Read the whole article, and the quote in context makes a bit more sense.
Essentially, Google has the ability, and is starting to display the technology to enable full featured web applications. Once you get fully hooked in, the WPA would do all the heavy lifting of prefecting, high end compression. It would serve you up Web based spreadsheets, Word Processors, heck, even Image editing applications.
Somewhere in the article he talks about Google essentially deploying a cluster of ~2-20M machines. These machines would run those web based applications. You'd save your data on their storage. The WPA is the first step in this process. You start there. Then they have the ability to serve up more content, and take over more responsibilities from your computer.
So eventually, any computer you walk to, as long as it is hooked into Google's WPA, you have all of the standard functionality and data you need when you use your computer.
I'm not sure I believe it, but in context, that quote makes a lot more sense.
Finally, a lot of people don't precisely agree with you on what a thin client is. A thin client most definitly runs it's own OS. A lot of times, it's the same OS you would use on a desktop. My definition of a thin client is: You can throw it away, and replace it with a fresh machine, and modulo minor configuration, you didn't lose any data or functionality. So, by my definition, a fully functionally WPA that stores your data, and has web enabled applications is pretty close. You need something capable of getting onto the internet, and a web broswer that is compatible with WPA. That's pretty close.
I have thin clients that are essentially diskless work stations. They run a full Linux install, but they have no floppy, CD, or disk. They boot off of the network, and use network filesystems to store thing. You still use the local CPU to run all your applications. In terms of administration, you just have to maintain the boot image. Now, on some of those, I've got them setup so that only Mozilla runs on the local CPU, while all other applications run over X. Thus, I only have to maintain a very small boot image, and for web based work, the user gets pretty much the full capacity of the machine. It's cheaper to buy full desktops and strip them, then to buy honest to goodness terminals from what I've seen.
What precisely is RPM sub-standard at? I've seen people talk about the beauty of.deb packages. It's my understanding that.deb is fairly isomorphic to RPM. Name something specific, or link to something specific that an RPM can't accomplish that some other packaging file format does so much better. Don't talk about dependency resolution, that's a function of apt-get, yum, or some other program.
As a general rule the strength or weakness of the distributions packages has less to do with the package file format, and more to do with the tender loving care devoted to each package in terms of specifing all of it'd dependencies, what it obsoletes, what functionality it provides.
There are some packages that are a pain in the ass in RPM format (RedHat's BIND/named packages jump to mind). Not having used a.deb based distro I long term, I don't know of any historically badly packaged applications from Debian.
As a general rule, I haven't had any serious problems with RPM's in years. They work just as well as any others. I use almost exclusively from RedHat (I do use a handful from freshrpms and Dag). They work just fine. Especially since I started using yum, it's generally a command line to update my system. So stop using the "Chewbakka Offense", and actually be specific. I've seen you make several posts that just assume that it's mathematically proven that RPM's are incapable of caputing the esscense of package management. I'm unaware of it's deficiencies.
I've always liked it conceptually, as then I can have it has all the advantages of separate machines on a single machine, with few of the downsides.
It makes it much easier to do security. I can limit which machines with with services can interact via a firewall much easier as each machine will have a different IP.
If one kernel deadlocks, or I need to change a setting that needs a reboot (very rare with a Linux box, but they do exist). I can reboot each service independently. Generally, any PAM or libc update is an automatic reboot for me (I want to ensure that everything is using the new libc or PAM authentication libraries, and it's very difficult to ensure that without essentially doing a reboot).
If administered properly (assuming the Hypervisor or whatever it is called in the technology you are deploying gives you control), it means that I can have a great deal of control about which services can dominate the machine. Despite all the promises of the UNIX scheduler and ulimits, I've found that I still can't control which services can use how much memory, or control how much they thrash the I/O channels as well as I'd like. I'm told that VMWare has much better support for this.
Finally, it's a very, very good way to sandbox things. If you want to know how something will act under load, or see how the new distribution will act under heavy load on the server hardware, you can just start a new image and install it. Now naturally, you can do this on a development box, but it means if you want to migrate the base OS that the service runs on, you can have them running in parallel and cut over when you want (which normally would require two machines).
I understand your point of view. I hate the fact that we have lots of redundant servers around our office. We have lots of machines that do essentially nothing most of the time. However it's still the safest way to ensure redundancy and failure conditions can be met. I'd much rather run two huge machines with virtual images for all of our stock services, each being a hot failover of the other.
If someone steals SSNs of college students and uses them 10-20 years down the road, chances are these people will have perfect credit, and won't even know where the attack came from
Where did you go to school? They actually teach college students about money management and how to improve your credit score. Don't post where it is, Discover will go there, and dump credit cards until they ruin a good thing.
In my experience, most college students do more harm to their credit scores in college then they can recover from in 10 years. Maybe 20 they could recover from. Most people leave college so debt laden it's silly. Credit card companies prey on students on college campuses. I was always shocked at home many places on campus had credit card offers. Remember, college is the new high school. College in the 1960's was a 25% of HS grads went. Now it's more like 75% go. Going to college isn't the indicator it used to be.
I happen to have decent credit, but that has a lot more to do with watching my family memebers have poor credit, and poor money management. I sure didn't learn a thing about it in college.
Thank you captain obvious. When you define "customers" to be whomever gives me money. Yes they are in fact customers.
However, sales is a bit easier when you are talking about dealing with companies who have departments (possibly divisions or business units), whose sole job it is to deal with companies who have cool technology they want to sell. Microsoft is always in the market for new features for their software. You don't have to convince them they need new features. You just have to convince them, they need YOUR new feature. Lots easier then a lot of the sales stuff I've seen people who have to do it.
Your also dealing with people who have a budget they need to spend. Serious bonus when selling.
There's no way in hell, I want to sell a Web Browser, or an e-mail client. However, there are lots of times where I could see selling a technology to significantly improve some aspect of one of those two things. I'd much rather deal with selling myself Microsoft, Sun, RedHat, or IBM instead of dealing attemting to convince the public at large they should use my crappy e-mail client because it has this one nice feature I have expertise in. There is no way I can compete with Microsoft's Outlook in terms of features and convince a user that I'm any good. I could however, convince Microsoft I have a very, very good anti-SPAM technology they could purchase from me for $2Mil instead of developing it themselves. There are lots of people who have done this.
The business model is about building expertise and technology in a specific realm that has a specific value to a large company with the assests to purchase you.
Yes, but the idea of a lot of startups isn't to find customers. It's to build something customers need, and then convince someone who has a product to integrate it the technology into their products.
So develop a wonderful grammer checker. YOu don't sell that to end users. You sell it to Sun, IBM, and Microsoft.
You don't develop an e-mail client, you develop a SPAM filter. You sell it to anyone with a good client that needs a better SPAM filter.
Any number of people have gotten wildly rich developing interesting technology and then getting purchased by Microsoft, Sun, HP, or IBM.
