MTAs (ie, no store) in that same role? Thats an infinite scale-out situation with exchange, same as with sendmail.
At the end of the earlier one, he clearly states that when his windows box was unstable, it was largely due to running as a local admin all the time, and installing things willy-nilly.
When he ran his windows box properly (dont work daily with a local admin account, dont install random shareware crap, etc), the box was fine, very reliable, and largely 'just worked'.
Having used XP since 2002, he's noticed that the more applications he installs, the slower and more unstable the operating system becomes. So to keep it in tip-top shape, he's keeping his software stack simple. He vowed to install as few additional applications as possible and to install only Microsoft manufactured and branded software at that (except for Firefox).
The other action he took was to create two separate log-ins: one with administrator privileges, which he would use on the rare occasions when he wants to install new software, and one with no administrator privileges, which he uses on a daily basis. The latter prevents websites he visits from downloading Active X controls. Halamka says these Active X controls, in addition to creating security holes, can introduce the software conflicts and hardware incompatibilities that cause crashes and slowdowns. The user-only log-in also prevents his computer from automatically downloading software updates from Microsoft at inopportune moments, like during presentations.
By taking those steps, Halamka says he's achieved "a version of XP that actually hasn't crashed in 30 days. "As long as I keep [the OS] in that totally static state, it'll be OK." Now granted, there are some factual errors in there (MS automatic updates requiring a login to download software updates, etc), but I'll chock those up to the shallow tech knowledge of a CIO, combined with the low-tech-knowledge of CIO reporters.
It also amuses me that he cites 30-days with a crash as something amazing. I think what he'll find if he continues to run his shop this way is that the OS installs will last years with little problems.
Halamka says it's possible to run a secure, stable and reliable version of Windows provided you configure XP properly and don't make any changes to it. "If you give yourself system administrator privileges and you install software and serve a lot of websites, the likelihood that the OS will be corrupted is high. You can prevent yourself from getting hurt, but you have to have a really locked down environment," he says. Note that he's saying this like its a big deal, like running Windows properly is some hugely unusual situation. If you want it to work you treat it like any other system, you make it a controlled environment that is managed by professionals, who know what they are doing.
Windows current security model only came into being with Windows NT, which came out just after Linux. The NT security model is basically a variation of the VMS security model. It's not new.
NT's security model is probably better than Vista's. The hooks between IE and the kernel were not there, and that is what makes Win XP and it's offspring less secure. I'm not even sure what this means. Vista IS NT.
And what hooks are you talking about? IE has no magical hook into the kernel. IE simply leverages system DLLs for various utility functions. But since the DLL is called from iexplore.exe (which is running in userspace), the DLL runs under that process' privileges.
There's no magical back-channel between IE and the NT Kernel. IE is just another userspace application. The only real difference is that it has library dependencies on system DLLs, same as many, many other MS and non-MS applications on the market.
In fact, by default, IE on Vista runs in protected mode, which means its running with significantly LESS privileges than a non-privileged user. It cant write to anywhere in the OS except its own cache, etc.
Strange as it may seem, apps written in the past have no foreknowledge of that which one day Vista deems illegal. Actually, they did. It's called (nowadays) the Logo Program, and has been around for 10 years, and describes EXACTLY what you should and shouldnt do like writing to Program Files, or Windows folders.
This is common, industry wide knowledge, that any competent ISV has known about for a decade.
If your app writers had followed these guidelines for the past 10 years, they wouldnt have failing apps now.
Doing apps on windows right with respect to these sorts of things is so easy to do its pathetic, and MS has been giving very explicit guidance for a decade on how to do it right. But so many ISVs are just crap, and never bother to learn anything about the platform they're running on.
Remote Desktop. You haven't looked at NX (or FreeNX). Faster than anything on the windows side, supports sound, network shares, and printing. Vastly superior to the Windows one, and extremely slick. I havent used NX much, but its based on X.
RDP is snappy, fast and responsive over a 56k dial-up modem.
RDP supports:
- remote or local sounds - shared printers - shared clipboard - shared smart-cards - shared serial ports - shared local drives to remote machine - shared locally-mounted network drives to remote machine - shared plug'n'play devices (some)
Not that I'm trying to bash NX/FreeNX, but its hard to get much simpler, slicker, or easier than RDP on windows. It's possible that its slightly faster in some scenarios, but given that its based on X and how fast RDP is in even low-bandwidth, high-latency situations, I'd be very surprised if its more than marginally faster.
I needed to rebuild a T43. I tried to use the rebuild partition included on the HD but it was corrupt. So I tried to make factory-install disks but the corrupt partition prevented it. Next option? Call Lenovo and get disks sent for $51. That process took five days and eight CD-ROMs from start to finish. With Ubuntu, this process takes three hours max, not four days and there's no software keys or other things to track down. The labor involved is less than a fifth with Ubuntu and the delivered product is a lot more productive - for my use models anyway.
