DLL Hell occurrs when a vendor decideds to overwrite one of microsoft's dlls, which breaks other software which relied on that DLL. It's not really an example of code reuse gone bad. It's more an example of vendors having too much ability to break the OS. MS is actually trying to address this problem in Win2K. After a couple of service packs, we may see if their efforts helped.
In order to get the install user friendly, Linux needs to be more drvice driver writer friendly. Until Linux gets a device driver interface, drivers will have to be recompiled for every kernel version, single and multiprocessors, and even processor optimations. With design cycles for some hardware such as Video cards being as low as 6 months, there's no way the vendors can keep their drivers current and test them on all the kernel versions.
I understand that the Linux kernel is a constantly evolving beast, and that any device driver interface would have to change occasionally, but not having one at all is going to be a tremendous barrier to Linux ever becoming a popular desktop OS.
Even if a company chooses to release Open Source drivers recompiling the kernel is a little beyond what you can expect from the average user. This forces hardware companies to produce several binary versions for each kernel, and maintain them for several kernel versions. This means they have to maintain these versions with bug fixes and test them all. They also have to write complicated install scripts to match the correct binary with the kernel on the user's machines, and deal with all the tech support questions which are greatly compounded by all this.
This pretty much makes Linux the hardest platform to support, and customer support is a large expense for most companies. Since the Linux customer base is still relatively small, this just gives hardware companies one more reason not to bother with Linux.
For some devices the manufacturer can publish a clear specification and a little sample source, and someone will write and contribute the driver. However, if you're targeting consumers, someone has to provide support. Someone is also likely going to have to polish the installation tools so that the average consumer can get the driver working.
We support multiple OSs in all of our drivers where I work. With a little work you can abstract our a lot of the OS specific parts of the driver. You can then write most of the driver in OS independent manner and port it easily. Most of the OS specific parts are also reusable in other drivers. Unfortunately for this to work well you need to have a reasonable device driver interface.
We currently support Linux in our drivers, but creating and testing all the binaries, and the customer support problems are already becomming an issue. It's a lot of effort, and the sales numbers just arent justifing the effort. I don't expect this to continue indefinately.
The reason you the drivers have to be recompiled for each kernel is because Linux doesn't provide a binary driver interface. The interface consists of of calls that do not from one kernel version to the next even though the underlying implementation does change. Every other major OS that I know of provides an interface for drivers. It should even make the kernel itself more maintainable. However, there seems to be considerable oposition to it from the "we don't want anyone writing binary drivers" crowd.
Linux doesn't have a binary driver interface. This means that the drivers need to be recompiled for each kernel. If you want to take advantage of MMX or SSE you'll probably need a version that supports it, and one that doesn't as well. Pretty soon you're testing and shipping a dozen different binaries in order to keep your diverse customer base happy.
Is it REALLY that hard to recompile drivers to make everyone happy?
Recompiling the drivers isn't that difficult. Properly testing the drivers on the different kernels and platforms is often time consuming, tedious work. The open source community may consider it acceptable to just compile the different binaries, do a quick 5 minute test and send out the binary. Our comercial customers expect a bit more. Our drivers go through weeks of testing for reliability, compatibility, and interoperability.
How different is *BSD from Linux as far as drivers go?
BSD has two desirable features in the kernel that Linux does not. Kernel threads (with priorities), and the ability to map user buffers into kernel space. Kernel threads make it easier to write drivers that share resources well. It also makes it easier for me to make a large portion of my driver OS independent so it's not only portable, but works more consistently across OSs. This keeps the different code bases managable, and reduces debugging efforts. Mapping user memory into kernel space is necessary in order to avoid copying the users data into or out of a kernel buffer. With the devices for which I'm writing drivers it's not uncommon to be sending data in blocks of 1 MB or more at throughputs arount 100 MB/s. Copying the data is relatively slow, and will quickly peg your CPU load at 100%. If you want the throughput, you need to map the memory into kernel space. There is a definate drawback to mapping user memory into kernel space. The author of the device driver MUST make sure that the user is passing in a valid buffer, and that the pages remain locked in physical memory. Failure to do this will cause the system to crash. A bad pointer in user space causes the application to crash. A bad pointer in kernel brings down the system. When MS claims the bigest reliability problem NT has is poorly written 3rd party drivers, they aren't making it up.
I'm already paying $0.50 a gallon in taxes on every gallon. Gasoline is a huge industry, and $0.50 a gallon adds up to a huge amount of money. I keep hearing the argument that it's the oil companies that are keeping new technologies from making gasoline obsolete. We live in an economy where there are ambitious individuals with billions of dollars, who would love to make billions more. I really doubt the oil industry is bribing all of them so the don't invest in alternative fuels. Most of the alternatives to gasoline have a common problem. They don't have as high of an eneryg density as you can get with gasoline. In order to give consumers the 250 to 300 mile range they want with their cars, they batteries or natural gas tanks just have to be too large. Fuel cells may be a better solution in the future, but the technology is still too expensive. The solution of making the car smaller and lighter only works so well. Families have a lot of stuff to haul around. Those little fuel efficient, aerodynamic cars also don't do so well in rural areas where the gravel road gets 6 inch ruts in it from the rain washing it out several times a year. It's also hard to drive in snow when you're pushing snow with your bumper. There are a lot of us that aren't going to buy a little electric car. If you're commuting in a city a little Honda Insight may be a great choice. However, if you're driving it in the mountains, it just doesn't have enough horsepower to make it up the hills (although you can recharge the batteries on the downslope).