I've seen roughly this same concept described on business2.com. I can't find the link in my Sent e-mails. It'd be no good unless you are a subscriber anyways (I'm not). Essentially developing cool technology that is useful when integrated into another product is a great way to get bought up if you are in Silicon Valley. That essentially, you need to bank two years of living money. Work until the money runs out, or you get bought up. Use that money to start the next startup. Never work longer then two years on the project. Repeat until you decide you have enough money to retire.
So go find something that Excel, Outlook, Linux, WebSphere, or something else does that is stupid that needs to be improved. Improve it. Show the proof of concept to the big company that should want your technology. Sell it to them. It's a novel concept.
Yes, and no. Using that sort of logic, Sun could never add a new feature to their JRE until they'd added to to everyone elses.
The problem Sun had with Microsoft's Java was that Microsoft was giving access to Win32-only API's, so that the source that used them would run only on a platform that supported Win32. Sun accused them of attempting to take a language they had worked hard to make platform independent, and tie it directly to Win32.
If Microsoft was making extensions that were useful and didn't need Win32 to implement, my guess is that Sun wouldn't have been so upset about it.
The analogous problem would be if Sun implemented stuff that would only work under Solaris. The JRE works on all of the platforms it's released on (at least it's supposed to).
Nothing I've read about the Sun Only extenions are inheriently unimplementable by anyone else on a standard Java platform. When they refer to them as Sun only. I'm guessing it's an API that Sun is working on standardizing, but wants to make it available for use to shake out defects in the API. That, and/or its so new that no one else has had time to implement it.
Past employers cannot legally give any information beyond "This person worked here from xx to xx".
You are going to have to cite a statue for that one. I'm incredibly dubious of the claim that I'd be in violation of the law to say more then "this person worked here". I'm unaware of any legal princepal of "Employee-Employer Privacy". Heard of it in the case of laywers and doctors, but never in the case of my boss. I'm fairly confident they can say if you were terminated with cause or not (that's got a legal definition, I'm unaware of all of the details).
As a general rule, most employers will not say any more then that out of fear of legal retaliation if you fail to get the job. There is precedent for suing former employers due to bad references. Employees have one such cases. I'm unaware of the details, but I see that it is making things harder for interviewers to get enough information to get such an assessment. Hence as a general rule, there's a bit of a wink, wink, nudge, nudge going on during some reference calls to employers who want to say bad things, but don't want to face the legal repercussions.
I know that it's in our hand book that if I am called as a reference for an employee who worked at my company, I can be fired for discussing past employees. However, that's because I'm creating a legal liability for the company and have no formal HR training to know what the laws and repercussions of what say.
As a general rule, always say: "I didn't give that person permission to use me as a reference". Anyone who asks who you would not be a good reference for inform them that shouldn't be using you as a reference. Thus what you are saying is honest and true. Anyone you want to say good things about give them permission. Then say only really nice things about them.
Here
is what was the best description I could find of the details behind the guidgen based off the top ten links from Google when putting "guidgen" into the system.
Even putting your MAC addr and using the time, it's still completely possible for two completely different machines to generate exactly the same GUID. I've know people who changed their MAC address (various reasons, including the cable modem providers that only accept the original MAC address you signed up with). I know some who just change it to DEADBEEF for giggles. I've know lots of people with clocks that are set wrong.
There's no "guaranteed" about it. It is "so incredibly unlikely that you should be more worried about being hit by lightening and eaten by a shark at the same time". Oddly enough that was the metaphor that was the subject of a college graduation speech I listened to. Still good advice (there are some things worth worring about, and some things not). In the end, it's nothing but a Random Number generator that uses your MAC address and the time as seeds in the generation. We can debate the semantics all you want, but in the end, that's all it is. In the end, there's nothing mathematically eliminating a collision.
Hell, theres nothing to say it can't run so fast on a machine built 30 years from now that the same machine couldn't generate one twice within a single clock resolution. Before you run off and tell me that's impossible, you really should read up on the great phone blackout in the North East seaboard. Some AT&T people thought the same thing about messages being generated. They were given unique identifiers based on time. Eventually the hardware got fast enough that it could generate two messages within the resolution of the clock units they used. One of the most spectacular failures of one of the most reliable networks on the planet, because of just that reasoning.
Got me, I've haven't installed Ubuntu. The last time I installed debian was probably 4 years ago. It was a black and white text screen. However, the screen shots that got linked to have basically the same color scheme as RedHat's installer. The "pick languages" screen appears to be just like the Anaconda one in terms of size and layout.
The progress meters are the same size and shape as Anaconda ones. If it wasn't the same code, they should have used the same code.
There might be plenty of differences under the hood in terms of code, but the look and feel of those screens shots strongly reminded me of Anaconda, and for that matter, any of the RedHat System Admin tools that are the text version. Up2date, redhat-config-network, whatever. They all look roughly like that. I put that together with the Progeny port of Anaconda and wondered aloud if the screen shots above were the result of that work.
That sure has a look and feel of the Anaconda installer that RedHat developed, and that Progeny was working towards getting Debian installable by.
In fact there was a Slashdot story about that Here. I wonder if it is. Granted it looks like one of many, many crappy curses applications, but nearly everything about it is just like the text installer that I associate with RedHat. It certainly looks like it'd fit right in with any of RedHat's system configuration tools also.
That's blantantly wrong if guidgen uses a random number generator (I'm unfamiliar with that one, but I've read up on other libraries).
The sequence "2", "2" is just as likely any other 64-bit number coming out of a random number generator.
It might be highly improbably, but then so is any other series of numbers you get out of a 64 or 128 bit random number generator. It's thinking like your is that convinces people like my friends, that they have a way to beat the house at craps by using past performance as a guide to future performance.
It might be 10^10 more likely that you'll win the lottery twice on consecutive week using the same numbers then generate two UUID that match, but that doesn't mean it's "really guaranteed" as in "mathematically proven to be impossible", which seems to be your implication.
Nothing like some guy who works part-time after high-school at EB lecturing the Net on the console biz.
Hmmm, that's funny. For whatever it's worth, I'm a College Grad, majored in Math and CS got a BS. I took several economics classes in College and have read up on economics, and statistical measures in economics in particular. Both as an interest in investing, purely for fun, and because I've considered becoming an actuary. I've got the necessary skills and background to become an economist if I really wanted to.
It doesn't take any insight into the console business in order to realize that's a really stupid mistake to make that comparison.
Here
is some good reference material were a world famous economist goes into detail about economic measures and statistics. One of the points he drives home in detail, is that all statistics have to be considered in context. One of the major considerations is seasonal adjustments. It's important in any retail business. It's important in any personal real estate stats. It's important when considering the price of Oil.
He talks about how a couple of bad snow storms in April can make the housing market look horrific. When if you take into consideration the weather, the numbers aren't so bad. He discusses how the Department of Labor does such things, and how they make refinements over time to their methodology. That comparing numbers between two different methodologies can at points be misleading.
