Well, it does depend on what I mean by luck... I think anyone who manages to come up with an innovative product before everyone else is lucky. Expanding that initial good fortune to gain substantial market share requires competence. Competence is easy enough to find. But I'd expect most experiencd businessmen would have managed to do as well as Google, had they got their foothold.
Well yes, Duke is an exception. 35 people would be quite a large team unless it also includes QA and admin. And 85 is huge even if it also includes staff shared with other projects.
Okay - 30 developers, 100K each, 2 years would be $6 million. These are fairly typical ballpark figues but there other costs as well. Dev teams range from half that size to about 3 times that size, employee costs are probably fairly variable and depend a lot on location. Development times are usually at least a year and rarely more than 3 (BOCTAOE). So lets say between $3 million and $60 million.
MySpace is not fundamentally different from offering generic webspace. The safe harbor provisions cover this. It's hard to argue that MySpace is not an ISP under the terms of the DMCA.
The fact that to deny responsibility, the ISP is better off not policing their network is hardly the ISP's fault. It's a badly drafted law. Perhaps Universal should have thought about thiswhen lobbying for it.
It's a fairly weak cypher. You can do it with pen and paper given a little patience. Replace a, with a, b with b and so on... (Wrapping around at z, of course). Then it all becomes clear
I think MicroSoft's best bet at success would be a heavily stripped down version of Windows CE.
This would work pretty well. At least according to MS (and why would they lie? Oh. right.) CE is very customisable and runs on some realtively low specced PDAs. The bits that aren't needed can be removed, and the licencing costs are lower than XP. When you have a $100 machine this is a very significant part of the overall cost.
There somewhat is. There's a namespace inside std that includes templated functions for ==, !=...
Useful. But surely the == operator is less efficient. C# automatically gives you a += operator when you overload + but doesn't do the same for comparisons. This would be quite simple to implement in any given language. A > B is equivalent to B It's not very fine-grained unfortunately; like you can't specify "I want to provide the other operators for this type but not this type" I think.
Well, having an operator you don't need isn't going to cause any real problems and you can still override them. Although that isn't always going to behave quite how you'd like. This is my favourite gripe with C++. Templates and ovrrides have the wrong precedence. If you have a template function or operator, and you provide an explicit implementation for a certain class, C++ still uses the template for subclasses.
I did hear of an attempt to reduce tailgating by automatically flashing cars when they were too close. This meant that people would slam on the brakes when they got close causing the guy who was tailgating them to crash into them.
But how many serious accidents does tailgating cause? Cars are quite good with rear end shunts. And dents can be fixed. People require a little more work.
Popular culture is a significant problem. There are far too many Star {Wars|Trek|Gate} articles. There's a Wikipedia article for every Star Wars comic book.
It's a geek culture rather than a pop culture thing. Kim possible's mum gets an entry, but Hasselhoff's character Mitch Buchannon redirects to Baywatch. Star Trek has every episode covered and many of the starship classes. It's not like Baywatch was unpopular but it was clearly less so amongst the wikipedia editting crowd.
If I leave some drugs in the back of a car, and someone steals that car, then later on, he decides to give me a gift of a few thousand dollars, am I really a drug dealer?
I mean I'm not totally sure whether Jimmy Wales really qualifies under the guidelines, but I guess his name has been mentioned in a fe other journals. And while there have been a number of editors of fairly successful print magazines who aren't listed, irregular webcomic creator David Morgan-Mar seems to get a mention because he publishes a webcomic and a few silly computer programs. Okay, I like the comic, but does he really pass the eligibility guidelines?
If I decide my garage door opener uses a frequency somewhere in the middle of the FM band, should this prevent the FCC from allocating that space to any radio station?
The garage door openers are at fault for using a military frequency. The fact that they got there first doesn't give them the right to use it. They could have used the 2.4GHz band with no problems from anything.
Actually, I'm not sure why I'm being so coy. I haven't worked for the company for 6 years... So, good guess.
But really it's not worth being too hard on nVidia. STM is a huge company, roughly the same size as Texas Instruments so loss of the graphics division hardly made an impact on them. If they really wanted to, they could have stood up to nVidia by legal action or by playing the same dirty tricks.
And if your "compiler doesn't understand", hwo on earth would doing TWO assignments instead of ONE be faster?
A good optimiser will eliminate simple assignments. It's just assigning a new variable idnetifier to the same register or address.