Lets try an example. I'm going camping with some friends in West Virginia. There are four of us going along with all our camping gear. The Insight is a two seater, so we'd have to take 2. Actually, we'd probably have to take three to fit all the gear, but let's say we can do it with two. The next problem is that the 60 to 70 MPG is based on driving on relatively flat ground. You aren't going to get that in the mountains, and the weight of the gear will make it considerably worse. If you're lucky you'll get 50 MPG. Instead of the 2 Insights, lets take 1 full sized Chevy Silverado with the extended cab. It seats 4 adults comfortably. Plenty of room for gear in the back. A cap on the back is recomended to keep the gear dry. If it's a 2WD truck you'll likely average about 25 - 26 MPG in the mountains. Since you have enough HP the hills don't hurt your MPG as badly. You also don't have that long string of cars behind you as you go up every slope at 25 MPH. You save a time (as long as your not stuck behind the guy in the Insight), and you can actually drive back the dirt road to where you're camping without getting stuck. You even get as good, if not better gas milage on average. If you go with the 4WD you likly drop a few MPG, but you can go more places, and it's a lot better on icy roads. If you live in a city, and have a short commute, the Insight may be for you, but car pooling will likely do just as good. If you really want good gas milage, I'm sure there are motorcycles that get 70+ MPG. They're not so good in the rain, and the AC doesn't work so well in traffic jams, but if you're looking at the Insight you've already accepted that you're going to have to make some tradeoffs.
The Insight is tiny, and has a 3 cylinder engine in it. I haven't seen any crash test ratings on it yet, but I doubt it holds up too well in a accident. And please don't tell me that the solution to the problem is to get rid of all the SUVs. Larger families need larger vehicles. A minivan would suffice for most, but there are some of us that actually need the ground clearance and the 4wd. I'd love to have better gas mileage, but my next vehicle will likely be a gas guzzling, 4wd truck, because a compact car just doesn't suit my needs. A natural gas truck might be a nice alternative if natural gas were available at gas stations.
I personally don't know what CSS is, but if browser developers aren't bothering to implement it, I doubt that there's a whole lot of demand for it. Standards are created by comittee. The result is usually a comprimais that no one is really happy with, or a bloated standard that has tons of features that are of little use to the majority of developers.
Some are, some aren't. OEMs will install another OS if there's significant demand for it. However, as far as consumer OS's are concerned they haven't had many choices. OS/2 wasn't a consumer OS, too hard to configure and use for the average user. LINUX distributions have made great strides recently, but they still have a ways to go before they are a good choice for the average user.
We live in a world where one person can post copyrighted material on the web, and millions can steal it. The copyright holder can go after the person who posted the material, if they can find them, but they can't practically go after all the people who stole their copyrighted works. How can you sue a million individuals. Our legal system can't deal with this type of crime, and it costs the copyright holder more to pursue the criminals than it's worth.
The result of this probelm has often been copy protection schemes. These schemes usually end up as a deterrent to only those who don't feel it's worth the effort to get around the copy protection. So now we have a law which prohibits circumventing the copy protection. Such a law even makes some sense to me at first thought. The probelm is that the law is being abused. The recording industry is using this law to dictate how you can use the copyrighted materials. They are using "copy protection" to extend limit how you use the material beyond the rights they are given by law. If you circumvent thier "protections" they say you are breaking the law, even if you your actions would have been considered fair use.
I doubt this is what Congress intended when they passed the law, at least not all of the Congressmen that voted for it. The courts need to overturn the parts of the law that interfere with fair use, and Congress needs to create a better balanced copyright law. This time they have to remember that the goal of copyright law is not to simply protect the rights of copyright holders. The law must also limit those rights. The framers of the law must keem in mind the actions of the recording industry to subvert the law and take away the rights of individuals. It's the recording industry that's trying to circumvent the law in this case. The law provides for fair use, and the recording industry is attempting to subvert that. Hopefully the recording industry will lose the lawsuits, but be forced to pay the legal costs of those they are persecuting.
It's reasonable to say that software that has no bugs would comply to the specification. It's however unreasonable to imply the because the software complies with the specification that it has no bugs, unless the spec is intentionally vague. The only way a spec can take into account all the unexpected user inputs that the user may try is to include vague specifications such as "will not crash due to unexpected input" or "will operate in a user friendly manner", both of which I've seen in specifications.
In reality, you can't use complience with the spec to determine if something is or is not a bug. Someone needs to look at the issues, determine if they are worth fixing, and assign some sort of priority if they are worth fixing.
It sounds tike the problem in this case is that the issues were submitted as bugs, not as design issues. However, since the product is a user interface, some of them might be considered bugs.
From what I read, he does have considerable experience in writing user interfaces. It sounds like he went through the user interface in a through fasion and submitted the things he had issues with as bugs.
Things he did right: 1. Tested the portion of the software in which he has considerable experience. 2. Submitted all the issues at once so developers don't have to keep going back and revisiting the issues (hopefully). Things he did wrong: 1. Submitted things as bugs that others didn't consider to be bugs.
Unless there's some history here that I'm not aware of, or he keeps submitting things this way after he's told to submit them in another way, this doesn't seem like a very big deal.
If you just move people off the project because you don't like their opinions, you end up with poorly designed products, which also happens often enough on commercial products. There's a balancing act to be done, and everyone on the project needs to be somewhat flexible.
From the AP story "The circuit court acted shortly after Microsoft almost simultaneously filed a notice of appeal with Jackson and a motion with the circuit court to stay Jackson's penalties, even though Jackson was also considering a Microsoft stay motion."
Microsoft filed the appeal with Jackson. There seems to be a lot of non-lawyers arguing how M$ actions are unlawful and unconstitutional. I think the DOJ has enough experience with the law that they would be pointing that out if it were true.
It does seem like the appeals court acted very quickly on this, but considering the actions of Jackson and the DOJ it's not that surprising. The behaviorial modifications in Jackson's ruling go into effect in 90 days. 90 days is a pretty short time for a company Microsoft's size to significantly change how they do business. They need to renegociate/rewrite their licensing contracts with all their vendors. This will cost them a large amount of time and money. They obviously don't want to do it if they don't have to. They really don't want to do it twice. Once now and again if they get some or all of the ruling overturned. Microsoft requested a stay on the remedies until the case was appealed. Justice then filed a huge reply and seemed to be stalling. They even went as far as to request that Jackson not rule on the stay until M$ had filed the appeal. It seems like both Justice and Jackson are trying to get the case to the Supreme who won't have time to rule on staying the remedies until after they go into effect.