Finally, my favorite math book of all time, is "How to Lie with Statistics". Wonderful book. It should be required reading by anyone who reads news articles. People misuse, or misinterpret statistics all the time.
The author of the article clearly didn't take seasonal issues into consideration when he compared the retail Q4 to retail Q1 (if he did, he should have mentioned it). You can call me a High School drop out if you like (no matter how wrong that is). Sure I'm an arm chair economist, but I sure didn't make such a stupid comparison while writting for "GamesIndustry.biz". I didn't post such a silly quote on the front page of an/. either. However, I can see a bogus comparsion when it's blatantly obvious.
but can equally be interpreted in light of the fact that the previous quarter, ended on December 31st
Okay, let me explain this to you. Christmas is the biggest, baddest ass time of year when it comes to retail business. I want to say like 35% of all retail business takes place in the month of December.
Game publishers try to get their best games out around Christmas seaons, because they know it's a great way to have a block buster game.
If X-Box has equalled the quarter that had Christmas during any quarter that doesn't have Christmas in it. Either they did something incredibly wrong during Christmas, or they did something incredibly right during the non-Christmas quarter.
There's a reason they seasonally adjust all retail numbers when doing economic analysis. It's because not all quarters are equal.
Unless LOGO counts (and even then it wasn't really programming), I know I started with BASIC on an Apple IIc. Pounding out code in 1987. Programmed in basic until 1992, when I first learned Pascal. Didn't learn C until 1995. Had done some C++ prior to that, but didn't really learn how to be good at that until sometime in the 96 or 97. Been programming professionally in C/C++ since '96. Did some assembly around 94 or 95 for fun.
Earning above median pay for a while now, and I started with BASIC. I believe out of the "Basic Apple BASIC" book. Still have it at home.
I know there are some people who cut their teeth on Java nowadays. In my experience, they are more broken then those of us who had to do assembly and C at one point. Java coddles you too much, and lets you do too many things you think are sophisticated without real work. In my experience, Pascal and C/C++ are still good learning languages. Java isn't so bad at it, but you have to dump classes and GUI's in fairly early, because standard terminal I/O in java is horrific, and you can't write a free standing function. It's nice to know that if I have to look at a hex dump of machine code with the assembly next to it, I can figure it out. You can't tell if your compilers is broken if you can't do that.
So I'll have to provide a second piece of anecdotal evidence that BASIC doesn't fundamentally break you.
The concept I remember most from "Code Complete" is comments should tell you why or what are you doing something, but it shouldn't tell you how you are doing. The how should be obvious from the code.
Never tell me you are adding shifting the varaible by 4 bits, instead tell me, you are converting from bytes to 16 byte blocks due to a chunks size conversion because that's the size the output device expects blocks to be written in.
Don't tell me you are swapping variables, tell me why the values should be exchanged.
Don't tell me you searching for the max value, tell me what the max value is used for when you find it. That sort of concept.
One easy way to do that, is to look at code and the comments. If the code and the comments could get out of sync because you changed the implementation, the comment is wrong. You documented the how not the what.
There are cases where the how is important, especially when there are things where the "how" doesn't behave naturally. Oh, this sorts numbers, but it is intentionally sorting the ASCII values of the numbers when represented as a string. That'd be very useful to know. I'd expect a sort to work based on the binary values.
The other rule I remember from code complete, and from "Writting Solid Code", was the concept of laying out the pseudo code that explained what you wanted to do at a high level. Then filling in between the comments with the implementation. It was a very good way to end up with documented code.
I'm confident there should be. Like I said, I know several browsers, including galeon have the XUL (or whatever the interfaces are written in), to do such a thing.
Googling around, this page shows that there is something called "Firefox Ultrabar". I don't use Firefox (but I do use Mozilla). I'm not much of a plug-in head. However, assembling a dictionary.com HTTP request isn't significantly different, so presuambly, it should be trivial. Try putting these into google: "Firefox dictionary.com plugin". I'm sure the answer is there somewhere. I believe both of these links will be interesting, the second more then the first:
Lots of plug ins and Dictionary specific plug ins
The other way, is just put the misspelled word into google. Most of the time, it will give you a better spelling (it did for your word). If you just want to know what the word means, you can do put this into the query box: "define: fubar".
dictionary.com
never leave home without it. Oh, and don't misspell it, as you'll get sent to a crappy squatter sites.
There are several very good dictionary sites on the web. Heck, I believe that Galeon used to ship with it and Google up next to the tool bar so you always had an quick link to them.
The same reasons why the FCC shouldn't regulate an emerging telecom industry (like VoIP), is relatively similar to why patents shouldn't hold up emerging markets.
Personally, I think patents should be shorter period. However, I think for software that's doubly true. The role of patents and copyright are to "to promote the Progress of Science and the useful Arts" (quoted from the US Constitution). I'm fairly confident that more patents are currently stifling innovation then promoting it in the area of software. Right now, there's literally no program you can write that doesn't infringe on some sort of patent. You can't write an Operating system. You can't write a compiler. You can't write a word processor. You can't write a mapping application. You can't write a spreadsheet. You probably can't write a GUI application period. I'll be slashcode infringes on patents. The entire field is covered with patents. A lot of them are unenforced, and a lot of the patents are probably bogus. However, it does mean that big patent holders can come use that leverage to execute a cashectomy, or force you to license your patent to them for terms favorable to them. It's a lot easier to write something using techniques that are well known from 4-5 years ago, instead of using techniques well known 20 years ago.
I mean if you really want to get to the meat of the problem. There should be finanical incentive for the patent offices to reject a patent. Second, the patent office shouldn't be a profit center for the gov't. (Right now the patent office is, and has been for a while. Congress cut off some of their funding, and they are now generating more revenue in fees then cost to operate). Finally, the Patent office should start paying people assine amounts of money. Being a patent clerk should be the single most competitive job in America. Simply because the patent office is staffed by people who couldn't get jobs in industry (or they really wanted to work for the patent office). The patent office doesn't get the best and the brightest in a lot of fields. They could if they paid well, and gave the job some prestige.
Finally, patents should be forced to be written as originally intended. They should clearly explain how to accomplish something. If I hand a patent to an expert in the field, and he can't tell you exactly what it does in a reading or two, it should be rejected out of hand as obscure. You get a gov't granted monopoly for spelling out how you are doing something. The concept that IBM got a patent for flushing a toilet is assine. All of it came down to either overworked and non-expertise in the patent office, or the clever language used to hide what was really being described.