Why did he throw away the original discs that came with the laptop? His choice to toss them and leave himself in that position has nothing to do with windows. Even if for some reason Lenovo didnt send original discs with the box (which means he bought a consumer-level model of the Lenovo, which no one in their right mind does), that is not standard practice in the industry, and Lenovo's silly business practices have nothing to do with Windows. This was a complaint about Lenovo, not Windows.
I needed to resubscribe to Symantec on a Windows machine. Again this is a 30-60 minute timeout from production AND a $49 charge AND a hassle with product keys and sending data about my machine and purchases around to companies that I'd choose not have it if I had a choice. But I didn't since Windows XP needs Symantec's products badly even though these scanning and cleaning products degrade machine performance badly - even with a gig of RAM.
Where to start with this one.
Why is he repurchasing? If its the same machine, just a fresh install, then just use your existing subscription. Buying a new one is a contrived counter example.
There are free anti-virus products if you arent mis-managing your equipment so badly that you need the full blown, as-bad-as-a-rootkit-invasive Symantec.
Machine degredation due to A/V has nothing to do with memory. It has to do with the A/V software hooking every file & memory kernel call in the system and getting in the way on each one. This is primarily a processing power problem, not a memory one.
And I now hear that Windows Vista renames the partition it's installed on what used to be the C: partition. I need to check out this story but the very idea of automatic partition renaming is insane to even contemplate.
I dont even know what this means. But apparently something 'he heard' that Vista might do, that he cant even explain properly, might be bad. Run for the hills.
I've used rsync for backups for years. I back up my mail, my Thunderbird data, and "my document" directory (i.e.,/home/xxxx/). One of these backup commands looks like this and sits in a single shell script and runs from cron once a day
So... he is forced to write scripts and run backups on his machine, and thats his idea of fast & easy backups? What he should do in windows is turn on roaming profiles and redirected mydocuments, and then he gets full userdata backups, for free, with ZERO effort, and never having to write or run a script. I mean come-on, this is technology that was mature in the Windows 2000 days, and the way he was doing it was by copying files by hand every 2 weeks?
I've got my home router set up to pass the ssh port 22 through to a Linux server sitting in my attic.
This is the single biggest mistake most linux 'admins' make, and their equipment gets owned. In many organizations, one of the single most voluminous bits of malware network traffic is SSH brute force attempts. Having port 22 exposed on a unix machine is essentially the same as exposing CIFS ports on a modern windows OS (ie, one that has all the legacy CIFS & SMB stuff turned off). If you're not careful, you're going to get beat on constanly, every day, for years. You may or may not get pwned, but its not a good configuration.
So if I'm understanding correctly... you want structured and fast (ie, indexed) data storage, without having to deal with the separate domain (and language) of set theory, which is what SQL allows you to operate on.
The problem I see is that if you want structured data storage without SQL (and therefore, without set theory), then you're going to have to give up something. You either give up the incredible value of a relational data store (ie, normalized data, referential integrity, constraints, joins, indexes, etc) and basically use a big cache of dictionaries/hash-maps, or you use a DAL/ORM.
I guess what I'm saying is that the best way we (we as in humans) know how to store data in a robust fashion, that maintains data integrity, minimizes redundancy, and maximizes searchability, is the good ol' Relational Database. Done properly (ie, avoiding just storing everything in one big wide table with many sparse columns), you end up with many tables. In fact, you basically end up with something that minimally represents an ERD, with extra rounds of normalization done on it to reduce data redundancy.
But unfortunately, now you're dealing with a storage structure that is heavily optimized for data integrity (and low data redundancy), and high speed of access (through indexing and partitioning optimizations within the DB). This optimization creates a data structure that is non-trivial to retrieve information out of. In other words, the data is so highly normalized (ie, broken into many small but full tables) that it requires complex set algebra to get that data back out.
Now you could ditch relational data structures entirely, and use something simpler. That will give you simple key-based access to long rows of data from the store.
And that will seem to work great in the simple case, with a relatively small amount of data. But as your data size and complexity grows, you'll start to see problems. You'll have to put lots of data-integrity code in your database, and if you make a mistake, you'll have bad data in your system (orphaned records with no parent, null values where there shouldnt be, etc).
So the problem with that approach is that it doesnt scale. It works decently in small apps where your table structure is nearly exactly 1:1 with your object model (ie, one table per 'entity' in your object model). But you run into problems when your data structure needs grow more complex. These usually revolve around data integrity issues (ie, data that is 'bad' is allowed into the system).
I've seen this happen many times over the years where, for example, someone doesnt believe in using Referential Integrity in the database, and they think it should all be done in the applciation layer. And that works great, until people do direct maintenance loads or modifications against the DB directly, without going through the application.
There's also big performance benefits to a normalized data structure. Because the structure of the data adheres to certain theoretical concepts, the query optimizer has great power to make decisions about how to efficiently access this data. If you move to a big hash-table type structure, you lose this. Performance would be fine at small scale, but access times to data rows would see linear (or worse) scalability as data size expanded. Contrast that with a relational data store, whose access times are very insensitive to data size (ie, a nearly flat response curve compared to data-size).
So I know I'm getting a little strayed from the original concern, but there is a connection.