Taking the address of a local variable may make the compiler put the variable into local memory. It would then have to be copied back. So not only could this require 2 copies, but it would also have to do this via memory which will usually be slower than via a register.
Why didnt your company tried to build their own boards ?
Not sure. Might have been the corporate policy of not competing with our customers. If it was our only product we would have had to do this. But this was just the graphics division of a huge larger company. If they wanted to, they could have given the chips away and competed in the same underhand way as nVidia. But they didn't want to.
Or pair-up with a smaller board manufacturer company that did not buy from nVidia.
You'd be surprised just how few of these exist. I think just about any PC hardware company with any interest in making graphics cards was already buying from nVidia.
It wasn't exactly a performance chip, but did compete with nVidias more profitable, older chips. And yes, there were other problems. It was a reasonably successful chip, but we could only really sell to Hercules.
It did depend somewhat on the game, and the resolution. T&L at the time wasn't all that well supported for DirectX games. If you used lightmaps instead of vertex lighting, then the Kyro would have done well because it was very efficient at texturing. Anything that did a lot of overdraw and blending worked very well, and the way the texturing unit worked, there was a significant drop in improvement in a geForce if you went to a fifth texture stage since that required multipass. Granted, this is kind of academic since nobody would that, but we did make a Quake level that did this and a few other tricks, and the Kyro outperformed a GF2 GTS.
As our first chip in the line, it had a few problems. We didn't quite get the clockspeeds we wanted. I think we were 10-20% off our target speed, but really the problem was that the Kyro was almost a year later than it should have been. Releasing a DX7 part shortly after DX8 was released was a bad idea. In hindsight, we could have done things differently. I think if they'd stuck with it, it would have been a pretty successful line of chips.
Slashdot Mirror
Well, it does depend on what I mean by luck... I think anyone who manages to come up with an innovative product before everyone else is lucky. Expanding that initial good fortune to gain substantial market share requires competence. Competence is easy enough to find. But I'd expect most experiencd businessmen would have managed to do as well as Google, had they got their foothold.
Their product is text ads. Not search technology. The search engine is just a hook.
Someone else could have come up with the text ads earlier. They didn't. Google got there first.
People are forgetting the secret to Google's success.
Luck.
They developed the right product at the right time. Microsoft did the same. They happened to be home when IBM called and got the DOS contract.
heir graduates can come up with quality product but will they be able to provide somethign the market really needs?
Well yes, Duke is an exception. 35 people would be quite a large team unless it also includes QA and admin. And 85 is huge even if it also includes staff shared with other projects.
Okay - 30 developers, 100K each, 2 years would be $6 million. These are fairly typical ballpark figues but there other costs as well. Dev teams range from half that size to about 3 times that size, employee costs are probably fairly variable and depend a lot on location. Development times are usually at least a year and rarely more than 3 (BOCTAOE). So lets say between $3 million and $60 million.
MySpace is not fundamentally different from offering generic webspace. The safe harbor provisions cover this. It's hard to argue that MySpace is not an ISP under the terms of the DMCA.
The fact that to deny responsibility, the ISP is better off not policing their network is hardly the ISP's fault. It's a badly drafted law. Perhaps Universal should have thought about thiswhen lobbying for it.
It's a fairly weak cypher. You can do it with pen and paper given a little patience. Replace a, with a, b with b and so on... (Wrapping around at z, of course). Then it all becomes clear
I think MicroSoft's best bet at success would be a heavily stripped down version of Windows CE.
This would work pretty well. At least according to MS (and why would they lie? Oh. right.) CE is very customisable and runs on some realtively low specced PDAs. The bits that aren't needed can be removed, and the licencing costs are lower than XP. When you have a $100 machine this is a very significant part of the overall cost.
There somewhat is. There's a namespace inside std that includes templated functions for ==, != ...
Useful. But surely the == operator is less efficient. C# automatically gives you a += operator when you overload + but doesn't do the same for comparisons. This would be quite simple to implement in any given language. A > B is equivalent to B
It's not very fine-grained unfortunately; like you can't specify "I want to provide the other operators for this type but not this type" I think.
Well, having an operator you don't need isn't going to cause any real problems and you can still override them. Although that isn't always going to behave quite how you'd like. This is my favourite gripe with C++. Templates and ovrrides have the wrong precedence. If you have a template function or operator, and you provide an explicit implementation for a certain class, C++ still uses the template for subclasses.
I'm talking about tailgate detectors.