I'm not sure how I feel about polynomial approximation to NP. I'd have to see the whole of it and know the details. I'm fairly confident that it shouldn't be patented. It might be patented as applied to a given scope (lets say for optimizing plane routing). However, it doesn't mean anyone who uses a similar technique applied to other industries should be subject to your patent. Now because you figured it out, the all of the US industries that could use that are beholden to you. That's incredibly stifiling. It means we'll have to wait 20 years for you to give it up, or for you to license it. Depending on exactly the terms you use to license it, you could do industry exclusive licensing. Which isn't terribly capitalistic, and it really isn't doing much to advance science or the useful arts for the next 20 years.
That was a problem with management at the company, not anything wrong with the Research people themselves. Several of the research people (See Bob Metcalf? The guy who founded 3Com and invented Ethernet), left the company precisely because they knew the management of the company had no clue what to do with the new stuff.
Xerox couldn't see past the end of their domination in copiers. My Dad worked for Xerox (not for PARC). It was just a culture thing there. If it wasn't a copier they just didn't have much interest in diversifing into that area. While they made computers and used them, they didn't do much to push them as a product. I used to use WordStar on their computers as a kid.
I'm thinking the parent has it slightly wrong, and your interpretting it even more wrong...
When you say "Xerox PARC" as an example of an R&D place. Uhh, they built a machine with a mouse. They built an Ethernet card. They had working networking. They are an example of everything that is right about R&D. They did good stuff. It's really too bad Xerox didn't think outside of the realm of copiers. They pretty much owned the computer market 5-10 years before it existed.
I think the parent to your post had it conceptually wrong in terms of "product". I think he should modify that to be "once they have a working proof of concept", they should be allowed to patent it to protect themselves while they turn the concept into a product (I think this is the one legitimate use of patents, to protect smaller companies from larger ones while they are turning their concept into a product).
The problem with software patents are that some of them are just stupid. Anything that is "I'm automating something done by hand with a computer", shouldn't be patentable. It is no longer "novel" to automate any kind of process thru software. So everyone who patents essentially a business model thru software (my software does X, and I re-sell X as a service so I'm given a government granted monopoly on the concept of automating service X). I know there were a couple of guys who did this for automating importing and exporting. It automatically filled in some gov't forms. Got a patent, essentially tried to run every one of his competitors out of business because they used computers for some form of automation.
The other problem with software patents, is that 17 years (or 20 years from application date), is just assinely long in terms of computers. Just think if someone had patented the "mouse" when the Mac came out. That would mean you'd have had bought their mouse up until Jan of 2001 (using the 17 years from application rule).
Conceptually no one will get to implement "one click" purchasing until what, 2017 (I think they applied for the patent in 1997)? Geez, that sounds like a fair amount of time. Lets see, how much has the computing world advanced since 1997? How much since 1987?
Just think if HTML, or a Web Browser were patented so that we would have to nicely ask permission to use such concepts? It'd badly stifle innovation. If they we're talking about letting you have a patent that could protect you for up 20 years, but you only got a gov't granted monopoly once you've productized it for say 2-4 years. I'd say that's a bit more sane the then current system. It could probably still be "gamed" to gain an unfair advantage, but it sure would be nice to see fast moving markets be relatively patent free.
The GPL requires that you put no further limitations on the re-distribution of said document.
I'll also point out that for a LaTeX document, I'm not giving you the prefered source format. If I generated the document from LaTeX source, and gave you the PDF version with an embedded GPL'ed font, you could easily claim I didn't give you the "preferred format". If I printed the document and handed it to you, that's not in electronic format, which is against the GPL (I'd have to at the very least give you a written offer for the document in electronic format good for at least 3 years if I remember the clauses correctly).
You couldn't print Trade Secrets in a font that is GPL'ed, as it would inheriently have more limitations added to it that the GPL doesn't allow.
Personally, I think GPL'ing a font is a lot like me saying I'm GPL'ing the blueprints to my house. It doesn't make any sense. I suppose it makes some, in the context of a derived work. However, I'm fairly doubtful it'd stand up to scruity in court. It probably means you are a copyright infringer, but as a license, it's fairly incoherent when applied to a font.
Now imagine if the 100,000+/. readers all donated $5 - $10 a year./. wouldn't need stupid ads.
I'm fairly sure that the economics of Slashdot are a lot higher then you think. I'm nearly positive that the IT costs of running slashdot on a yearly basis, exceeds $0.5-1.0Mil USD. Call it a hunch. I've never priced out that high end a set of bandwidth. However, at $40-$85K a month, might get you the kind of bandwidth and Co-Lo's needed to support Slashdot. However, when you couple that with paying what 5-10 full time employees, it'd never happen.
All that aside, and replying to the story, I mostly agree with you. I wish they sold two versions of everything. I'd happily pay evaluate cable as a "No commercials" thing. Even if it meant there were 8 minutes of dead air. I'd prefer the kind of feed that crazy fella in Toronto has (there was a slashdot story about a guy who among a half dozen other things, essentially setup his on neigborhood cable provider service and bought TV content in bulk).
I'd pay extra for no commercials. Happily. I buy a lot of shows on DVD specifically for that purpose. I've never ripped any DVD's, but eventually I'll get to it.
I'm waiting for the day when producing serial shows direct to DVD is tried (or TV on demand like was described a couple of weeks ago in Cringly's PBS column). Just think of the kind of fun that JMS, Joss Wedon, or Matt Groening (Babylon 5, Firefly, and The Simpsons creators respectively) could have marketing their shows directly. If they had the finanical muscle to pull it off, it'd be tons of fun. The problem is the startup capital. However, there are several people who have enough "cult of personality" to help launch that as a economically viable business model. (The biggest problem would be the timing and lack of feedback about the direction the show was taking). There are several people who'd I'd buy a series they produced on DVD for $75 for 22hours of entertainment, sight unseen.
No Apollo mission killed anyone. The plugs out test that was re-named "Apollo 1" afterwards killed three astronauts, including the second American to go into space. Grissom, White, and Jaffie if I remember the names correctly. The next space flight was "Apollo 8". More people die in the space program flying back and forth during training then they have in space.
I'm unaware of the mission that so badly missed Mars (not to say it didn't happen, but it's not ringing any bells). I know there was one where a British contractor expected input in feet and NASA fed it data in meters. I believe that crashed into Mars.
I know that the EU managed to plunge one of their satallites into Mars. Not sure anyone figured that one out (Beagle?)
Everyone knows that's just the Martian missle defense...
Challenger blew up because the people at NASA in charge caved had fairly systemic failures. From what I've read, they literally died because that teacher was on board. They didn't want to miss the launch as it had strong political implications. Google for Richard Feynmann's Appendix on the Challenger Accident. He discusses the wrong headedness of NASA's decision, and that they had a lot of the information to realize it was a problem. His analogy to Russian Roulette is scary, but true.
Here are some decent links:
http://www.westgard.com/guest25.htm
http://www.fotuva.org/feynman/challenger-appendix. html
You really should have taken more accounting, and learn a bit more about the IT industry if you think that is all there really is to it.