Reliable data storage requires some sort of specialized storage structure (like a normalized data structure).
This requires a relational database engine (or something like it) to manage the data, and keep it in line.
Accessing data in this format requires a specialized language, to deal with the highly specialized nature of the data storage domain.
So in other words, if you give up SQL, then you must necessarily g
What is it that you are proposing as an alternative? I hear alot of complaining, but you dont see to be offering anything else that might work better.
I believe the reason that RDBMS's are the way they are (at the core, not the peripheral feature sets) is that the problem they are trying to solve is a fundamentally 'hard' one (hard in the CS sense).
I believe if you tried to solve the data persistence problem yourself (assuming a reasonable sized data set, figure items in the millions to billions range), you would end up building the core engine of a relational (or maybe hierarchical) database all over again, but it would be tuned to the particular character of your first app's dataset.
Then as you re-used your code and tried to re-use the storage layer, you'd end up just generalizing your code. Do this for about 10 years of evolution, and whammo, you've just re-invented Oracle DB.
Also note that you can get by just fine for data storage using just standardized SQL (maybe a few dialectical issues like semi-colons, statement delimeters, etc). So from the app's point of view you're just talking to a generic database. Hence the upsurge in use of DAL's and ORM's in the industry the last couple years.
Then do your optimization at the DB layer, where it is invisible and abstracted away from your app layer.
But... back to the original point. Storing large amounts of heterogenous data in a form that can be quickly (ie, nearly instantly) read back and added to (in a performant way) is hard. It's easy for small data sets, but gets very hard to solve in a way that scales linearly.
Now I'll grant you that all the major DB vendors do add alot of cruft onto them, and seem to be moving towards a full blown mini-OS and platform. But thats really tangential to their primary use, which is just storing and retrieving structure data while maintaining data integrity. And you can do that with anything, including Oracle, without ever touching PL/SQL or any other oracle-specific bloat features.
I know of a number of organizations that were bit by the Dell Inspiron 1100/1150/5100/5100/5160 systems. (As someone that works day-to-day with fixing notebook computers, we run into a lot of these systems). Maybe I'm missing something, but why would an 'organization' ever even buy an Inspiron?
Inspiron's are consumer level garbage. Random parts that differs from one minor version to another, cutting edge components with cutting edge drivers, etc.
If an organization is buying dell, it should be buying Latitudes, OptiPlex's and Precisions. Lots of bennies there too, in addition to the higher quality parts.
For example, the power cord we bought for our Latitude D800 (15.4" wide-screen, 1st gen) will work just fine with a D420 (12.1", 3rd gen), same for docking station/port-replicator, hot-swap optical drives, etc etc etc.
What I dont know is how early Daimler started making light trucks (smaller than what is called a rig or tractor-trailer here now).
But at least in the US its a good 70-80 years old phenomenon. And Daimler and the other manufacturers in Europe had to be making them too, as trucks server a very significant utility purpose.
What about how they wreck my planet, cause my friends to have to go to war to keep them fueled up, and are 17 times more likely to kill me in a side impact collision than a normal car? Wreck your planet? Are you kidding me? If the SUV is wrecking your planet, then ANY OTHER PETROLEUM BURNING VEHICLE is wrecking your planet as well, the SUV just does it at a very slightly faster rate. If you really believe that, then you need to stop driving altogether. You have fun with that.
Your friends have to go to war because of:
1. Incredibly bad leadership in the US. 2. Choosing to enroll in the armed forces.
If your friends were voluntarily enrolling in armed forces with the hope they'd never have to fight, then they gambled and lost. The smart move (IMO, I'm sure others will disagree) would be not to enlist in the first place, peacetime or war.
If you're worried about a more massive vehicle causing disproportionate damage to your vehicle in a wreck, then it sounds like you need to get a more massive vehicle.
You know what the state of the streets are. If you choose to buy a lightweight (and effectively more dangerous vehicle, due to what else is out there), then you need to accept the consequences of your actions.
No one made you buy a crappy little plastic toy car. You could choose to buy a crappy beat up (but cheap) pickup truck if you wanted to.
And I already pay extra in taxes for my gas usage. Because I use more gas. And taxes are tax$ per gallon.
The vehicles don't do ANYTHING well, at all Hmmm, lets list all the things my pickup truck does well:
Goes fast for a truck (big engine)
Pulls a boat or a trailer full of quads
Comfortable, roomy, and safe daily driver
Accelerates uphill on a 7% grade in top gear
Makes me happy!
And you know what? That last one is the only one that really matters. It's my money, and as long as I decide the cost of gas is worth the pleasure I get from my truck, then I'll continue to make the same choice.
They're unsafe, slow, can't handle, can't stop, don't have much interior space... I dont know where you get the 'dont have much interior space' thing, unless you're focusing on just the mini-SUVs out there.
My pickup truck is the most comfortable, roomiest vehicle I've ever owned. It's a pleasure to drive.
