Re:Manual length and Macs vs. PC
on
Macintosh Clustering
·
· Score: 5, Insightful
But if the one-page document is a "Quick Start" guide (and the document is entitled "Pooch Quick Start") and the 230 page book is a detailed technical reference discussing all of the important aspects of designing, building, using, administering, and programming a cluster, as appears to be the case in this instance, then the relative sizes of the documents says absolutely nothing about any human factors.
In fact, my first inclination is to try to use the Beowulf stuff rather than Pooch simply because such a detailed work exists and is available for Beowulf clusters, but I don't know if any such information exists for Pooch.
Lois Bujold has said in public that Science Fiction is a reflection of the society that exists when the work is created, not a prediction of the future, and I believe her. It is, in my opinion, a fool's errand to talk about how one writer or another predicted something. Most of the time, an SF writer simply takes a currently existing invention and plays games with it. The other times, the writer talks about something he or she earnestly wants, but hasn't seen yet. Heinlein's waterbed is one of those sorts of things.
For example, by the time Friday was released, in 1984, as I recall, publically available computer terminals were in existence, BBSes were how you got on-line (except for the fortunate few how knew about and had access to Usenet) and networked BBSes were about to be invented.
My own personal favorite example of an SF prediction is in Bellamy's Looking Backward which, among other things, talked about how the broadcasting of music (live performances over telephone lines as neither audio recording nor radio had been invented or conceived of when the book was written) had become common. I also seem to recall that it had some bit in there about how that led to fewer people being able to play the piano, but that may be my subsequent experiences leaking over as it's been 20+ years since I read that book.
However, it seems to me that the question is not about predictions in SF that come true, but about how SF has driven invention. If, as I say I believe above, SF is a reflection of the culture it's written in, then there can be no direct link. However, I also believe that invention is also a product of the culture it is in, so it is certainly fair to say that, if a work doesn't have a direct effect on invention, then it will necessarily reflect the environment in which the invention is made. Rarely is this made more clear than in "The Man Who Sold the Moon" where Delos D. Harriman talks about what it was like to grow up in the early part of the 20th century.
Further, if one wishes to look at that aspect more closes, I think that one could do worse than looking at the work of Dr. Lienhard of the University of Houston (not his son, who is a professor at MIT) who has a 5-minute daily radio program (and book derived therein) called "The Engines of our Ingenuity" which discusses the whole process of invention and covers quite well the methods by which people derive inspiration. The URL to reach the radio show's transcripts is
http://www.uh.edu/engines
I did not say that the Web is filled with "free, almost-instant, accurate help". Instead, I said that it's more useful than telephone tech support. You snipped the part about the "well-formed question with a known answer" that's the critical bit. If the answer isn't known, it doesn't matter how you search for the answer because you're not going to find it.
For the cases you describe, it sounds as if the answer isn't known. Too bad. In that case, you can either figure out the problem is yourself or suffer through it, which, at 03:17, are basically the same choices you have with commercial software and expensive support agreements. If they've got to call the developers in to write a patch, you're screwed anyway.
Of course, documentation that doesn't suck would be nice. It'd be nice if one got documentation that doesn't suck from Microsoft, Oracle, and Sun, too. However, you have to take what you can get because, well, it's what you get. Oh, and personally, I think there's quite a bit of freeware documentation that doesn't suck. Not nearly all of it, by any stretch of the imagination, but quite a bit.
I don't "poo poo" the notion that support is an issue with free software, I just wonder why you define what you get when you call a telephone support line as "support".
Look, say the database goes down at 3:17 AM. You're running Microsoft SQL server with one of those premium contracts, (this is a mission-critical application after all and time is, therefore, a lot of money,) so you call Microsoft's tech support line. You don't get to talk to a human, of course. Instead, you get to listen to "Microsoft Radio" for four hours while the only awake human in the place finishes up some hot chat. Once you do get a support person on the line, you can ask your question and hear the known answer or you can ask your question and get "we don't know" or "it's in the third-party package" or "it's in your application". Not a lot of support there!
Telephone tech support only works at all if you're asking a well-defined question with a known answer. If you've got a well-defined question with a known answer, then that question and the answer should be in the documentation which should be on the Web because it will change with time and printed books are a pain to distribute.