There are several other considerations:
Taxes
The tax issues alone could override the costs of the equipment itself. Depending on the situation the company is it and how much taxable income it has, will completely change the dynamic of this. That's not even discussing the costs of figuring out the taxes (at points, the man power to accurately figure out what the savings would be costs more then the savings, gotta love the IRS).
Cash Flow
We lease a ton of equipment and arrange for a $1 buyout at the end of the lease. Which essentially means we are taking a loan out and purchasing the computers. We structure this as a lease because a bank wouldn't loan us money at a reasonable interest rate given our credit risk (small company w/ no profit history is a huge risk), so we could finance the debt rather then have to pony up all of the capital to purchase the computer equipment we needed.
In the end, we probably would have gone out of business if we hadn't done this. At some point, we wouldn't have been able to make payroll, but we would have plenty of assests for the bankruptcy auction. Instead, on paper we were in the red for a while, but still had the cash on hand to make it thru to profitability. It simplifies the taxes, and allows us to hold onto maximal amounts of cash which is what you need to keep employees coming back in to work.
Migration Costs
Those are the business ends of things. There are still more things to think about in terms of the IT end of things. If you have a real lease, the buyout price, and migrating off the hardware has a serious cost associated with it. So you'll have to account for that when evaluating the options. We've had desktops that had a buyout price of $500, for a machine that would cost $300 to purchase new. However, the deployment costs of the new machines, it was nearly a wash when it came to manpower to deploy them (we probably would have replaced them if only because 2-3 years out, the parts failure would have eaten us alive). We ended up threating to send the machines back, at which point the vendor said: "KEEP THEM!, we don't know what the heck we can do with a 3 year old machine". We've slowly replaced the big wave of leases with smaller sets of purchased machines.
Slashdot Mirror
Essentially, Google has the ability, and is starting to display the technology to enable full featured web applications. Once you get fully hooked in, the WPA would do all the heavy lifting of prefecting, high end compression. It would serve you up Web based spreadsheets, Word Processors, heck, even Image editing applications.
Somewhere in the article he talks about Google essentially deploying a cluster of ~2-20M machines. These machines would run those web based applications. You'd save your data on their storage. The WPA is the first step in this process. You start there. Then they have the ability to serve up more content, and take over more responsibilities from your computer.
So eventually, any computer you walk to, as long as it is hooked into Google's WPA, you have all of the standard functionality and data you need when you use your computer.
I'm not sure I believe it, but in context, that quote makes a lot more sense.
Finally, a lot of people don't precisely agree with you on what a thin client is. A thin client most definitly runs it's own OS. A lot of times, it's the same OS you would use on a desktop. My definition of a thin client is: You can throw it away, and replace it with a fresh machine, and modulo minor configuration, you didn't lose any data or functionality. So, by my definition, a fully functionally WPA that stores your data, and has web enabled applications is pretty close. You need something capable of getting onto the internet, and a web broswer that is compatible with WPA. That's pretty close.
I have thin clients that are essentially diskless work stations. They run a full Linux install, but they have no floppy, CD, or disk. They boot off of the network, and use network filesystems to store thing. You still use the local CPU to run all your applications. In terms of administration, you just have to maintain the boot image. Now, on some of those, I've got them setup so that only Mozilla runs on the local CPU, while all other applications run over X. Thus, I only have to maintain a very small boot image, and for web based work, the user gets pretty much the full capacity of the machine. It's cheaper to buy full desktops and strip them, then to buy honest to goodness terminals from what I've seen.
Kirby
As a general rule the strength or weakness of the distributions packages has less to do with the package file format, and more to do with the tender loving care devoted to each package in terms of specifing all of it'd dependencies, what it obsoletes, what functionality it provides.
There are some packages that are a pain in the ass in RPM format (RedHat's BIND/named packages jump to mind). Not having used a .deb based distro I long term, I don't know of any historically badly packaged applications from Debian.
As a general rule, I haven't had any serious problems with RPM's in years. They work just as well as any others. I use almost exclusively from RedHat (I do use a handful from freshrpms and Dag). They work just fine. Especially since I started using yum, it's generally a command line to update my system. So stop using the "Chewbakka Offense", and actually be specific. I've seen you make several posts that just assume that it's mathematically proven that RPM's are incapable of caputing the esscense of package management. I'm unaware of it's deficiencies.
Kirby
It makes it much easier to do security. I can limit which machines with with services can interact via a firewall much easier as each machine will have a different IP.
If one kernel deadlocks, or I need to change a setting that needs a reboot (very rare with a Linux box, but they do exist). I can reboot each service independently. Generally, any PAM or libc update is an automatic reboot for me (I want to ensure that everything is using the new libc or PAM authentication libraries, and it's very difficult to ensure that without essentially doing a reboot).
If administered properly (assuming the Hypervisor or whatever it is called in the technology you are deploying gives you control), it means that I can have a great deal of control about which services can dominate the machine. Despite all the promises of the UNIX scheduler and ulimits, I've found that I still can't control which services can use how much memory, or control how much they thrash the I/O channels as well as I'd like. I'm told that VMWare has much better support for this.
Finally, it's a very, very good way to sandbox things. If you want to know how something will act under load, or see how the new distribution will act under heavy load on the server hardware, you can just start a new image and install it. Now naturally, you can do this on a development box, but it means if you want to migrate the base OS that the service runs on, you can have them running in parallel and cut over when you want (which normally would require two machines).
I understand your point of view. I hate the fact that we have lots of redundant servers around our office. We have lots of machines that do essentially nothing most of the time. However it's still the safest way to ensure redundancy and failure conditions can be met. I'd much rather run two huge machines with virtual images for all of our stock services, each being a hot failover of the other.
Kirby
Where did you go to school? They actually teach college students about money management and how to improve your credit score. Don't post where it is, Discover will go there, and dump credit cards until they ruin a good thing.
In my experience, most college students do more harm to their credit scores in college then they can recover from in 10 years. Maybe 20 they could recover from. Most people leave college so debt laden it's silly. Credit card companies prey on students on college campuses. I was always shocked at home many places on campus had credit card offers. Remember, college is the new high school. College in the 1960's was a 25% of HS grads went. Now it's more like 75% go. Going to college isn't the indicator it used to be.
I happen to have decent credit, but that has a lot more to do with watching my family memebers have poor credit, and poor money management. I sure didn't learn a thing about it in college.
Kirby
However, sales is a bit easier when you are talking about dealing with companies who have departments (possibly divisions or business units), whose sole job it is to deal with companies who have cool technology they want to sell. Microsoft is always in the market for new features for their software. You don't have to convince them they need new features. You just have to convince them, they need YOUR new feature. Lots easier then a lot of the sales stuff I've seen people who have to do it.
Your also dealing with people who have a budget they need to spend. Serious bonus when selling.