Why would the appeals court want to hear this case? Maybe because Jackson no longer appears to be an impartial Jurist. Maybe it's because Jackson pretty much ignored their ruling that M$ could integrate IE to Windows. They've ruled on issues related to this case before. Jackson and Justice appear to be trying to bypass them and subvert their authority.
Jackson has been anything but sympathetic to Microsoft. The trial has been rushed, the witnesses very limited. From the outset everyone was predicting that M$ would lose the case, but win on appeal. Now the DOJ and Jackson appear to be trying to limit the appeals process. Everyone has a right to a fair trial, even M$. The more of this crap that gets pulled, the less I have faith in the DOJ and the Judicial system. The worst thing that could happen is that a mistrail could be declared by the appeals court. Then everything starts over again.
IANAL. The president could appoint an Attourney General who would drop the case, but that would be a very politically dangerous thing to do. I think the states could still pursue the case, but I really don't know.
The Sherman act provides for both criminal and civil penalties, however this is a civil case. No criminal charges have been filed, and criminal charges would be handled in a seperate case from the civil case. It may be possible for the government to still file criminal charges under the Sherman act, but it seems unlikly to happen.
So when LINUX becomes a common desktop OS user are going to have to save their email attachments that are to be executed. They're going to have to figure out that the file is to be executed in some way, with no pretty icon (not to mention an extension). They're even going to have to run chmod on the file to get a script to run.
These limitations are all perfectly reasonable on a server OS. They obviously make the system considerably more secure. But if you think that Joe and Jane user who use this as a desktop OS at work or at home are going to figure all this out, I think your overly optimistic.
Companies can probably afford a few minor disasters from viruses than losing the productivity they gain from clicking on e-mail attachments and having them do what the sender intended. I know this is a huge security hole which requires the user to determine if the attachment is safe based on who they think sent them the message.
The question is how much ease of use you want to trade for security. I think the solution that many have suggested of showing a dialog box before outlook lets an application send an email is a good place for MS to start. However, it appears they have made some patches to fix some of their security problems, and that sys admin are very lax in applying them. Hopefully both MS and those sys admins have learned their lessons.
Let me preface this with saying that standard APIs are definatley the most desirable solution, and that I have very little experience in developing OpenGL software. However, there is a problem with how slowly standards tend to evolve. When you have a mature standard such as POSIX or TCP/IP it's much easier to develop to the standard and still put in all the features and performance that your market desires.
Gaming hardware and software is evolving at an amazing rate right now, and deciding standards by committee isn't usually a quick process. You also have the problems of people on the standards committee having conflicting goals. 3dfx may have some feature that no one else has, and want support for it integrated into OpenGL as quickly as possible. Other compainies are going to have other priorities, and feel that 3dfx's feature is just slowing down the implementation of the features they think are improtant.
There are also companies that are unwilling to put their technology into an open standard such as OpenGL. Those same companies may be willing to lisence their technology to Microsoft to put it in DirectX.
Standards like OpenGL are also very succeptable to feature bloat. You get the I'll vote for your feature if you vote for my feature syndrome going, and you end up with a standard which includes so many features that no one can really implement them in drivers.
Microsoft's method of unilaterally creating DirectX can allow them to more quickly add features which take advantage of new technologies. Because it's a closed standard, they can also start working with hardware vendors to add new features under NDA. This allows Microsoft to get a head start on implementing new features, and allows the hardware vendor to wait to announce new features and not tip their hands to their competitors.
If you goal is protability, then standards are the way to go. If your goal is to get the newest features out to your customers, and you're willing to accept that your customer base will be smaller, then a propriatary API may be a reasonable choice.
Say for an example you have 3 different Word Processors with the same file format. You've just trippled your training budget for Word processors. Your tech support staff now has to support 3 times as many products. You're also 3 times as likely to run into a bug that keeps you from setting up standard templates you're trying to develop, and now the people working on them need to know 3 different word processors and spend time testing on all of them to make sure it all works.
It doesn't take too long until managment and support figure out that this is just too costly. A virus can take them down for several days each quarter, and they're still better off. So they standardize on one of the word processors for their company. Other companies may standardize on a different word processor, but then you have to train people when you hire them from other companies. Soon enough a clear leader shows up, and you have a monopoly in the word processor world.
This is assuming they all have the same features. If one has better features (including stability) than the others, then it may win out sooner.
Diversity is a pretty costly solution to the problem. You always have to balance security against other benifits. The problem is that for a lot of companies, a system with poor security suits their needs pretty well. Managers can always hope that a virus won't take them down at a really bad time, but even if it does it may be less costly than some of the solutions.
I'm wondering if Microsoft will have any luck appealing the Judge's decision that IE and Windows are two seperate products. IE is software that allows users to run applications and view information formatted in HTML. How different does this sound than other features of Microsoft's OS. Sure they tested the technology outside of the OS, worked out some bugs then integrated it. That doesn't necessisarily mean it's a seperate product. If they aren't seperate products, there's no tying, and no second market they were trying to monopolize. A lot of it comes down to the definition of the OS that's used. If you use a definition that only includes the kernel and a few utilities in the OS then IE is definately a seperate product. However, there is no OS of significant quantities, which meets that definition in the applicable market. Yet that definition was used. Remember, the court determined that MS has 95% of the market, so they pretty much define the market. If adding browsing features to the OS were a reasonable extention to their product at that time, this whole case quickly becomes irrevelent. Remember, integration is permitted, it's tying that's illegal.
Using longwave tranceivers (1300 nm) FibreChannel can use cable lengths of up to 10 km with 1 Gbps data rate, and you'll start seeing 2 Gbps products real soon. These are't experimental products either, you can buy them now. FibreChannel can aslo use fabric switches to make networks considerable larger than anyone in their right mind would want to make.
The price can be a bit of a problem for a lot of applications right now, but if you need the bandwidth, expandability, and cable lengths, FibreChannel can provide them.