On the other hand, if you are willing to use the World-Wide Web, you can go to google.com, type in a few keywords and go straight to some page with the answer. A lot of the time, that page will be the vendor's. As a matter of fact, I had a question about Windows NT yesterday and instead of trying to get help by calling someone on the phone, I did a Web search through google and it happened to point me to sites in microsoft.com. Ten minutes later, the question was answered and everything was working again. No hold time, and especially no waiting for "the next available operator"
The companies that used to provide traditional telephone tech support are abandoning it in favor of Web-based support. Microsoft has been shifting their tech support stuff to their "Knowledge Base" for years. Telephone tech support is expensive for the vendor and nearly useless for the customer. If Microsoft thinks providing Web-based tech support is better than telephone support, so why shouldn't it also be a win for free software? I can get my questions answered at postgresql.org just like at microsoft.com, can't I? Of course I can!
I'm sorry, I'm having cognitive dissonance over the idea that there are Windows developers who would be willing to develop applications in Notepad. Most Windows programmers I know cannot conceive of using anything other than a fully-blown IDE with all the bells and whistles. They seem to think that there is no way to develop anything at all using the style I use (Emacs or Vi, command-line compilers, and make.)
So, it doesn't seem to me to be FUD. Nobody familiar with Windows programming is going to even consider using such a "primitive" development "environment" as worthy of their time.
Some people write and freely-release software because they want to change the way the world thinks about intellectual property. That's what the Free Software Foundation is about. Some people write and freely-release software because it enhances the value of a product that they sell. That's why Livingston released their RADIUS daemon: so they could sell more Portmasters. Some software is funded by the public in the public interest and it's only fair for the public who paid for it to be able to use it.
However, I think the most people produce software simply for their own use. That is, we want to have the software to accomplish some task and there's no efficient way of getting it other than to write it ourselves, so we write it. Then, we give it away because there's no particular reason to keep it. If the primary value of a program is derived from the use you get out of it, and the potential commercial value of most software is small, then giving it away doesn't reduce the value of the software at all. Indeed, since others are encouraged to enhance the software and release their enhancements, the value of the software to you may actually increase if you share it.
I'm not out to change the world and I believe that intellectual property is as real as any other kind. I think that copyrights, trademarks, and even patents have boosted innovation and commerce for centuries so I see no reason to do away with them. Further, I have no products to sell on my own and my funding for these things comes out of my own pocket, but I was taught that it is polite to be neighborly and to share what you have with those around you. That is sufficient justification, in and of itself, for me to share the software that I write for myself. The fact that it has sometimes come back to me significantly enhanced is just icing on the cake.
I'm sorry, but I didn't see the part of the article where it said that "nothing MS works and everything crashes daily." The article did talk about unexpected downtime, but the main point was about the fact that the costs for the Microsoft-based solution were a lot higher than expected and the benefits were a lot lower than they were supposed to be. This seems to match my experience fairly well.
In fact, the most important thing about the article is the observation that Linux can be adopted piecemeal while Microsoft tends to want you to change all your software, and often much of your hardware, at once. In an economic downturn, the last thing you want to do is spend a bunch of money for the chance to take a leap of faith and shift your paradigm. Instead, more evolutionary tactics are called for, which just happens to be what Linux or *BSD is good for.
The use of Linux doesn't promise a radical improvement in the way you do business, but it also doesn't have a lot of the risk associated with a paradigm shift. Companies hedging their bets would do well to at least consider not buying Microsoft.
the guy sounds very clever, but when i see statistics like that i start wondering about what they're really measuring.
Umm, what difference does it make? If you're suffering more losses from frontal attacks, you need to beef up your frontal defense. It doesn't really matter why the losses are greater, the reaction is going to be the same.
Obsessed with Linux with good Reason!
on
IBM Wants Linux
·
· Score: 1
Don't you mean that the usage of Linux in some embedded systems is extreme bloat?
I used to think like you do, but the term "embedded system" covers a wide range of devices. I've programmed everything from wimpy-little 8-bit systems (not quite the 4-bit bottom end that some people still work on) to fairly large process control systems. They are all considered embedded systems because of the fact that they're special-purpose devices, but they have very different software requirements.
The little 8-bit devices were typically used for real-time control of local hardware and were programmed, often in assembly language, at the "bare metal". Big process control stuff often has significant database and display requirements so I tend to suggest running PostgreSQL and X11 under Linux. (I'm also involved in a system that makes extensive use of Windows NT Embedded, so it's not just something Linux fanatics do.)