There's no way in hell, I want to sell a Web Browser, or an e-mail client. However, there are lots of times where I could see selling a technology to significantly improve some aspect of one of those two things. I'd much rather deal with selling myself Microsoft, Sun, RedHat, or IBM instead of dealing attemting to convince the public at large they should use my crappy e-mail client because it has this one nice feature I have expertise in. There is no way I can compete with Microsoft's Outlook in terms of features and convince a user that I'm any good. I could however, convince Microsoft I have a very, very good anti-SPAM technology they could purchase from me for $2Mil instead of developing it themselves. There are lots of people who have done this.
The business model is about building expertise and technology in a specific realm that has a specific value to a large company with the assests to purchase you.
Kirby
So develop a wonderful grammer checker. YOu don't sell that to end users. You sell it to Sun, IBM, and Microsoft.
You don't develop an e-mail client, you develop a SPAM filter. You sell it to anyone with a good client that needs a better SPAM filter.
Any number of people have gotten wildly rich developing interesting technology and then getting purchased by Microsoft, Sun, HP, or IBM.
I've seen roughly this same concept described on business2.com. I can't find the link in my Sent e-mails. It'd be no good unless you are a subscriber anyways (I'm not). Essentially developing cool technology that is useful when integrated into another product is a great way to get bought up if you are in Silicon Valley. That essentially, you need to bank two years of living money. Work until the money runs out, or you get bought up. Use that money to start the next startup. Never work longer then two years on the project. Repeat until you decide you have enough money to retire.
So go find something that Excel, Outlook, Linux, WebSphere, or something else does that is stupid that needs to be improved. Improve it. Show the proof of concept to the big company that should want your technology. Sell it to them. It's a novel concept.
Kirby
The problem Sun had with Microsoft's Java was that Microsoft was giving access to Win32-only API's, so that the source that used them would run only on a platform that supported Win32. Sun accused them of attempting to take a language they had worked hard to make platform independent, and tie it directly to Win32.
If Microsoft was making extensions that were useful and didn't need Win32 to implement, my guess is that Sun wouldn't have been so upset about it.
The analogous problem would be if Sun implemented stuff that would only work under Solaris. The JRE works on all of the platforms it's released on (at least it's supposed to).
Nothing I've read about the Sun Only extenions are inheriently unimplementable by anyone else on a standard Java platform. When they refer to them as Sun only. I'm guessing it's an API that Sun is working on standardizing, but wants to make it available for use to shake out defects in the API. That, and/or its so new that no one else has had time to implement it.
Kirby
You are going to have to cite a statue for that one. I'm incredibly dubious of the claim that I'd be in violation of the law to say more then "this person worked here". I'm unaware of any legal princepal of "Employee-Employer Privacy". Heard of it in the case of laywers and doctors, but never in the case of my boss. I'm fairly confident they can say if you were terminated with cause or not (that's got a legal definition, I'm unaware of all of the details).
As a general rule, most employers will not say any more then that out of fear of legal retaliation if you fail to get the job. There is precedent for suing former employers due to bad references. Employees have one such cases. I'm unaware of the details, but I see that it is making things harder for interviewers to get enough information to get such an assessment. Hence as a general rule, there's a bit of a wink, wink, nudge, nudge going on during some reference calls to employers who want to say bad things, but don't want to face the legal repercussions.
I know that it's in our hand book that if I am called as a reference for an employee who worked at my company, I can be fired for discussing past employees. However, that's because I'm creating a legal liability for the company and have no formal HR training to know what the laws and repercussions of what say.
As a general rule, always say: "I didn't give that person permission to use me as a reference". Anyone who asks who you would not be a good reference for inform them that shouldn't be using you as a reference. Thus what you are saying is honest and true. Anyone you want to say good things about give them permission. Then say only really nice things about them.
Kirby
Even putting your MAC addr and using the time, it's still completely possible for two completely different machines to generate exactly the same GUID. I've know people who changed their MAC address (various reasons, including the cable modem providers that only accept the original MAC address you signed up with). I know some who just change it to DEADBEEF for giggles. I've know lots of people with clocks that are set wrong.
There's no "guaranteed" about it. It is "so incredibly unlikely that you should be more worried about being hit by lightening and eaten by a shark at the same time". Oddly enough that was the metaphor that was the subject of a college graduation speech I listened to. Still good advice (there are some things worth worring about, and some things not). In the end, it's nothing but a Random Number generator that uses your MAC address and the time as seeds in the generation. We can debate the semantics all you want, but in the end, that's all it is. In the end, there's nothing mathematically eliminating a collision.
Hell, theres nothing to say it can't run so fast on a machine built 30 years from now that the same machine couldn't generate one twice within a single clock resolution. Before you run off and tell me that's impossible, you really should read up on the great phone blackout in the North East seaboard. Some AT&T people thought the same thing about messages being generated. They were given unique identifiers based on time. Eventually the hardware got fast enough that it could generate two messages within the resolution of the clock units they used. One of the most spectacular failures of one of the most reliable networks on the planet, because of just that reasoning.
Kirby
The progress meters are the same size and shape as Anaconda ones. If it wasn't the same code, they should have used the same code.
There might be plenty of differences under the hood in terms of code, but the look and feel of those screens shots strongly reminded me of Anaconda, and for that matter, any of the RedHat System Admin tools that are the text version. Up2date, redhat-config-network, whatever. They all look roughly like that. I put that together with the Progeny port of Anaconda and wondered aloud if the screen shots above were the result of that work.
In fact there was a Slashdot story about that Here. I wonder if it is. Granted it looks like one of many, many crappy curses applications, but nearly everything about it is just like the text installer that I associate with RedHat. It certainly looks like it'd fit right in with any of RedHat's system configuration tools also.
Kirby
The sequence "2", "2" is just as likely any other 64-bit number coming out of a random number generator.
It might be highly improbably, but then so is any other series of numbers you get out of a 64 or 128 bit random number generator. It's thinking like your is that convinces people like my friends, that they have a way to beat the house at craps by using past performance as a guide to future performance.
It might be 10^10 more likely that you'll win the lottery twice on consecutive week using the same numbers then generate two UUID that match, but that doesn't mean it's "really guaranteed" as in "mathematically proven to be impossible", which seems to be your implication.
Kirby
Hmmm, that's funny. For whatever it's worth, I'm a College Grad, majored in Math and CS got a BS. I took several economics classes in College and have read up on economics, and statistical measures in economics in particular. Both as an interest in investing, purely for fun, and because I've considered becoming an actuary. I've got the necessary skills and background to become an economist if I really wanted to.
It doesn't take any insight into the console business in order to realize that's a really stupid mistake to make that comparison.
Here is some good reference material were a world famous economist goes into detail about economic measures and statistics. One of the points he drives home in detail, is that all statistics have to be considered in context. One of the major considerations is seasonal adjustments. It's important in any retail business. It's important in any personal real estate stats. It's important when considering the price of Oil.