If your card is using a long wave laser (1300 nm)you can reach distances of 10 km without a repeater. This does not mean that the average optical FC card (850 nm) can do this, but versions many vendors cards with long wave lasers do exhist. You can also get 30 m using copper HSSDC, which is a bit cheaper.
1) In their own words, Microsoft indicated that they were giving away IE to undercut Netscape.
IE is used for a lot of things in Windows. It's used for displaying help files in MSDN. They use it for their Windows update. MS uses it for a lot of things. They couldn't do all these things if IE wasn't included in Windows. If they used someone else's browser, such as Netscape, they would be dependent on them to make sure the features they required exhisted, and worked properly. Personally, I wouldn't want core features of my flagship product depending on another company, especially one that appeared hostile. Remember that Netscape was also a buggy piece of crap, even MS developed better software.
How can you tell that MS didn't make a design decision as to how they were going to develop their software. Sure they said they were going to "cut off Netscape's air supply", but was that their original purpose, or an effect of their decision.
2) They went out and paid off several large ISPs (from AOL on down) so that the ISPs distributed IE instead of Netscape.
I think AOL actually paid MS, but as far as some of the other ISP's I think they did pay them for the advertising. This is a pretty common practice, but possibly a bit unfair for a Monopoly.
3) They threatened to raise prices on OEMs such as Compaq if they shipped Netscape preinstalled.
I haven't heard this one before, but it's possible I missed it.
4) They forbade OEMs from placing any other Internet or ISP-related icon on the preinstalled Windows desktop.
Funny, I bought a laptop that came with icons for Netscape, and for setting up a number of ISPs right on the desktop. Actually, what Microsoft was accused of was giving preferential treatment to ISPs who only distributed IE and not Netscape. Specifically, if you wanted an icon on the desktop, you had to only distribute IE.
5) They paid and rewarded large websites that included IE-specific features like Channels.
Sounds like competition to me. You've got to advertise your products.
6) They tied key Windows system DLLs to Internet Explorer, so that some 3rd party applications would be forced to have IE4 as a prerequisite for install.
Were these DLLs related to the browser functionality they put in Windows? If they were, distributing updated DLLs with ID kind of make sense to me.
Any time Microsoft makes a design decision, lots of developers and companies are effected. If they have to go to court everytime they change the way they do things, they can't innovate. People scoff at MS stance that they're just trying to maintain their freedom to innovate, but there seems to be a lot of truth to it. If you think Microsoft is unhelpful, try and get help out of SGI or Sun, when they're developing a competing product for their OS.
The things in this case that MS can't link to a design decision are their OEM pricing, and linking ISP's spot on the desktop to IE. Considering that these are things that most compainies out there do, it's hard to justify breaking up MS over them. These are also things that Microsoft is willing to negociate a settlement about. The appeals court has already put forth their opinion that the courts shouldn't be making design decisions. Now we'll see if Judge Jackson will listen, and deal with the appropriate issues, or if this'll be tied up in appeals for years.
How do you split up a software company? Who gets the key people with the key knowledge of how the software works. You can split MS along product lines, but that really doesn't keep them from doing the things they've done in the past. You can split them into several identical baby Bills, but how do split them evenly when the important resources are likley the employees, who could simply decide to flock to one of the baby Bills, and you're back where you started. The only thing it would really succeed in doing is make Microsoft a confused mess for a while. It's hard to justify that as a long term solution to help consumers, not to mention get it to hold up in an appeal.
If you want to fine them you make it much larger than a couple billion. Then you take the money and pump it into the competitors or perhaps some really nice OSS projects. Imagine just how good linux could become with say $50 billion to spend on getting SMP working *for sure* on say a good sized Sun Enterprise server.
The government may be able to fine Microsoft, but giving the money to competitors or an open source project is not something they can do. It's also in my opinion an extremely bad idea. The last thing you want is the government doling out cash to create a "level playing field". How would they decide who gets the money? What would that do to the stock market? The opertunities for corruption are horrible. Would they give it all to US companies, or share the cash with foreign companies that were allegedly harmed? It would be a nightmare of litigation that would never end.
If Microsoft is ruled a monopoly, the companies they harmed have a right to sue, and I'm sure many will. This case is about how Microsoft may of harmed consumers, not competitors. The government brought this case against Microsoft, not their competitors. If their competitors want to sue them they are welcome to do so.
IMO Microsoft has harmed consumers, but mainly through their abusive pricing mothods and their bundling deals with OEMs. In this case the government seems to be more concerned with protecting the likes of Netscape and AOL than protecting consumers. If the government would quit trying to tell Microsoft how to develop software, and concentrate on their business practices we would likely have a settlement which provides relief to consumers by now.
Although people seem reluctant to admit it there are likly thousands of known bugs in your typical Linux distribution. I don't work on Linux myself, but it seems like our developers that do run into their share of bugs in the kernel, many of which people already know about, and that's just the kernel. Win2000 contains a lot more than what *NIX users would consider a kernel, so comparing it to the Linux kernel isn't really fair.
I'm always skeptical when I hear how open sourcing a project makes bugs disappear. It seems like when there's a bug in LINUX, it's ok, people know about it, and they're working on a fix. It'll likely make it in the next stable kernel release in 6 months or so. If there's a bug in Win2000, then Win2000 is horribly buggy crap, and people people shouldn't have to pay for software that has to be patched. I understand not liking to pay for something that's not exactly what I wanted, but face it, no large software package is ever bug free. MS products do seem to reach a lower quality standard than I'd like, but they do seem to be trying to address this, if somewhat slowly.
Remedies should address the charges. The charges were tying IE to the OS, and something about how MS unfairly prices Windows and Office to OEMs.
OS/2 isn't involved. If Microsoft changed their pricing strategies and seperated out IE, that should address the charges.
However, this case now appears to be a political issue, where the AGs are out to advance their political careers rather than address the needs of consumers. Maybe I'm sceptical, but it looks to me that wether MS wins or loses the consumer is likly to get screwed in the end.