With smaller cars, I'm always touching some other part of the car. The window with my left shoulder, the roof with my head, the center console with my hip, etc. In my pickup truck, I've got room to *gasp* actually put my arms at my sides if I want, without a door or center console getting in the way.
Thats actually a very fair and reasonable approach.
If I decide that the likelihood I'll be involved in an accident with my kids is higher than I like, then damn straight I'm going to select a vehicle that minimizes my chances of significant injury.
Speaking for the OP, since I'm in a similar situation...
Since its my life, my money, and my decisions, then I need whatever I decide I need. Some days I need to drive my truck real fast, some days I dont.
Thats the great thing about being free. I'm not beholden to people like yourself to explain my choices. I need what I want, and I dont have to justify it to someone with a (from my point of view) warped world view.
Lars, if thats what you want to do, then more power to you!
Go get that ambulance and 18-wheeler, and enjoy yourself!
After all, as we both know, its your money, your life, and so you can make whatever decisions about them you like.
There is no obligation or expectation that you have to explain your personal choices to anyone else, or justify them to some pathetic little busy-body who has so little going on in their own life that they have to go meddle in other people's.
I was just called an enviro-fascist for asking why people are having 4-5 children and/or driving Hummers when there are no obvious advantages to either. We live in a (more or less) free society and a (more or less) market economy.
This means that people get to decide for themselves what a good cost/benefit tradeoff is for their spending.
And we dont like it when busybodies try to muddle in people's personal affairs and question why they spend their money the way they want to. And rightly so.
I'll try to make it clear: No one needs to justify, explain, or rationalize how they spend their hard earned money to you or anyone (some.gov excepted). It's their money, their world-view, their choices.
In other words, the answer to every one of your questions as to why they 'need' something is: because they want to.
What did people do 20+ years ago before this truck and SUV thing happened? Did people just not pull trailers? They used trucks. This 'truck thing' has been going on since pretty darn close to the beginning of automobiles existing.
If the person has 4-5 kids, why do they have that many kids? We don't live in an agrarian society anymore, so there's no need for that many children. Do they have a religious reason for it, or are they just too ignorant/uninformed to know about birth control? Umm, maybe because they want to have 4-5 kids? I know its crazy, but some people actually love their kids, so much so that they want to have more.
What does 'need' have to do with anything?
And what about the person driving the Hummer? Do they have a big family? Do they need to haul cargo/tow trailers, or are they just trying to impress people with the amount of money they can afford to waste? Maybe its because they like driving a big powerful vehicle. There doesnt have to be any other reason.
I fit into this category. I drive a Dodge Ram with the 5.7L hemi engine. I rarely tow anything (though it does come up now and then), and if I was big into towing, I'd get a bigger, heavier, more powerful truck (3/4 ton with a diesel).
And you know what, the only reason I got that truck with that engine, is because it gives me pleasure. I enjoy driving it. I enjoy the feel of 350 horsepower and 400 lb-ft of torque. I just plain like it.
And for me, the extra cost of the low mileage (10 in the city, 17 on the highway), is fine, and worth it for the pleasure I get out of it.
I know its hard to understand, but not everyone in the world thinks like you, and has the same value system as you do.
The point is that unless you have a definite, consistent use for a vehicle that only gets 18 miles per gallon, you're wasting a resource that could be put to better use. Better by whose definition? Yours or mine? By my definition, the cost is a valid tradeoff. And its a market economy on oil, so as long as its sold on the open market and I can afford to buy it, I'll make my own cost/benefit trade-offs, thank you very much.
We dont live in a command economy where you are in charge. And this is a good thing, because not everybody is the same. The things you value, only a relatively small portion of the population values as well. And I'm not saying that your views are a minority, but there are just so many damn views, and opinions, that the only way we can make it work is by us each deciding how we want to spend the fruits of our labor.
Note that what you need for towing is torque, not horsepower.
And the only real way to get high torque is through high engine volume. In other words, trucks have big engines because thats the only way to get big torque. And big torque is what you need to pull the boat or the enclosed trailer full of quads to the sand dunes.
Now, that being said, shutting off some cylinders during low load _is_ a valid way to reduce fuel consumption when you dont need the torque. But it doesnt reduce the overall weight of the vehicle. So you're still going to get crappy gas mileage, you just will get a little bit better than you would without it.
Making patches available as soon as possible, the administrators can schedule testing and patching as most convenient, maybe weekends are preferred for rolling out patches. And they can decide which patches to fast track, and which to delay to the regular updates according to how their systems are affected. This doesnt really work in the real world. As soon as patches are released to the world, they get very quickly reverse engineered, and massive active exploitation tools follows by a few days to a week.
This effectively forces admins to install patches asap as soon as they come out, as there's only a short window before the active exploits flood the wild.
The net result of all this is that MS releasing patches once a month is effectively better security, even if it means they hold on to some patches for a couple weeks. At least in the general case. There are always exceptions.
Most windows admins (sadly) dont use tools like WSUS or SUS, SMS, etc. The general consensus is that this makes it easier on their lives. Mind you, I'm not making a value judgement on this observation, but it has been clearly and strongly what I've seen (and been reported in the media, and by MS).