He talks about how a couple of bad snow storms in April can make the housing market look horrific. When if you take into consideration the weather, the numbers aren't so bad. He discusses how the Department of Labor does such things, and how they make refinements over time to their methodology. That comparing numbers between two different methodologies can at points be misleading.
Finally, my favorite math book of all time, is "How to Lie with Statistics". Wonderful book. It should be required reading by anyone who reads news articles. People misuse, or misinterpret statistics all the time.
The author of the article clearly didn't take seasonal issues into consideration when he compared the retail Q4 to retail Q1 (if he did, he should have mentioned it). You can call me a High School drop out if you like (no matter how wrong that is). Sure I'm an arm chair economist, but I sure didn't make such a stupid comparison while writting for "GamesIndustry.biz". I didn't post such a silly quote on the front page of an /. either. However, I can see a bogus comparsion when it's blatantly obvious.
Kirby
Okay, let me explain this to you. Christmas is the biggest, baddest ass time of year when it comes to retail business. I want to say like 35% of all retail business takes place in the month of December.
Game publishers try to get their best games out around Christmas seaons, because they know it's a great way to have a block buster game.
If X-Box has equalled the quarter that had Christmas during any quarter that doesn't have Christmas in it. Either they did something incredibly wrong during Christmas, or they did something incredibly right during the non-Christmas quarter.
There's a reason they seasonally adjust all retail numbers when doing economic analysis. It's because not all quarters are equal.
Kirby
Earning above median pay for a while now, and I started with BASIC. I believe out of the "Basic Apple BASIC" book. Still have it at home.
I know there are some people who cut their teeth on Java nowadays. In my experience, they are more broken then those of us who had to do assembly and C at one point. Java coddles you too much, and lets you do too many things you think are sophisticated without real work. In my experience, Pascal and C/C++ are still good learning languages. Java isn't so bad at it, but you have to dump classes and GUI's in fairly early, because standard terminal I/O in java is horrific, and you can't write a free standing function. It's nice to know that if I have to look at a hex dump of machine code with the assembly next to it, I can figure it out. You can't tell if your compilers is broken if you can't do that.
So I'll have to provide a second piece of anecdotal evidence that BASIC doesn't fundamentally break you.
Kirby
Never tell me you are adding shifting the varaible by 4 bits, instead tell me, you are converting from bytes to 16 byte blocks due to a chunks size conversion because that's the size the output device expects blocks to be written in.
Don't tell me you are swapping variables, tell me why the values should be exchanged.
Don't tell me you searching for the max value, tell me what the max value is used for when you find it. That sort of concept.
One easy way to do that, is to look at code and the comments. If the code and the comments could get out of sync because you changed the implementation, the comment is wrong. You documented the how not the what.
There are cases where the how is important, especially when there are things where the "how" doesn't behave naturally. Oh, this sorts numbers, but it is intentionally sorting the ASCII values of the numbers when represented as a string. That'd be very useful to know. I'd expect a sort to work based on the binary values.
The other rule I remember from code complete, and from "Writting Solid Code", was the concept of laying out the pseudo code that explained what you wanted to do at a high level. Then filling in between the comments with the implementation. It was a very good way to end up with documented code.
Kirby
Googling around, this page shows that there is something called "Firefox Ultrabar". I don't use Firefox (but I do use Mozilla). I'm not much of a plug-in head. However, assembling a dictionary.com HTTP request isn't significantly different, so presuambly, it should be trivial. Try putting these into google: "Firefox dictionary.com plugin". I'm sure the answer is there somewhere. I believe both of these links will be interesting, the second more then the first: Lots of plug ins and Dictionary specific plug ins
The other way, is just put the misspelled word into google. Most of the time, it will give you a better spelling (it did for your word). If you just want to know what the word means, you can do put this into the query box: "define: fubar".
Kirby
There are several very good dictionary sites on the web. Heck, I believe that Galeon used to ship with it and Google up next to the tool bar so you always had an quick link to them.
Kirby
Personally, I think patents should be shorter period. However, I think for software that's doubly true. The role of patents and copyright are to "to promote the Progress of Science and the useful Arts" (quoted from the US Constitution). I'm fairly confident that more patents are currently stifling innovation then promoting it in the area of software. Right now, there's literally no program you can write that doesn't infringe on some sort of patent. You can't write an Operating system. You can't write a compiler. You can't write a word processor. You can't write a mapping application. You can't write a spreadsheet. You probably can't write a GUI application period. I'll be slashcode infringes on patents. The entire field is covered with patents. A lot of them are unenforced, and a lot of the patents are probably bogus. However, it does mean that big patent holders can come use that leverage to execute a cashectomy, or force you to license your patent to them for terms favorable to them. It's a lot easier to write something using techniques that are well known from 4-5 years ago, instead of using techniques well known 20 years ago.
I mean if you really want to get to the meat of the problem. There should be finanical incentive for the patent offices to reject a patent. Second, the patent office shouldn't be a profit center for the gov't. (Right now the patent office is, and has been for a while. Congress cut off some of their funding, and they are now generating more revenue in fees then cost to operate). Finally, the Patent office should start paying people assine amounts of money. Being a patent clerk should be the single most competitive job in America. Simply because the patent office is staffed by people who couldn't get jobs in industry (or they really wanted to work for the patent office). The patent office doesn't get the best and the brightest in a lot of fields. They could if they paid well, and gave the job some prestige.
Finally, patents should be forced to be written as originally intended. They should clearly explain how to accomplish something. If I hand a patent to an expert in the field, and he can't tell you exactly what it does in a reading or two, it should be rejected out of hand as obscure. You get a gov't granted monopoly for spelling out how you are doing something. The concept that IBM got a patent for flushing a toilet is assine. All of it came down to either overworked and non-expertise in the patent office, or the clever language used to hide what was really being described.
I'm not sure how I feel about polynomial approximation to NP. I'd have to see the whole of it and know the details. I'm fairly confident that it shouldn't be patented. It might be patented as applied to a given scope (lets say for optimizing plane routing). However, it doesn't mean anyone who uses a similar technique applied to other industries should be subject to your patent. Now because you figured it out, the all of the US industries that could use that are beholden to you. That's incredibly stifiling. It means we'll have to wait 20 years for you to give it up, or for you to license it. Depending on exactly the terms you use to license it, you could do industry exclusive licensing. Which isn't terribly capitalistic, and it really isn't doing much to advance science or the useful arts for the next 20 years.
Kirby
Xerox couldn't see past the end of their domination in copiers. My Dad worked for Xerox (not for PARC). It was just a culture thing there. If it wasn't a copier they just didn't have much interest in diversifing into that area. While they made computers and used them, they didn't do much to push them as a product. I used to use WordStar on their computers as a kid.
Kirby
When you say "Xerox PARC" as an example of an R&D place. Uhh, they built a machine with a mouse. They built an Ethernet card. They had working networking. They are an example of everything that is right about R&D. They did good stuff. It's really too bad Xerox didn't think outside of the realm of copiers. They pretty much owned the computer market 5-10 years before it existed.