Slashdot Mirror
DLL Hell occurrs when a vendor decideds to overwrite one of microsoft's dlls, which breaks other software which relied on that DLL. It's not really an example of code reuse gone bad. It's more an example of vendors having too much ability to break the OS. MS is actually trying to address this problem in Win2K. After a couple of service packs, we may see if their efforts helped.
In order to get the install user friendly, Linux needs to be more drvice driver writer friendly. Until Linux gets a device driver interface, drivers will have to be recompiled for every kernel version, single and multiprocessors, and even processor optimations. With design cycles for some hardware such as Video cards being as low as 6 months, there's no way the vendors can keep their drivers current and test them on all the kernel versions.
I understand that the Linux kernel is a constantly evolving beast, and that any device driver interface would have to change occasionally, but not having one at all is going to be a tremendous barrier to Linux ever becoming a popular desktop OS.
Even if a company chooses to release Open Source drivers recompiling the kernel is a little beyond what you can expect from the average user. This forces hardware companies to produce several binary versions for each kernel, and maintain them for several kernel versions. This means they have to maintain these versions with bug fixes and test them all. They also have to write complicated install scripts to match the correct binary with the kernel on the user's machines, and deal with all the tech support questions which are greatly compounded by all this.
This pretty much makes Linux the hardest platform to support, and customer support is a large expense for most companies. Since the Linux customer base is still relatively small, this just gives hardware companies one more reason not to bother with Linux.
For some devices the manufacturer can publish a clear specification and a little sample source, and someone will write and contribute the driver. However, if you're targeting consumers, someone has to provide support. Someone is also likely going to have to polish the installation tools so that the average consumer can get the driver working.
We support multiple OSs in all of our drivers where I work. With a little work you can abstract our a lot of the OS specific parts of the driver. You can then write most of the driver in OS independent manner and port it easily. Most of the OS specific parts are also reusable in other drivers. Unfortunately for this to work well you need to have a reasonable device driver interface.
We currently support Linux in our drivers, but creating and testing all the binaries, and the customer support problems are already becomming an issue. It's a lot of effort, and the sales numbers just arent justifing the effort. I don't expect this to continue indefinately.
The reason you the drivers have to be recompiled for each kernel is because Linux doesn't provide a binary driver interface. The interface consists of of calls that do not from one kernel version to the next even though the underlying implementation does change. Every other major OS that I know of provides an interface for drivers. It should even make the kernel itself more maintainable. However, there seems to be considerable oposition to it from the "we don't want anyone writing binary drivers" crowd.
Linux doesn't have a binary driver interface. This means that the drivers need to be recompiled for each kernel. If you want to take advantage of MMX or SSE you'll probably need a version that supports it, and one that doesn't as well. Pretty soon you're testing and shipping a dozen different binaries in order to keep your diverse customer base happy.
Is it REALLY that hard to recompile drivers to make everyone happy?
Recompiling the drivers isn't that difficult. Properly testing the drivers on the different kernels and platforms is often time consuming, tedious work. The open source community may consider it acceptable to just compile the different binaries, do a quick 5 minute test and send out the binary. Our comercial customers expect a bit more. Our drivers go through weeks of testing for reliability, compatibility, and interoperability.
How different is *BSD from Linux as far as drivers go?
BSD has two desirable features in the kernel that Linux does not. Kernel threads (with priorities), and the ability to map user buffers into kernel space.
Kernel threads make it easier to write drivers that share resources well. It also makes it easier for me to make a large portion of my driver OS independent so it's not only portable, but works more consistently across OSs. This keeps the different code bases managable, and reduces debugging efforts.
Mapping user memory into kernel space is necessary in order to avoid copying the users data into or out of a kernel buffer. With the devices for which I'm writing drivers it's not uncommon to be sending data in blocks of 1 MB or more at throughputs arount 100 MB/s. Copying the data is relatively slow, and will quickly peg your CPU load at 100%. If you want the throughput, you need to map the memory into kernel space.
There is a definate drawback to mapping user memory into kernel space. The author of the device driver MUST make sure that the user is passing in a valid buffer, and that the pages remain locked in physical memory. Failure to do this will cause the system to crash. A bad pointer in user space causes the application to crash. A bad pointer in kernel brings down the system. When MS claims the bigest reliability problem NT has is poorly written 3rd party drivers, they aren't making it up.
I'm already paying $0.50 a gallon in taxes on every gallon. Gasoline is a huge industry, and $0.50 a gallon adds up to a huge amount of money. I keep hearing the argument that it's the oil companies that are keeping new technologies from making gasoline obsolete. We live in an economy where there are ambitious individuals with billions of dollars, who would love to make billions more. I really doubt the oil industry is bribing all of them so the don't invest in alternative fuels. Most of the alternatives to gasoline have a common problem. They don't have as high of an eneryg density as you can get with gasoline. In order to give consumers the 250 to 300 mile range they want with their cars, they batteries or natural gas tanks just have to be too large. Fuel cells may be a better solution in the future, but the technology is still too expensive.
The solution of making the car smaller and lighter only works so well. Families have a lot of stuff to haul around. Those little fuel efficient, aerodynamic cars also don't do so well in rural areas where the gravel road gets 6 inch ruts in it from the rain washing it out several times a year. It's also hard to drive in snow when you're pushing snow with your bumper.
There are a lot of us that aren't going to buy a little electric car. If you're commuting in a city a little Honda Insight may be a great choice. However, if you're driving it in the mountains, it just doesn't have enough horsepower to make it up the hills (although you can recharge the batteries on the downslope).
Lets try an example. I'm going camping with some friends in West Virginia. There are four of us going along with all our camping gear. The Insight is a two seater, so we'd have to take 2. Actually, we'd probably have to take three to fit all the gear, but let's say we can do it with two. The next problem is that the 60 to 70 MPG is based on driving on relatively flat ground. You aren't going to get that in the mountains, and the weight of the gear will make it considerably worse. If you're lucky you'll get 50 MPG.