What's interesting is that as the price on more powerful processors and memory drops, the embedded systems built from those parts get faster and more capable of running software built with techniques that weren't efficient enough before. One example is the fact that higher-level languages are now commonly used on new embedded designs. Ten years ago, that didn't happen all that often. The use of operating systems instead of programming to the "bare metal" is another example. The more advanced techniques allow greater programmer productivity, enhanced reliability, and lower maintenance costs and the reduced costs mean that it's economical enough to do on smaller and smaller systems.
With the wide use of so-called "embedded PC" devices, that trend has accelerated because Wintel PC's have to be both immensely powerful and quite inexpensive to compete with the market and the hardware manufacters take advantage of the technology developed for desktops to reduce the prices of their own products. The fact that many embedded systems only have to be "real time enough" and the fact that embedded systems tend to be very sensitive to marginal costs bodes well for the use of Linux for embedded systems into the long term.
You know, less than a decade ago Continental Airlines went through "Chapter 11", and their airplanes still take to the air every day. Yes, a bankruptcy is a dramatic step, but smart people who own a business will consider it long before the company is on its last legs.
It can also be good for the creditors, too. You can't get money from a company that's filed for Chapter 7, as that means the company has gone out of business, but generally the payments are rescheduled in a Chapter 11 bankruptcy so the creditors at least have a shot at getting all of what they're owed.
Of course, the folks at Loki probably would rather have not filed for protection, as it is a very public admission of being in over your head, but this doesn't mean that they're going to go under no matter what.
"we're all going to EISA"? Where did that come from? Use of the EISA was every bit as restricted as use of the MCA (that is, you had to pay the licensing fees to the owners) and got used in not a whole lot more computers. I believe that the fact that neither EISA nor MCA ever was very popular with PC users (IBM based a whole lot more on the MCA than the PS/2) was likely due to a combination of economic factors and technical factors.
Most people, even heavy computer users, have never seen an EISA based computer. I've never seen an EISA peripheral card. I have seen EISA computers, but then I run some of my software on certain models of Dell servers. It certainly wasn't a major selling point for most Dell customers which is no doubt why more recent models in the same line have dropped EISA and use PCI. The MCA was released around 1986 or 1987, and the EISA came out about then. The ISA dominated from before then until VLB and PCI came out and a few years after, in fact. PCI didn't start being popular until 1995 or so, after a brief fight with VLB, and has become dominant only in the last three years or so.
What really happened is that the PC buyers looked at machines based on MCA, EISA, and ISA, and decided that the benefits of the more powerful busses weren't worth the additional cost and stuck with ISA. I don't know if the perceived benefits were higher with VLB and PCI or if the costs were lower, but the situation was obviously different when they came on the scene. (I suspect that the costs were lower because Intel gave away the rights to use the PCI with the expectation that they would be in the best position to produce motherboards for PCI-based computers and so would make their money from hardware sales, but that's not based on any hard knowledge, just on my observation of what actually happened.)
I'd like to point out that you still haven't explained what I'm missing at the lower resolution. To me, HDTV vs analog is like BetaMax vs VHS. BetaMax had a noticeably better picture than VHS, but most people who bought VCR's valued picture quality less than things that VHS does well. Most people who watch television simply don't care very much about picture quality. They're not wrong any more than those who do care deeply about picture quality. However, it is wrong to forecast sales or attempt to build a mass market based upon the idea that the average American TV viewer is just waiting to leap at the chance to watch their Saturday-morning cartoons on a wider screen and at higher resolution.
With respect to the conversion to color television from black-and-white, it really was something that the "grass roots" end users wanted. The most certainly did not need to be "pushed" to accept it. When the only televisions were B&W, everyone expected that there would be an eventual conversion to color and that expectation drove the search for and the eventual widespread acceptance of a color television standard. That situation really isn't comparable to HDTV. HDTV isn't something that very many end users want. Instead, it was something that some electronics manufacturers sold to the US government as something necessary to keep the US up technically with the rest of the world. The end users didn't get any choice in the matter except the one they're taking: They simply don't buy the televisions or watch the programming.
Barring a radical change in display technology, I don't think the prices for HDTV's are going to decline quickly any time soon. While the technology to build CRT's for computer monitors may be similar to that needed to make CRT's for HDTV use, the tooling is different and the details are different (there are significantly more holes in the shadow mask and each one must be very precisely drilled) and the production lines are not optimized for the 16:9 aspect ratio CRT's.