The majority of their customers want less freedom as to when they install a patch? I call shenanigans. The majority of their customers want patching to occupty as small a piece of their time as possible. This provides that, as it concentrates the patches into one lump per month (usually).
I mean, is it really possible that the majority of paid Windows NT sysadmins with enough licenses to make Microsoft care about what they have to say are so stupid that they don't know that they can install a patch as long after the release as they want? Again, ignoring the nastiness and focusing on the content of what you said: Thats just not how they work. For the most part, most MS shops install the patches as soon as they are released. They have no desire or need to wait, they want to get their machines patched asap.
Now mind you, many bigger shops have to go through change control boards, and scheduled downtime periods and such, but the once-per-month release is even more important for environments like that, as it consolidates the administrative overhead of downtime to once per month.
Slashdot Mirror
Actually, there are tons of us out there that have run decent-size windows networks with no unexpected problems. If you do it right, ie RTFM (so to speak) and manage your systems correctly, things tend to 'just work'.
There are lots of us out there that have been there, done that, and bought the t-shirt.
So to get to your situation specifically, what caused the crashes? Were the machines getting a blue-screen, or were the apps crashing? If the former, what did the dump logs say was the cause of the crash? If it was a third-party software or driver, did you contact that vendor with the situation and what was their response?
The bottom line is, if your systems are just spontaneously rebooting, then its:
1. Bad Hardware.
2. Bad Power.
3. Bad Drivers.
Dont get me wrong, Windows is far from perfect, but I get so sick of people blaming their organizational incompetence on the vendor. The information is widely and readily available on how to run windows securely, reliably, and consistently. And what you'll find is that its damn near identical to how you you run Unix systems securely, reliably, and consistently. But so often, people cant be bothered to institute some operational excellence, and 'do things right'. So they do a crappy job on managing the systems, and then yell and scream when things dont work well.
It's the equivalent of never changing the oil on your car (or doing any sort of maintenance) and then crying to Honda when the engine fails at 30,000 miles.
At the end of the earlier one, he clearly states that when his windows box was unstable, it was largely due to running as a local admin all the time, and installing things willy-nilly.
When he ran his windows box properly (dont work daily with a local admin account, dont install random shareware crap, etc), the box was fine, very reliable, and largely 'just worked'. Having used XP since 2002, he's noticed that the more applications he installs, the slower and more unstable the operating system becomes. So to keep it in tip-top shape, he's keeping his software stack simple. He vowed to install as few additional applications as possible and to install only Microsoft manufactured and branded software at that (except for Firefox).
The other action he took was to create two separate log-ins: one with administrator privileges, which he would use on the rare occasions when he wants to install new software, and one with no administrator privileges, which he uses on a daily basis. The latter prevents websites he visits from downloading Active X controls. Halamka says these Active X controls, in addition to creating security holes, can introduce the software conflicts and hardware incompatibilities that cause crashes and slowdowns. The user-only log-in also prevents his computer from automatically downloading software updates from Microsoft at inopportune moments, like during presentations.
By taking those steps, Halamka says he's achieved "a version of XP that actually hasn't crashed in 30 days. "As long as I keep [the OS] in that totally static state, it'll be OK." Now granted, there are some factual errors in there (MS automatic updates requiring a login to download software updates, etc), but I'll chock those up to the shallow tech knowledge of a CIO, combined with the low-tech-knowledge of CIO reporters.
It also amuses me that he cites 30-days with a crash as something amazing. I think what he'll find if he continues to run his shop this way is that the OS installs will last years with little problems. Halamka says it's possible to run a secure, stable
and reliable version of Windows provided you configure XP properly and don't make any changes to it. "If you give yourself system administrator privileges and you install software and serve a lot of websites, the likelihood that the OS will be corrupted is high. You can prevent yourself from getting hurt, but you have to have a really locked down environment," he says. Note that he's saying this like its a big deal, like running Windows properly is some hugely unusual situation. If you want it to work you treat it like any other system, you make it a controlled environment that is managed by professionals, who know what they are doing.
And what hooks are you talking about? IE has no magical hook into the kernel. IE simply leverages system DLLs for various utility functions. But since the DLL is called from iexplore.exe (which is running in userspace), the DLL runs under that process' privileges.
There's no magical back-channel between IE and the NT Kernel. IE is just another userspace application. The only real difference is that it has library dependencies on system DLLs, same as many, many other MS and non-MS applications on the market.
In fact, by default, IE on Vista runs in protected mode, which means its running with significantly LESS privileges than a non-privileged user. It cant write to anywhere in the OS except its own cache, etc.
This is common, industry wide knowledge, that any competent ISV has known about for a decade.
If your app writers had followed these guidelines for the past 10 years, they wouldnt have failing apps now.
Doing apps on windows right with respect to these sorts of things is so easy to do its pathetic, and MS has been giving very explicit guidance for a decade on how to do it right. But so many ISVs are just crap, and never bother to learn anything about the platform they're running on.
RDP is snappy, fast and responsive over a 56k dial-up modem.