Instead of the 2 Insights, lets take 1 full sized Chevy Silverado with the extended cab. It seats 4 adults comfortably. Plenty of room for gear in the back. A cap on the back is recomended to keep the gear dry. If it's a 2WD truck you'll likely average about 25 - 26 MPG in the mountains. Since you have enough HP the hills don't hurt your MPG as badly. You also don't have that long string of cars behind you as you go up every slope at 25 MPH. You save a time (as long as your not stuck behind the guy in the Insight), and you can actually drive back the dirt road to where you're camping without getting stuck. You even get as good, if not better gas milage on average. If you go with the 4WD you likly drop a few MPG, but you can go more places, and it's a lot better on icy roads.
If you live in a city, and have a short commute, the Insight may be for you, but car pooling will likely do just as good. If you really want good gas milage, I'm sure there are motorcycles that get 70+ MPG. They're not so good in the rain, and the AC doesn't work so well in traffic jams, but if you're looking at the Insight you've already accepted that you're going to have to make some tradeoffs.
The Insight is tiny, and has a 3 cylinder engine in it. I haven't seen any crash test ratings on it yet, but I doubt it holds up too well in a accident. And please don't tell me that the solution to the problem is to get rid of all the SUVs. Larger families need larger vehicles. A minivan would suffice for most, but there are some of us that actually need the ground clearance and the 4wd. I'd love to have better gas mileage, but my next vehicle will likely be a gas guzzling, 4wd truck, because a compact car just doesn't suit my needs. A natural gas truck might be a nice alternative if natural gas were available at gas stations.
I personally don't know what CSS is, but if browser developers aren't bothering to implement it, I doubt that there's a whole lot of demand for it. Standards are created by comittee. The result is usually a comprimais that no one is really happy with, or a bloated standard that has tons of features that are of little use to the majority of developers.
Some are, some aren't. OEMs will install another OS if there's significant demand for it. However, as far as consumer OS's are concerned they haven't had many choices. OS/2 wasn't a consumer OS, too hard to configure and use for the average user. LINUX distributions have made great strides recently, but they still have a ways to go before they are a good choice for the average user.
We live in a world where one person can post copyrighted material on the web, and millions can steal it. The copyright holder can go after the person who posted the material, if they can find them, but they can't practically go after all the people who stole their copyrighted works. How can you sue a million individuals. Our legal system can't deal with this type of crime, and it costs the copyright holder more to pursue the criminals than it's worth.
The result of this probelm has often been copy protection schemes. These schemes usually end up as a deterrent to only those who don't feel it's worth the effort to get around the copy protection. So now we have a law which prohibits circumventing the copy protection. Such a law even makes some sense to me at first thought. The probelm is that the law is being abused. The recording industry is using this law to dictate how you can use the copyrighted materials. They are using "copy protection" to extend limit how you use the material beyond the rights they are given by law. If you circumvent thier "protections" they say you are breaking the law, even if you your actions would have been considered fair use.
I doubt this is what Congress intended when they passed the law, at least not all of the Congressmen that voted for it. The courts need to overturn the parts of the law that interfere with fair use, and Congress needs to create a better balanced copyright law. This time they have to remember that the goal of copyright law is not to simply protect the rights of copyright holders. The law must also limit those rights. The framers of the law must keem in mind the actions of the recording industry to subvert the law and take away the rights of individuals. It's the recording industry that's trying to circumvent the law in this case. The law provides for fair use, and the recording industry is attempting to subvert that. Hopefully the recording industry will lose the lawsuits, but be forced to pay the legal costs of those they are persecuting.
It's reasonable to say that software that has no bugs would comply to the specification. It's however unreasonable to imply the because the software complies with the specification that it has no bugs, unless the spec is intentionally vague. The only way a spec can take into account all the unexpected user inputs that the user may try is to include vague specifications such as "will not crash due to unexpected input" or "will operate in a user friendly manner", both of which I've seen in specifications.
In reality, you can't use complience with the spec to determine if something is or is not a bug. Someone needs to look at the issues, determine if they are worth fixing, and assign some sort of priority if they are worth fixing.
It sounds tike the problem in this case is that the issues were submitted as bugs, not as design issues. However, since the product is a user interface, some of them might be considered bugs.
From what I read, he does have considerable experience in writing user interfaces. It sounds like he went through the user interface in a through fasion and submitted the things he had issues with as bugs.
Things he did right:
1. Tested the portion of the software in which he has considerable experience.
2. Submitted all the issues at once so developers don't have to keep going back and revisiting the issues (hopefully).
Things he did wrong:
1. Submitted things as bugs that others didn't consider to be bugs.
Unless there's some history here that I'm not aware of, or he keeps submitting things this way after he's told to submit them in another way, this doesn't seem like a very big deal.
If you just move people off the project because you don't like their opinions, you end up with poorly designed products, which also happens often enough on commercial products. There's a balancing act to be done, and everyone on the project needs to be somewhat flexible.
From the AP story "The circuit court acted shortly after Microsoft almost simultaneously filed a notice of appeal with Jackson and a motion with the circuit court to stay Jackson's penalties, even though Jackson was also considering a Microsoft stay motion."
Microsoft filed the appeal with Jackson. There seems to be a lot of non-lawyers arguing how M$ actions are unlawful and unconstitutional. I think the DOJ has enough experience with the law that they would be pointing that out if it were true.
It does seem like the appeals court acted very quickly on this, but considering the actions of Jackson and the DOJ it's not that surprising. The behaviorial modifications in Jackson's ruling go into effect in 90 days. 90 days is a pretty short time for a company Microsoft's size to significantly change how they do business. They need to renegociate/rewrite their licensing contracts with all their vendors. This will cost them a large amount of time and money. They obviously don't want to do it if they don't have to. They really don't want to do it twice. Once now and again if they get some or all of the ruling overturned. Microsoft requested a stay on the remedies until the case was appealed. Justice then filed a huge reply and seemed to be stalling. They even went as far as to request that Jackson not rule on the stay until M$ had filed the appeal. It seems like both Justice and Jackson are trying to get the case to the Supreme who won't have time to rule on staying the remedies until after they go into effect.