That means that it is not necessarily just a matter of time before $500 HDTV displays are available, and the cost per unit probably has to get down to about half that before it'll truly be mass-market. You'll have to wait for the economies of scale to kick in and that means that there has to be substantial demand for the product, which simply isn't materializing, for the cost to come down to the point where it'll be competitive with analog TV's.
Go ahead and be confident, but if the people who are making budgeting decisions for their households don't view HDTV as "progress" then they won't spend the money to buy the equipment and what I have predicted will actually happen. I guess we'll see in a decade or so. I can wait.
There is a shortage of stores that actually show HDTV samples on their HDTV sets, but if there's one near you go take a look. The differences are apparent, especially on standard CRT TVs (most of the rear projection, big screen models look like CRAP IMO).
I think you miss his point. I've seen HDTV up close and in person, and I won't argue that the differences aren't apparent, but I will argue that my TV-watching experience isn't significantly enhanced by the extra resolution. I mean, how many of Jay Leno's jokes am I going to not get because I'm still watching him on an analog TV? I can tell you that my enjoyment of "Mission Impossible 2" wasn't significantly reduced because I watched the opening on HDTV and the end on regular analog.
The point is that most people think that their televisions work well enough and don't see any crying need to make any significant changes. This is not the same as color television because it was generally acknowledged by the great unwashed masses that color was enough better than black and white to justify paying a hefty (at one point, on the order of 3-5 times the cost of the B&W TV) premium.
Anyway, since there is no general acknowledgement, especially among the "aluminum foil-enhanced rabbit-ears" crowd that makes up the bulk of television audiences, that the improved display significantly enhances what people watch television for, it is not clear to me that the "progression is inevitable" or that the economies of scale will ever kick in and drive the costs of HDTV lower before HDTV itself is abandoned and the equipment can be bought on closeout for ten cents or so on the dollar.
Yes, it looks better (when it works at all, and HDTV is significantly less robust than analog television) but it's a quantitative difference rather than qualitative. That's why HDTV is such a hard sell and why it's adoption is way slower than most HDTV advocates expected.
Yes, but it certainly can be profitable to market, distribute, and sell those devices needed to take advantage of those free somethings. In fact, the worldwide demand for solar cells has kept a number of companies in business for quite some time now and the demand is increasing. That demand has paid for research into improved solar cells which has made possible things like calculators that can be powered from the ambient light in a well-lit room. The efficiency of the improved cells opens new markets, like those solar-powered school zone lights I keep seeing, which is what is fuelling the increased demand.
However, there are some things about photovoltaics that you have to understand. First, they're not zero-maintenance. For example, if a solar panel gets covered by dust or other debris, it has to be cleaned. Second, they don't have an infinite lifespan. When I last paid attention to this sort of thing, the figures I heard was that any given solar panel was expected to last 5 years, but the efficiency of any photovoltaic cell decreases over time. When you add to those two problems the fact that there is absolutely no way to efficiently store the energy that the cell produces, you find that there's a reason that almost nobody uses it if other sources of power are available. The total installed cost of the system has to be less than the cost of the electricity that would otherwise have to be paid for over the life of that system or it makes no sense to use solar power.
It's also interesting to note that the early demand for solar power equipment came, in large part, from those mean old nasty oilmen who, according to some people, are the only thing standing between the world and clean, efficient power. If it weren't for the solar cells used to power offshore oil production platforms and similar on-shore equipment in the often undeveloped places where petroleum is found, the modern high-efficiency amorphous silicon photovoltaics likely wouldn't exist.
The problem that "alternative" energy forms have is that they are difficult to store or it is difficult to convert their energy into something useful. Also, every form of energy production has some sort of negative environmental impact whether it be lakes whose level goes up and down with the energy demand, forests of windmills, or acres of shade where there was none before.
Actually, I've run sid on a number of computers for several months now and I've been quite happy with it. Other than occasional glitches, like the time the Perl update basically removed the working perl from the system, the time the lilo package maintainer decided that always starting the lilo configuration over from scratch was a good idea, and the ssh that wasn't synced with the encryption libraries, everything has worked reasonably well. The machines in question aren't servers, but are workstations in daily use. I suppose you might consider that "production"
The primary reason to use sid is to get the up-to-date software. Among other things, I got 3d accelerated video with DRI well before it was available in stable. The primary burden with using sid (other than the obvious---things break sometimes) is the fact that you have to keep updating it. Of course, even if you're running stable you should add the security patch source to your apt-get and upgrade your system periodically, but with stable you don't see enhancements coming at the fast and furious pace of sid.