RDP supports:
- remote or local sounds
- shared printers
- shared clipboard
- shared smart-cards
- shared serial ports
- shared local drives to remote machine
- shared locally-mounted network drives to remote machine
- shared plug'n'play devices (some)
Not that I'm trying to bash NX/FreeNX, but its hard to get much simpler, slicker, or easier than RDP on windows. It's possible that its slightly faster in some scenarios, but given that its based on X and how fast RDP is in even low-bandwidth, high-latency situations, I'd be very surprised if its more than marginally faster.
From TFA:
I needed to rebuild a T43. I tried to use the rebuild partition included on the HD but it was corrupt. So I tried to make factory-install disks but the corrupt partition prevented it. Next option? Call Lenovo and get disks sent for $51. That process took five days and eight CD-ROMs from start to finish. With Ubuntu, this process takes three hours max, not four days and there's no software keys or other things to track down. The labor involved is less than a fifth with Ubuntu and the delivered product is a lot more productive - for my use models anyway.
Why did he throw away the original discs that came with the laptop? His choice to toss them and leave himself in that position has nothing to do with windows. Even if for some reason Lenovo didnt send original discs with the box (which means he bought a consumer-level model of the Lenovo, which no one in their right mind does), that is not standard practice in the industry, and Lenovo's silly business practices have nothing to do with Windows. This was a complaint about Lenovo, not Windows.
I needed to resubscribe to Symantec on a Windows machine. Again this is a 30-60 minute timeout from production AND a $49 charge AND a hassle with product keys and sending data about my machine and purchases around to companies that I'd choose not have it if I had a choice. But I didn't since Windows XP needs Symantec's products badly even though these scanning and cleaning products degrade machine performance badly - even with a gig of RAM.
Where to start with this one.
And I now hear that Windows Vista renames the partition it's installed on what used to be the C: partition. I need to check out this story but the very idea of automatic partition renaming is insane to even contemplate.
I dont even know what this means. But apparently something 'he heard' that Vista might do, that he cant even explain properly, might be bad. Run for the hills.
I've used rsync for backups for years. I back up my mail, my Thunderbird data, and "my document" directory (i.e., /home/xxxx/). One of these backup commands looks like this and sits in a single shell script and runs from cron once a day
So ... he is forced to write scripts and run backups on his machine, and thats his idea of fast & easy backups? What he should do in windows is turn on roaming profiles and redirected mydocuments, and then he gets full userdata backups, for free, with ZERO effort, and never having to write or run a script. I mean come-on, this is technology that was mature in the Windows 2000 days, and the way he was doing it was by copying files by hand every 2 weeks?
I've got my home router set up to pass the ssh port 22 through to a Linux server sitting in my attic.
This is the single biggest mistake most linux 'admins' make, and their equipment gets owned. In many organizations, one of the single most voluminous bits of malware network traffic is SSH brute force attempts. Having port 22 exposed on a unix machine is essentially the same as exposing CIFS ports on a modern windows OS (ie, one that has all the legacy CIFS & SMB stuff turned off). If you're not careful, you're going to get beat on constanly, every day, for years. You may or may not get pwned, but its not a good configuration.
So if I'm understanding correctly ... you want structured and fast (ie, indexed) data storage, without having to deal with the separate domain (and language) of set theory, which is what SQL allows you to operate on.
The problem I see is that if you want structured data storage without SQL (and therefore, without set theory), then you're going to have to give up something. You either give up the incredible value of a relational data store (ie, normalized data, referential integrity, constraints, joins, indexes, etc) and basically use a big cache of dictionaries/hash-maps, or you use a DAL/ORM.
I guess what I'm saying is that the best way we (we as in humans) know how to store data in a robust fashion, that maintains data integrity, minimizes redundancy, and maximizes searchability, is the good ol' Relational Database. Done properly (ie, avoiding just storing everything in one big wide table with many sparse columns), you end up with many tables. In fact, you basically end up with something that minimally represents an ERD, with extra rounds of normalization done on it to reduce data redundancy.
But unfortunately, now you're dealing with a storage structure that is heavily optimized for data integrity (and low data redundancy), and high speed of access (through indexing and partitioning optimizations within the DB). This optimization creates a data structure that is non-trivial to retrieve information out of. In other words, the data is so highly normalized (ie, broken into many small but full tables) that it requires complex set algebra to get that data back out.
Now you could ditch relational data structures entirely, and use something simpler. That will give you simple key-based access to long rows of data from the store.
And that will seem to work great in the simple case, with a relatively small amount of data. But as your data size and complexity grows, you'll start to see problems. You'll have to put lots of data-integrity code in your database, and if you make a mistake, you'll have bad data in your system (orphaned records with no parent, null values where there shouldnt be, etc).
So the problem with that approach is that it doesnt scale. It works decently in small apps where your table structure is nearly exactly 1:1 with your object model (ie, one table per 'entity' in your object model). But you run into problems when your data structure needs grow more complex. These usually revolve around data integrity issues (ie, data that is 'bad' is allowed into the system).