Why would the appeals court want to hear this case? Maybe because Jackson no longer appears to be an impartial Jurist. Maybe it's because Jackson pretty much ignored their ruling that M$ could integrate IE to Windows. They've ruled on issues related to this case before. Jackson and Justice appear to be trying to bypass them and subvert their authority.
Jackson has been anything but sympathetic to Microsoft. The trial has been rushed, the witnesses very limited. From the outset everyone was predicting that M$ would lose the case, but win on appeal. Now the DOJ and Jackson appear to be trying to limit the appeals process. Everyone has a right to a fair trial, even M$. The more of this crap that gets pulled, the less I have faith in the DOJ and the Judicial system. The worst thing that could happen is that a mistrail could be declared by the appeals court. Then everything starts over again.
IANAL. The president could appoint an Attourney General who would drop the case, but that would be a very politically dangerous thing to do. I think the states could still pursue the case, but I really don't know.
The Sherman act provides for both criminal and civil penalties, however this is a civil case. No criminal charges have been filed, and criminal charges would be handled in a seperate case from the civil case. It may be possible for the government to still file criminal charges under the Sherman act, but it seems unlikly to happen.
So when LINUX becomes a common desktop OS user are going to have to save their email attachments that are to be executed. They're going to have to figure out that the file is to be executed in some way, with no pretty icon (not to mention an extension). They're even going to have to run chmod on the file to get a script to run.
These limitations are all perfectly reasonable on a server OS. They obviously make the system considerably more secure. But if you think that Joe and Jane user who use this as a desktop OS at work or at home are going to figure all this out, I think your overly optimistic.
Companies can probably afford a few minor disasters from viruses than losing the productivity they gain from clicking on e-mail attachments and having them do what the sender intended. I know this is a huge security hole which requires the user to determine if the attachment is safe based on who they think sent them the message.
The question is how much ease of use you want to trade for security. I think the solution that many have suggested of showing a dialog box before outlook lets an application send an email is a good place for MS to start. However, it appears they have made some patches to fix some of their security problems, and that sys admin are very lax in applying them. Hopefully both MS and those sys admins have learned their lessons.
Let me preface this with saying that standard APIs are definatley the most desirable solution, and that I have very little experience in developing OpenGL software. However, there is a problem with how slowly standards tend to evolve. When you have a mature standard such as POSIX or TCP/IP it's much easier to develop to the standard and still put in all the features and performance that your market desires.
Gaming hardware and software is evolving at an amazing rate right now, and deciding standards by committee isn't usually a quick process. You also have the problems of people on the standards committee having conflicting goals. 3dfx may have some feature that no one else has, and want support for it integrated into OpenGL as quickly as possible. Other compainies are going to have other priorities, and feel that 3dfx's feature is just slowing down the implementation of the features they think are improtant.
There are also companies that are unwilling to put their technology into an open standard such as OpenGL. Those same companies may be willing to lisence their technology to Microsoft to put it in DirectX.
Standards like OpenGL are also very succeptable to feature bloat. You get the I'll vote for your feature if you vote for my feature syndrome going, and you end up with a standard which includes so many features that no one can really implement them in drivers.
Microsoft's method of unilaterally creating DirectX can allow them to more quickly add features which take advantage of new technologies. Because it's a closed standard, they can also start working with hardware vendors to add new features under NDA. This allows Microsoft to get a head start on implementing new features, and allows the hardware vendor to wait to announce new features and not tip their hands to their competitors.
If you goal is protability, then standards are the way to go. If your goal is to get the newest features out to your customers, and you're willing to accept that your customer base will be smaller, then a propriatary API may be a reasonable choice.
Unfortunately, it's just not that simple.
Say for an example you have 3 different Word Processors with the same file format. You've just trippled your training budget for Word processors. Your tech support staff now has to support 3 times as many products. You're also 3 times as likely to run into a bug that keeps you from setting up standard templates you're trying to develop, and now the people working on them need to know 3 different word processors and spend time testing on all of them to make sure it all works.
It doesn't take too long until managment and support figure out that this is just too costly. A virus can take them down for several days each quarter, and they're still better off. So they standardize on one of the word processors for their company. Other companies may standardize on a different word processor, but then you have to train people when you hire them from other companies. Soon enough a clear leader shows up, and you have a monopoly in the word processor world.
This is assuming they all have the same features. If one has better features (including stability) than the others, then it may win out sooner.
Diversity is a pretty costly solution to the problem. You always have to balance security against other benifits. The problem is that for a lot of companies, a system with poor security suits their needs pretty well. Managers can always hope that a virus won't take them down at a really bad time, but even if it does it may be less costly than some of the solutions.
I'm wondering if Microsoft will have any luck appealing the Judge's decision that IE and Windows are two seperate products. IE is software that allows users to run applications and view information formatted in HTML. How different does this sound than other features of Microsoft's OS. Sure they tested the technology outside of the OS, worked out some bugs then integrated it. That doesn't necessisarily mean it's a seperate product. If they aren't seperate products, there's no tying, and no second market they were trying to monopolize. A lot of it comes down to the definition of the OS that's used. If you use a definition that only includes the kernel and a few utilities in the OS then IE is definately a seperate product. However, there is no OS of significant quantities, which meets that definition in the applicable market. Yet that definition was used. Remember, the court determined that MS has 95% of the market, so they pretty much define the market. If adding browsing features to the OS were a reasonable extention to their product at that time, this whole case quickly becomes irrevelent. Remember, integration is permitted, it's tying that's illegal.
Using longwave tranceivers (1300 nm) FibreChannel can use cable lengths of up to 10 km with 1 Gbps data rate, and you'll start seeing 2 Gbps products real soon. These are't experimental products either, you can buy them now. FibreChannel can aslo use fabric switches to make networks considerable larger than anyone in their right mind would want to make.
The price can be a bit of a problem for a lot of applications right now, but if you need the bandwidth, expandability, and cable lengths, FibreChannel can provide them.