On the other hand, it's difficult to distribute a meaningful distribution from a collection that's continuously changing without occasional freezes. If you're not willing to keep up with the bleeding edge, you want a single version number to work from to let you know where you are in relation to what's available. Having a collection revision makes it easier to know where you stand. That's why most of my systems are running stable. Also, I prefer to install from CD-ROM, which is another thing that is difficult to do without freezes.
Can a linux box read a PC floppy or HD? How about one for Mac? Can a redhat box access files from a Mandrake one?
What you are asking is if Linux is compatable with all the partitioning methods and file systems that those kinds of computers use. The answer is "yes". Of course, the PC is well supported because even early Linux systems had to coexist with and use the filesystems for the other PC operating systems installed on the same hardware, but the Macintosh is well supported as well. In fact, recently my office received a Mac ZIP disk containing works in progress from our advertising agency and of all the computers in the office, only those running Linux could read it.
Whether or not a kernel from one distribution (Mandrake) can read the file system used by another (RedHat) depends upon the options used to build the kernel. Some older kernels won't be able to read some file systems. However, most all kernels distributed understand the second extended file system (or ext2fs) as it's been around for a number of years now and is widely used.
BassGuy23 also asked:
Can I get a linux box to access the internet through the Windows network?
On the Internet, there is no such thing as a "Windows network". Everybody uses the same TCP/IP protocols, including Windows computers. Unless the other Windows computers on the network are running special software as well, you shouldn't have a problem with a Linux computer.
If I were starting an ASP (and I've got a business plan to do just that around here somewhere) I would accept liability for lost data. That liability would be limited, certainly, but giving performance-level guarantees is about the only way that you're going to get people to trust the service.
I would also specify the means by which people would get their data back in the event of the business going under or being sold. Again, this is essential part of the ASP business because what an ASP is really selling is security and selling security requires that people trust you.
On the contrary. In order to design a stealthy shape, one mostly just avoids shapes that are particularly unstealthy. The rule to accomplish that is simple: No concave right angles. Since, for aerodynamic reasons (the term "interference drag" is significant) right angles are rarely sought out by aircraft designers, the bulk of what makes an F-117 or B2 more stealthy than your typical aircraft is the radar-absorptive coating. However, it is quite difficult to create a radar-absorptive coating that works over a broad range of frequencies.
One example of a radar-absorptive coating is a layer of conductive paint a quarter wavelength above the metal surface of the aircraft. This is very simple to do, doesn't require any particular care, and doesn't require any exotic materials. If, however, the radar is at twice the frequency that the coating is designed to absorb, that particular coating enhances the return rather than attenuates it.
Please note that even an absolutely stealthy aircraft can't necessarily escape detection. There are techniques for using backscatter from the atmosphere (developed primarily for wind shear advisories around airports) that can detect an aircraft's passage from the air that is disturbed around it. Even flying slow (which defeats a doppler-based system for low-observability aircraft) won't defeat that because it doesn't look at the aircraft, but at the air and an airplane is going to move a lot more air than a bird and is going to move it a lot faster than the bird would.
I've actually set up reverse-DNS for my downstream customers such that they can configure it themselves. The primary issue is that the IPv4 reverse-DNS system is oriented around naming classful addresses. The minimum you can delegate is 256 addresses and nobody (well, almost nobody) gives out 256 addresses any more. So, you've got to use this crufty (but standard!--it's in the RFC's) hack to delegate parts of each address block.
A second problem is that DNS servers can be a major hassle and a misconfigured DNS server can cause things to stop working. An admin tends to not be real comfortable delegating domains to the typical customer because the typical customer hasn't proven he knows what he's doing. We often run both the forward and reverse DNS for our fixed-IP customers for this very reason.
On the other hand, as someone else pointed out, it should be enough for tcpd if there's a forward DNS entry that matches each reverse-DNS entry no matter what other DNS entries also map to that address. If each address has a default name and each address maps to that default name, then everything should work even if other names map to any given address. I consider that sort of thing to be basic to good network design.
When I think of Microsoft's failures, the one that always leaps to mind, jumping up and down waving its arms for attention is QuickPascal. Back in the late 80's, Microsoft was tired of Borland selling lots more copies of compilers than they were and so decided to compete head-to-head with Turbo Pascal and release QuickPascal, to go along with their QuickBASIC and QuickC offerings.
The fact that none of you have ever heard of QuickPascal shows how well that plan worked. Borland ran into their own problems, but they had more to do with the mistakes made by the Borland people than anything Microsoft did and Turbo Pascal is still around, it's just been renamed "Delphi".