I did hear of an attempt to reduce tailgating by automatically flashing cars when they were too close. This meant that people would slam on the brakes when they got close causing the guy who was tailgating them to crash into them.
But how many serious accidents does tailgating cause? Cars are quite good with rear end shunts. And dents can be fixed. People require a little more work.
I'd prefer my taxes to go up.
Asshatism is a little too subjective. I'd be concenred that I might qualify under some definition.
Is the ticketting reducing the number of accidents?
If it fails to do this then it's taxation rather than law enforcement.
Don't know. I'd imagine it was crushed, but I'd just abandoned it anyway since it had negligible value.
Popular culture is a significant problem. There are far too many Star {Wars|Trek|Gate} articles. There's a Wikipedia article for every Star Wars comic book.
It's a geek culture rather than a pop culture thing. Kim possible's mum gets an entry, but Hasselhoff's character Mitch Buchannon redirects to Baywatch. Star Trek has every episode covered and many of the starship classes. It's not like Baywatch was unpopular but it was clearly less so amongst the wikipedia editting crowd.
If I leave some drugs in the back of a car, and someone steals that car, then later on, he decides to give me a gift of a few thousand dollars, am I really a drug dealer?
I mean I'm not totally sure whether Jimmy Wales really qualifies under the guidelines, but I guess his name has been mentioned in a fe other journals. And while there have been a number of editors of fairly successful print magazines who aren't listed, irregular webcomic creator David Morgan-Mar seems to get a mention because he publishes a webcomic and a few silly computer programs. Okay, I like the comic, but does he really pass the eligibility guidelines?
I declare you the winner.
Not quite sure what you;re the winner of. This was a battle of wits and I'm sure at least one opponent was unarmed.
If I decide my garage door opener uses a frequency somewhere in the middle of the FM band, should this prevent the FCC from allocating that space to any radio station?
The garage door openers are at fault for using a military frequency. The fact that they got there first doesn't give them the right to use it. They could have used the 2.4GHz band with no problems from anything.
You're right. I tried this and get exactly the same code with both methods.
Actually, I'm not sure why I'm being so coy. I haven't worked for the company for 6 years... So, good guess.
But really it's not worth being too hard on nVidia. STM is a huge company, roughly the same size as Texas Instruments so loss of the graphics division hardly made an impact on them. If they really wanted to, they could have stood up to nVidia by legal action or by playing the same dirty tricks.
Seven lines of code is clearer than one line now?
Yes. Unless you like code like *a += (*b-=3).
And if your "compiler doesn't understand", hwo on earth would doing TWO assignments instead of ONE be faster?
A good optimiser will eliminate simple assignments. It's just assigning a new variable idnetifier to the same register or address.
Taking the address of a local variable may make the compiler put the variable into local memory. It would then have to be copied back. So not only could this require 2 copies, but it would also have to do this via memory which will usually be slower than via a register.
I worked for STMicro. Not PowerVR. So we were the big company that shoved you out of the way. Not the smaller company that froze you out.
Why didnt your company tried to build their own boards ?
Not sure. Might have been the corporate policy of not competing with our customers. If it was our only product we would have had to do this. But this was just the graphics division of a huge larger company. If they wanted to, they could have given the chips away and competed in the same underhand way as nVidia. But they didn't want to.
Or pair-up with a smaller board manufacturer company that did not buy from nVidia.
You'd be surprised just how few of these exist. I think just about any PC hardware company with any interest in making graphics cards was already buying from nVidia.
It wasn't exactly a performance chip, but did compete with nVidias more profitable, older chips. And yes, there were other problems. It was a reasonably successful chip, but we could only really sell to Hercules.
It did depend somewhat on the game, and the resolution. T&L at the time wasn't all that well supported for DirectX games. If you used lightmaps instead of vertex lighting, then the Kyro would have done well because it was very efficient at texturing. Anything that did a lot of overdraw and blending worked very well, and the way the texturing unit worked, there was a significant drop in improvement in a geForce if you went to a fifth texture stage since that required multipass. Granted, this is kind of academic since nobody would that, but we did make a Quake level that did this and a few other tricks, and the Kyro outperformed a GF2 GTS.
As our first chip in the line, it had a few problems. We didn't quite get the clockspeeds we wanted. I think we were 10-20% off our target speed, but really the problem was that the Kyro was almost a year later than it should have been. Releasing a DX7 part shortly after DX8 was released was a bad idea. In hindsight, we could have done things differently. I think if they'd stuck with it, it would have been a pretty successful line of chips.