I think the parent to your post had it conceptually wrong in terms of "product". I think he should modify that to be "once they have a working proof of concept", they should be allowed to patent it to protect themselves while they turn the concept into a product (I think this is the one legitimate use of patents, to protect smaller companies from larger ones while they are turning their concept into a product).
The problem with software patents are that some of them are just stupid. Anything that is "I'm automating something done by hand with a computer", shouldn't be patentable. It is no longer "novel" to automate any kind of process thru software. So everyone who patents essentially a business model thru software (my software does X, and I re-sell X as a service so I'm given a government granted monopoly on the concept of automating service X). I know there were a couple of guys who did this for automating importing and exporting. It automatically filled in some gov't forms. Got a patent, essentially tried to run every one of his competitors out of business because they used computers for some form of automation.
The other problem with software patents, is that 17 years (or 20 years from application date), is just assinely long in terms of computers. Just think if someone had patented the "mouse" when the Mac came out. That would mean you'd have had bought their mouse up until Jan of 2001 (using the 17 years from application rule).
Conceptually no one will get to implement "one click" purchasing until what, 2017 (I think they applied for the patent in 1997)? Geez, that sounds like a fair amount of time. Lets see, how much has the computing world advanced since 1997? How much since 1987?
Just think if HTML, or a Web Browser were patented so that we would have to nicely ask permission to use such concepts? It'd badly stifle innovation. If they we're talking about letting you have a patent that could protect you for up 20 years, but you only got a gov't granted monopoly once you've productized it for say 2-4 years. I'd say that's a bit more sane the then current system. It could probably still be "gamed" to gain an unfair advantage, but it sure would be nice to see fast moving markets be relatively patent free.
Kirby
I'll also point out that for a LaTeX document, I'm not giving you the prefered source format. If I generated the document from LaTeX source, and gave you the PDF version with an embedded GPL'ed font, you could easily claim I didn't give you the "preferred format". If I printed the document and handed it to you, that's not in electronic format, which is against the GPL (I'd have to at the very least give you a written offer for the document in electronic format good for at least 3 years if I remember the clauses correctly).
You couldn't print Trade Secrets in a font that is GPL'ed, as it would inheriently have more limitations added to it that the GPL doesn't allow.
Personally, I think GPL'ing a font is a lot like me saying I'm GPL'ing the blueprints to my house. It doesn't make any sense. I suppose it makes some, in the context of a derived work. However, I'm fairly doubtful it'd stand up to scruity in court. It probably means you are a copyright infringer, but as a license, it's fairly incoherent when applied to a font.
Kirby
I'm fairly sure that the economics of Slashdot are a lot higher then you think. I'm nearly positive that the IT costs of running slashdot on a yearly basis, exceeds $0.5-1.0Mil USD. Call it a hunch. I've never priced out that high end a set of bandwidth. However, at $40-$85K a month, might get you the kind of bandwidth and Co-Lo's needed to support Slashdot. However, when you couple that with paying what 5-10 full time employees, it'd never happen.
All that aside, and replying to the story, I mostly agree with you. I wish they sold two versions of everything. I'd happily pay evaluate cable as a "No commercials" thing. Even if it meant there were 8 minutes of dead air. I'd prefer the kind of feed that crazy fella in Toronto has (there was a slashdot story about a guy who among a half dozen other things, essentially setup his on neigborhood cable provider service and bought TV content in bulk).
I'd pay extra for no commercials. Happily. I buy a lot of shows on DVD specifically for that purpose. I've never ripped any DVD's, but eventually I'll get to it.
I'm waiting for the day when producing serial shows direct to DVD is tried (or TV on demand like was described a couple of weeks ago in Cringly's PBS column). Just think of the kind of fun that JMS, Joss Wedon, or Matt Groening (Babylon 5, Firefly, and The Simpsons creators respectively) could have marketing their shows directly. If they had the finanical muscle to pull it off, it'd be tons of fun. The problem is the startup capital. However, there are several people who have enough "cult of personality" to help launch that as a economically viable business model. (The biggest problem would be the timing and lack of feedback about the direction the show was taking). There are several people who'd I'd buy a series they produced on DVD for $75 for 22hours of entertainment, sight unseen.
Kirby
I'm unaware of the mission that so badly missed Mars (not to say it didn't happen, but it's not ringing any bells). I know there was one where a British contractor expected input in feet and NASA fed it data in meters. I believe that crashed into Mars.
I know that the EU managed to plunge one of their satallites into Mars. Not sure anyone figured that one out (Beagle?)
Everyone knows that's just the Martian missle defense...
Challenger blew up because the people at NASA in charge caved had fairly systemic failures. From what I've read, they literally died because that teacher was on board. They didn't want to miss the launch as it had strong political implications. Google for Richard Feynmann's Appendix on the Challenger Accident. He discusses the wrong headedness of NASA's decision, and that they had a lot of the information to realize it was a problem. His analogy to Russian Roulette is scary, but true.
Here are some decent links:
http://www.westgard.com/guest25.htm
http://www.fotuva.org/feynman/challenger-appendix. html
Kirby
There are several other considerations:
Taxes
The tax issues alone could override the costs of the equipment itself. Depending on the situation the company is it and how much taxable income it has, will completely change the dynamic of this. That's not even discussing the costs of figuring out the taxes (at points, the man power to accurately figure out what the savings would be costs more then the savings, gotta love the IRS).
Cash Flow
We lease a ton of equipment and arrange for a $1 buyout at the end of the lease. Which essentially means we are taking a loan out and purchasing the computers. We structure this as a lease because a bank wouldn't loan us money at a reasonable interest rate given our credit risk (small company w/ no profit history is a huge risk), so we could finance the debt rather then have to pony up all of the capital to purchase the computer equipment we needed.
In the end, we probably would have gone out of business if we hadn't done this. At some point, we wouldn't have been able to make payroll, but we would have plenty of assests for the bankruptcy auction. Instead, on paper we were in the red for a while, but still had the cash on hand to make it thru to profitability. It simplifies the taxes, and allows us to hold onto maximal amounts of cash which is what you need to keep employees coming back in to work.
Migration Costs
Those are the business ends of things. There are still more things to think about in terms of the IT end of things. If you have a real lease, the buyout price, and migrating off the hardware has a serious cost associated with it. So you'll have to account for that when evaluating the options. We've had desktops that had a buyout price of $500, for a machine that would cost $300 to purchase new. However, the deployment costs of the new machines, it was nearly a wash when it came to manpower to deploy them (we probably would have replaced them if only because 2-3 years out, the parts failure would have eaten us alive). We ended up threating to send the machines back, at which point the vendor said: "KEEP THEM!, we don't know what the heck we can do with a 3 year old machine". We've slowly replaced the big wave of leases with smaller sets of purchased machines.
Kirby