Sometimes, Microsoft persists with a poor seller and makes it into a market power. With QuickPascal, they decided to go in a different direction (to wit, emphasize Windows development, where they have a natural advantage) and to bury the loser deep.
The moral of the story: Microsoft is not a magical company. They make mistakes.
Let's talk about point 3 for a moment. Most of what Microsoft says about TCO is nonsense. Yes, the cost of the software is negligible if you're trying to make use of TB-sized databases. However, most people aren't running TB-sized databases. For the sort of small-end stuff (up to 1 GB or so) that most companies need in order to continue in business, the cost of the system is dominated by the cost of the software, if the software used is commercial.
The supposed difference in the amount you have to pay a Linux administrator as opposed to having one of your nontechnical employees manage your W2K server part-time is buried by the fact that it takes a real admin to keep W2K running properly. Just like any other server system, W2K is reliable and secure only if properly administered. A real admin costs about the same whether he knows W2K or Linux or BSD or whatever.
If the cost of the system is dominated by the software (which it will be, in the typical case) and the cost of operating the system is dominated by administration (which it is) and administrating the system costs the same whether you run Linux or W2K, then just how does Microsoft win on TCO?
Finally, the test results, as given, say absolutely nothing about the performance of Linux on the same hardware as W2K was using. Neither do they say anything about the performance of W2K on the hardware that Linux was using. It is entirely possible that W2K would lose on the four-computer cluster just as it is possible that W2K would still win if Linux was running on the 8-way SMP box. We simply have no information either way.
And what if SQL Server under W2K pulls in a few (10, say) percent more performance than Linux on the same hardware? What difference does it make? I'll admit that I'm a free software (I run Linux, but the *BSD projects have some really cool stuff going on, too) advocate and I work really hard to get Linux accepted by the guys in suits, and I am encouraged by these results.
Why? Well among people who know what they're talking about, it is generally acknowledged that benchmarks are mostly bogus, and this benchmark especially so because it's more for the hardware manufacturers to strut their stuff than to show off the performance of the software, but it's tougher to convince someone with the argument that a benchmark isn't a "real world" test if the system you prefer doesn't even appear near the top of the benchmark results than if it does. Being at the top is nice, but being in the top 10 is almost as good. Linux doesn't have to be the best, it just has to be good enough.
With all due respect to that person of questionable intelligence who posted that rant about asking Slashdot about this as opposed to doing "research", well, I am the "pop" half of a "mom-and-pop" ISP in Houston, so I can probably give some sort of useful advice. In fact, in my opinion, and I've been doing this for most of the last decade, this sort of question asked in this venue is likely to produce more useful information than going to some library and trying to find a book that describes how to run an ISP. I really pity someone who tries to figure out how to start an ISP by reading back issues of "Boardwatch".
The first thing that I recommend to people who want to start their own ISP is professional psychiatric help. (In fact, I have the names of several very good psychiatrists and even a psychologist or two in and around Houston, TX, US.) If that doesn't convince them that it's more fun to go broke on a trip to Vegas than to accomplish the same task by starting an Internet business, then I can get down to brass tacks.
So, the advice (some of it contradictory) observations, and opinions in no particular order:
The first thing you need to know about the ISP business is that it is not primarily a "technology" business. What I mean by that is that there is essentially no technical risk. What I mean by that is that you can start an ISP with equipment and software purchased off-the-shelf. That fact is why the margins are so low in the ISP business, and why it's so difficult to stay in business.
I don't know what it costs in Australia, but it takes most people somewhere between 50,000 USD and 100,000 USD to start an ISP around here. The more technical expertise you have, the less you'll need to spend, but expect to spend at least as much on advertising as you do on ongoing service.
Unfortunately, people who have money to invest typically want to invest vastly more than that in hopes of getting vastly more return. Around here, this means that you aren't likely to be able to get any venture capital unless you can figure out how to write a business plan that calls for spending maybe 20,000,000 USD per year and breaks even in 48 months. (Lately, I've been toying with the idea of putting together such a plan, getting an investor's money, then continuing to operate on the cheap. I could then break even in 48 hours.)
If you do go for VC funding, try to talk only to people with money to invest. (Most of the people I've managed to talk to over the years have nothing but some fantasy about brokering a deal with some VC person for a share of the money. This works about as well as jet-propelled pigs.)