I've seen this happen many times over the years where, for example, someone doesnt believe in using Referential Integrity in the database, and they think it should all be done in the applciation layer. And that works great, until people do direct maintenance loads or modifications against the DB directly, without going through the application.
There's also big performance benefits to a normalized data structure. Because the structure of the data adheres to certain theoretical concepts, the query optimizer has great power to make decisions about how to efficiently access this data. If you move to a big hash-table type structure, you lose this. Performance would be fine at small scale, but access times to data rows would see linear (or worse) scalability as data size expanded. Contrast that with a relational data store, whose access times are very insensitive to data size (ie, a nearly flat response curve compared to data-size).
So I know I'm getting a little strayed from the original concern, but there is a connection.
Reliable data storage requires some sort of specialized storage structure (like a normalized data structure).
This requires a relational database engine (or something like it) to manage the data, and keep it in line.
Accessing data in this format requires a specialized language, to deal with the highly specialized nature of the data storage domain.
So in other words, if you give up SQL, then you must necessarily g
What is it that you are proposing as an alternative? I hear alot of complaining, but you dont see to be offering anything else that might work better.
... back to the original point. Storing large amounts of heterogenous data in a form that can be quickly (ie, nearly instantly) read back and added to (in a performant way) is hard. It's easy for small data sets, but gets very hard to solve in a way that scales linearly.
I believe the reason that RDBMS's are the way they are (at the core, not the peripheral feature sets) is that the problem they are trying to solve is a fundamentally 'hard' one (hard in the CS sense).
I believe if you tried to solve the data persistence problem yourself (assuming a reasonable sized data set, figure items in the millions to billions range), you would end up building the core engine of a relational (or maybe hierarchical) database all over again, but it would be tuned to the particular character of your first app's dataset.
Then as you re-used your code and tried to re-use the storage layer, you'd end up just generalizing your code. Do this for about 10 years of evolution, and whammo, you've just re-invented Oracle DB.
Also note that you can get by just fine for data storage using just standardized SQL (maybe a few dialectical issues like semi-colons, statement delimeters, etc). So from the app's point of view you're just talking to a generic database. Hence the upsurge in use of DAL's and ORM's in the industry the last couple years.
Then do your optimization at the DB layer, where it is invisible and abstracted away from your app layer.
But
Now I'll grant you that all the major DB vendors do add alot of cruft onto them, and seem to be moving towards a full blown mini-OS and platform. But thats really tangential to their primary use, which is just storing and retrieving structure data while maintaining data integrity. And you can do that with anything, including Oracle, without ever touching PL/SQL or any other oracle-specific bloat features.
The current gen are LSI MegaRaid cards
What cards do you prefer? LSI makes a pretty good RAID card, in my experience, and are used by many of the major manufacturers.
Which RAID card, what generation?
The different gens (PERC3, PERC4, etc) are completely different cards with completely different BIOS.
Inspiron's are consumer level garbage. Random parts that differs from one minor version to another, cutting edge components with cutting edge drivers, etc.
If an organization is buying dell, it should be buying Latitudes, OptiPlex's and Precisions. Lots of bennies there too, in addition to the higher quality parts.
For example, the power cord we bought for our Latitude D800 (15.4" wide-screen, 1st gen) will work just fine with a D420 (12.1", 3rd gen), same for docking station/port-replicator, hot-swap optical drives, etc etc etc.
If Wikipedia is to be believed on this, it looks like American pickups started in the late-20's, early-30's.
http://en.wikipedia.org/wiki/Pickup_truck#History
What I dont know is how early Daimler started making light trucks (smaller than what is called a rig or tractor-trailer here now).
But at least in the US its a good 70-80 years old phenomenon. And Daimler and the other manufacturers in Europe had to be making them too, as trucks server a very significant utility purpose.
Your friends have to go to war because of:
1. Incredibly bad leadership in the US.
2. Choosing to enroll in the armed forces.
If your friends were voluntarily enrolling in armed forces with the hope they'd never have to fight, then they gambled and lost. The smart move (IMO, I'm sure others will disagree) would be not to enlist in the first place, peacetime or war.
If you're worried about a more massive vehicle causing disproportionate damage to your vehicle in a wreck, then it sounds like you need to get a more massive vehicle.
You know what the state of the streets are. If you choose to buy a lightweight (and effectively more dangerous vehicle, due to what else is out there), then you need to accept the consequences of your actions.
No one made you buy a crappy little plastic toy car. You could choose to buy a crappy beat up (but cheap) pickup truck if you wanted to.
And I already pay extra in taxes for my gas usage. Because I use more gas. And taxes are tax$ per gallon.
And you know what? That last one is the only one that really matters. It's my money, and as long as I decide the cost of gas is worth the pleasure I get from my truck, then I'll continue to make the same choice. They're unsafe, slow, can't handle, can't stop, don't have much interior space
My pickup truck is the most comfortable, roomiest vehicle I've ever owned. It's a pleasure to drive.