If your card is using a long wave laser (1300 nm)you can reach distances of 10 km without a repeater. This does not mean that the average optical FC card (850 nm) can do this, but versions many vendors cards with long wave lasers do exhist. You can also get 30 m using copper HSSDC, which is a bit cheaper.
1) In their own words, Microsoft indicated that they were giving away IE to undercut Netscape.
IE is used for a lot of things in Windows. It's used for displaying help files in MSDN. They use it for their Windows update. MS uses it for a lot of things. They couldn't do all these things if IE wasn't included in Windows. If they used someone else's browser, such as Netscape, they would be dependent on them to make sure the features they required exhisted, and worked properly. Personally, I wouldn't want core features of my flagship product depending on another company, especially one that appeared hostile. Remember that Netscape was also a buggy piece of crap, even MS developed better software.
How can you tell that MS didn't make a design decision as to how they were going to develop their software. Sure they said they were going to "cut off Netscape's air supply", but was that their original purpose, or an effect of their decision.
2) They went out and paid off several large ISPs (from AOL on down) so that the ISPs distributed IE instead of Netscape.
I think AOL actually paid MS, but as far as some of the other ISP's I think they did pay them for the advertising. This is a pretty common practice, but possibly a bit unfair for a Monopoly.
3) They threatened to raise prices on OEMs such as Compaq if they shipped Netscape preinstalled.
I haven't heard this one before, but it's possible I missed it.
4) They forbade OEMs from placing any other Internet or ISP-related icon on the preinstalled Windows desktop.
Funny, I bought a laptop that came with icons for Netscape, and for setting up a number of ISPs right on the desktop. Actually, what Microsoft was accused of was giving preferential treatment to ISPs who only distributed IE and not Netscape. Specifically, if you wanted an icon on the desktop, you had to only distribute IE.
5) They paid and rewarded large websites that included IE-specific features like Channels.
Sounds like competition to me. You've got to advertise your products.
6) They tied key Windows system DLLs to Internet Explorer, so that some 3rd party applications would be forced to have IE4 as a prerequisite for install.
Were these DLLs related to the browser functionality they put in Windows? If they were, distributing updated DLLs with ID kind of make sense to me.
Any time Microsoft makes a design decision, lots of developers and companies are effected. If they have to go to court everytime they change the way they do things, they can't innovate. People scoff at MS stance that they're just trying to maintain their freedom to innovate, but there seems to be a lot of truth to it. If you think Microsoft is unhelpful, try and get help out of SGI or Sun, when they're developing a competing product for their OS.
The things in this case that MS can't link to a design decision are their OEM pricing, and linking ISP's spot on the desktop to IE. Considering that these are things that most compainies out there do, it's hard to justify breaking up MS over them. These are also things that Microsoft is willing to negociate a settlement about. The appeals court has already put forth their opinion that the courts shouldn't be making design decisions. Now we'll see if Judge Jackson will listen, and deal with the appropriate issues, or if this'll be tied up in appeals for years.
How do you split up a software company? Who gets the key people with the key knowledge of how the software works. You can split MS along product lines, but that really doesn't keep them from doing the things they've done in the past. You can split them into several identical baby Bills, but how do split them evenly when the important resources are likley the employees, who could simply decide to flock to one of the baby Bills, and you're back where you started. The only thing it would really succeed in doing is make Microsoft a confused mess for a while. It's hard to justify that as a long term solution to help consumers, not to mention get it to hold up in an appeal.
If you want to fine them you make it much larger than a couple billion. Then you take the money and pump it into the competitors or perhaps some really nice OSS projects. Imagine just how good linux could become with say $50 billion to spend on getting SMP working *for sure* on say a good sized Sun Enterprise server.
The government may be able to fine Microsoft, but giving the money to competitors or an open source project is not something they can do. It's also in my opinion an extremely bad idea. The last thing you want is the government doling out cash to create a "level playing field". How would they decide who gets the money? What would that do to the stock market? The opertunities for corruption are horrible. Would they give it all to US companies, or share the cash with foreign companies that were allegedly harmed? It would be a nightmare of litigation that would never end.
If Microsoft is ruled a monopoly, the companies they harmed have a right to sue, and I'm sure many will. This case is about how Microsoft may of harmed consumers, not competitors. The government brought this case against Microsoft, not their competitors. If their competitors want to sue them they are welcome to do so.
IMO Microsoft has harmed consumers, but mainly through their abusive pricing mothods and their bundling deals with OEMs. In this case the government seems to be more concerned with protecting the likes of Netscape and AOL than protecting consumers. If the government would quit trying to tell Microsoft how to develop software, and concentrate on their business practices we would likely have a settlement which provides relief to consumers by now.
Although people seem reluctant to admit it there are likly thousands of known bugs in your typical Linux distribution. I don't work on Linux myself, but it seems like our developers that do run into their share of bugs in the kernel, many of which people already know about, and that's just the kernel. Win2000 contains a lot more than what *NIX users would consider a kernel, so comparing it to the Linux kernel isn't really fair.
I'm always skeptical when I hear how open sourcing a project makes bugs disappear. It seems like when there's a bug in LINUX, it's ok, people know about it, and they're working on a fix. It'll likely make it in the next stable kernel release in 6 months or so. If there's a bug in Win2000, then Win2000 is horribly buggy crap, and people people shouldn't have to pay for software that has to be patched. I understand not liking to pay for something that's not exactly what I wanted, but face it, no large software package is ever bug free. MS products do seem to reach a lower quality standard than I'd like, but they do seem to be trying to address this, if somewhat slowly.
Remedies should address the charges. The charges were tying IE to the OS, and something about how MS unfairly prices Windows and Office to OEMs.
OS/2 isn't involved. If Microsoft changed their pricing strategies and seperated out IE, that should address the charges.
However, this case now appears to be a political issue, where the AGs are out to advance their political careers rather than address the needs of consumers. Maybe I'm sceptical, but it looks to me that wether MS wins or loses the consumer is likly to get screwed in the end.