Customer retention is the key to long-term success.
The biggest barrier to customer retention is persistent connection difficulties. Most of the connection problems your customers will have will be due to the telephone company. However, your customers will be mad at you about it and demand that you fix it despite the fact that you can do nothing to help or hinder the process.
If you can't balance a checkbook, hire someone who can and then watch them to make sure they don't have sticky fingers.
If you can't set up a router, RADIUS server, and access equipment, hire someone who can and pay them a salary. Make sure their bonuses are related to uptime rather than "face" time. ("I don't care if the sysadmin's not here as long as the network is running," should be your motto.)
Everyone at the ISP should resign themselves to the fact that, at a startup ISP, everybody does sales and everybody does tech support.
Put together service packages and every time someone wants you to bid on a special project, either make it out of those service packages (after the fashion of a "Chinese menu") or don't bid. Putting together bids is a major time sink if you let it become one.
Try to not lose money on anything you sell.
Billing systems suck. Some of them suck in different ways and some of them suck expensively, but they all suck.
Talk to a lawyer and an accountant about the form that the business should take. Limit your liability as much as you can. You won't be able to get out of all of it, because some of the creditors will insist on personal guarantees of payment, but get out of as much as you can.
Explore and develop ancillary sources of revenue. I know an ISP here in Texas that does for-pay computer training three or four times a year.
Learn how to work the telephone system to get what you want. I don't know what that means in context of an Australian ISP, because it's different from working the SBC system, but you'll need to do it.
Use access concentrators and digital phone lines wherever possible. I've used Ascend (now Lucent) Max equipment and they work. Others have used Cisco equipment with similar results.
It may be possible to find someone who will lease you access to their modem banks. If you can do that, it might be worth your while. However, it adds an additional layer for a customer's problem report to go through, so it may not be worth the hassle.
Keep backups of all customer data.
If a vendor neglects to bill you, put the money you would be paying them into an interest-bearing account and leave it there until they notice that you haven't paid. Just because you didn't get the bill doesn't mean that you don't have to pay the bill.
Try to make yourself superfluous as quickly as possible. Essential people don't get vacations.
The most necessary tech support training isn't technical. The most important skill a telephone tech support person has is the ability to control a call. The user should be responding to you, not the other way around.
They don't need to know any more than they do, since all they want is a glorified typewriter and calculator. Those people will use whatever OS is put in front of them. For them, Linux will be very easy to use, because someone else will have done all the configuration and installation.
I should point out that at the company I own, as opposed to the one I work for, we actually deployed Linux on the secretary's desktop and it did, indeed, work exactly this way. I set it up and managed it, and she used it and she had no problems at all, except she never did figure out how to run the games that were installed. (I don't know what was so hard about "stepping on the foot" and selecting "Games", but she's never managed it.)
Slashdot Mirror
In fact, my first inclination is to try to use the Beowulf stuff rather than Pooch simply because such a detailed work exists and is available for Beowulf clusters, but I don't know if any such information exists for Pooch.
For example, by the time Friday was released, in 1984, as I recall, publically available computer terminals were in existence, BBSes were how you got on-line (except for the fortunate few how knew about and had access to Usenet) and networked BBSes were about to be invented.
My own personal favorite example of an SF prediction is in Bellamy's Looking Backward which, among other things, talked about how the broadcasting of music (live performances over telephone lines as neither audio recording nor radio had been invented or conceived of when the book was written) had become common. I also seem to recall that it had some bit in there about how that led to fewer people being able to play the piano, but that may be my subsequent experiences leaking over as it's been 20+ years since I read that book.
However, it seems to me that the question is not about predictions in SF that come true, but about how SF has driven invention. If, as I say I believe above, SF is a reflection of the culture it's written in, then there can be no direct link. However, I also believe that invention is also a product of the culture it is in, so it is certainly fair to say that, if a work doesn't have a direct effect on invention, then it will necessarily reflect the environment in which the invention is made. Rarely is this made more clear than in "The Man Who Sold the Moon" where Delos D. Harriman talks about what it was like to grow up in the early part of the 20th century.
Further, if one wishes to look at that aspect more closes, I think that one could do worse than looking at the work of Dr. Lienhard of the University of Houston (not his son, who is a professor at MIT) who has a 5-minute daily radio program (and book derived therein) called "The Engines of our Ingenuity" which discusses the whole process of invention and covers quite well the methods by which people derive inspiration. The URL to reach the radio show's transcripts is http://www.uh.edu/engines
For the cases you describe, it sounds as if the answer isn't known. Too bad. In that case, you can either figure out the problem is yourself or suffer through it, which, at 03:17, are basically the same choices you have with commercial software and expensive support agreements. If they've got to call the developers in to write a patch, you're screwed anyway.