With smaller cars, I'm always touching some other part of the car. The window with my left shoulder, the roof with my head, the center console with my hip, etc. In my pickup truck, I've got room to *gasp* actually put my arms at my sides if I want, without a door or center console getting in the way.
Thats actually a very fair and reasonable approach.
If I decide that the likelihood I'll be involved in an accident with my kids is higher than I like, then damn straight I'm going to select a vehicle that minimizes my chances of significant injury.
Thats just good common sense.
Speaking for the OP, since I'm in a similar situation ...
Since its my life, my money, and my decisions, then I need whatever I decide I need. Some days I need to drive my truck real fast, some days I dont.
Thats the great thing about being free. I'm not beholden to people like yourself to explain my choices. I need what I want, and I dont have to justify it to someone with a (from my point of view) warped world view.
Lars, if thats what you want to do, then more power to you!
Go get that ambulance and 18-wheeler, and enjoy yourself!
After all, as we both know, its your money, your life, and so you can make whatever decisions about them you like.
There is no obligation or expectation that you have to explain your personal choices to anyone else, or justify them to some pathetic little busy-body who has so little going on in their own life that they have to go meddle in other people's.
This means that people get to decide for themselves what a good cost/benefit tradeoff is for their spending.
And we dont like it when busybodies try to muddle in people's personal affairs and question why they spend their money the way they want to. And rightly so.
I'll try to make it clear: No one needs to justify, explain, or rationalize how they spend their hard earned money to you or anyone (some
In other words, the answer to every one of your questions as to why they 'need' something is: because they want to.
What does 'need' have to do with anything? And what about the person driving the Hummer? Do they have a big family? Do they need to haul cargo/tow trailers, or are they just trying to impress people with the amount of money they can afford to waste? Maybe its because they like driving a big powerful vehicle. There doesnt have to be any other reason.
I fit into this category. I drive a Dodge Ram with the 5.7L hemi engine. I rarely tow anything (though it does come up now and then), and if I was big into towing, I'd get a bigger, heavier, more powerful truck (3/4 ton with a diesel).
And you know what, the only reason I got that truck with that engine, is because it gives me pleasure. I enjoy driving it. I enjoy the feel of 350 horsepower and 400 lb-ft of torque. I just plain like it.
And for me, the extra cost of the low mileage (10 in the city, 17 on the highway), is fine, and worth it for the pleasure I get out of it.
I know its hard to understand, but not everyone in the world thinks like you, and has the same value system as you do. The point is that unless you have a definite, consistent use for a vehicle that only gets 18 miles per gallon, you're wasting a resource that could be put to better use. Better by whose definition? Yours or mine? By my definition, the cost is a valid tradeoff. And its a market economy on oil, so as long as its sold on the open market and I can afford to buy it, I'll make my own cost/benefit trade-offs, thank you very much.
We dont live in a command economy where you are in charge. And this is a good thing, because not everybody is the same. The things you value, only a relatively small portion of the population values as well. And I'm not saying that your views are a minority, but there are just so many damn views, and opinions, that the only way we can make it work is by us each deciding how we want to spend the fruits of our labor.
Note that what you need for towing is torque, not horsepower.
And the only real way to get high torque is through high engine volume. In other words, trucks have big engines because thats the only way to get big torque. And big torque is what you need to pull the boat or the enclosed trailer full of quads to the sand dunes.
Now, that being said, shutting off some cylinders during low load _is_ a valid way to reduce fuel consumption when you dont need the torque. But it doesnt reduce the overall weight of the vehicle. So you're still going to get crappy gas mileage, you just will get a little bit better than you would without it.
This has apparently been dealt with, from what I'm reading.
u pdate-on.aspx
o ups/dgbrowser/
http://blogs.technet.com/wsus/archive/2007/04/28/
http://www.microsoft.com/technet/community/newsgr
This effectively forces admins to install patches asap as soon as they come out, as there's only a short window before the active exploits flood the wild.
The net result of all this is that MS releasing patches once a month is effectively better security, even if it means they hold on to some patches for a couple weeks. At least in the general case. There are always exceptions.
Most windows admins (sadly) dont use tools like WSUS or SUS, SMS, etc. The general consensus is that this makes it easier on their lives. Mind you, I'm not making a value judgement on this observation, but it has been clearly and strongly what I've seen (and been reported in the media, and by MS). The majority of their customers want less freedom as to when they install a patch? I call shenanigans. The majority of their customers want patching to occupty as small a piece of their time as possible. This provides that, as it concentrates the patches into one lump per month (usually). I mean, is it really possible that the majority of paid Windows NT sysadmins with enough licenses to make Microsoft care about what they have to say are so stupid that they don't know that they can install a patch as long after the release as they want? Again, ignoring the nastiness and focusing on the content of what you said: Thats just not how they work. For the most part, most MS shops install the patches as soon as they are released. They have no desire or need to wait, they want to get their machines patched asap.
Now mind you, many bigger shops have to go through change control boards, and scheduled downtime periods and such, but the once-per-month release is even more important for environments like that, as it consolidates the administrative overhead of downtime to once per month.