Of course, documentation that doesn't suck would be nice. It'd be nice if one got documentation that doesn't suck from Microsoft, Oracle, and Sun, too. However, you have to take what you can get because, well, it's what you get. Oh, and personally, I think there's quite a bit of freeware documentation that doesn't suck. Not nearly all of it, by any stretch of the imagination, but quite a bit.
Look, say the database goes down at 3:17 AM. You're running Microsoft SQL server with one of those premium contracts, (this is a mission-critical application after all and time is, therefore, a lot of money,) so you call Microsoft's tech support line. You don't get to talk to a human, of course. Instead, you get to listen to "Microsoft Radio" for four hours while the only awake human in the place finishes up some hot chat. Once you do get a support person on the line, you can ask your question and hear the known answer or you can ask your question and get "we don't know" or "it's in the third-party package" or "it's in your application". Not a lot of support there!
Telephone tech support only works at all if you're asking a well-defined question with a known answer. If you've got a well-defined question with a known answer, then that question and the answer should be in the documentation which should be on the Web because it will change with time and printed books are a pain to distribute.
On the other hand, if you are willing to use the World-Wide Web, you can go to google.com, type in a few keywords and go straight to some page with the answer. A lot of the time, that page will be the vendor's. As a matter of fact, I had a question about Windows NT yesterday and instead of trying to get help by calling someone on the phone, I did a Web search through google and it happened to point me to sites in microsoft.com. Ten minutes later, the question was answered and everything was working again. No hold time, and especially no waiting for "the next available operator"
The companies that used to provide traditional telephone tech support are abandoning it in favor of Web-based support. Microsoft has been shifting their tech support stuff to their "Knowledge Base" for years. Telephone tech support is expensive for the vendor and nearly useless for the customer. If Microsoft thinks providing Web-based tech support is better than telephone support, so why shouldn't it also be a win for free software? I can get my questions answered at postgresql.org just like at microsoft.com, can't I? Of course I can!
So, it doesn't seem to me to be FUD. Nobody familiar with Windows programming is going to even consider using such a "primitive" development "environment" as worthy of their time.
If you're interested in DOS and Windows development, you can also try the free (as in beer) Digital Mars compiler.
However, I think the most people produce software simply for their own use. That is, we want to have the software to accomplish some task and there's no efficient way of getting it other than to write it ourselves, so we write it. Then, we give it away because there's no particular reason to keep it. If the primary value of a program is derived from the use you get out of it, and the potential commercial value of most software is small, then giving it away doesn't reduce the value of the software at all. Indeed, since others are encouraged to enhance the software and release their enhancements, the value of the software to you may actually increase if you share it.
I'm not out to change the world and I believe that intellectual property is as real as any other kind. I think that copyrights, trademarks, and even patents have boosted innovation and commerce for centuries so I see no reason to do away with them. Further, I have no products to sell on my own and my funding for these things comes out of my own pocket, but I was taught that it is polite to be neighborly and to share what you have with those around you. That is sufficient justification, in and of itself, for me to share the software that I write for myself. The fact that it has sometimes come back to me significantly enhanced is just icing on the cake.
In fact, the most important thing about the article is the observation that Linux can be adopted piecemeal while Microsoft tends to want you to change all your software, and often much of your hardware, at once. In an economic downturn, the last thing you want to do is spend a bunch of money for the chance to take a leap of faith and shift your paradigm. Instead, more evolutionary tactics are called for, which just happens to be what Linux or *BSD is good for.
The use of Linux doesn't promise a radical improvement in the way you do business, but it also doesn't have a lot of the risk associated with a paradigm shift. Companies hedging their bets would do well to at least consider not buying Microsoft.
Umm, what difference does it make? If you're suffering more losses from frontal attacks, you need to beef up your frontal defense. It doesn't really matter why the losses are greater, the reaction is going to be the same.
Don't you mean that the usage of Linux in some embedded systems is extreme bloat?
I used to think like you do, but the term "embedded system" covers a wide range of devices. I've programmed everything from wimpy-little 8-bit systems (not quite the 4-bit bottom end that some people still work on) to fairly large process control systems. They are all considered embedded systems because of the fact that they're special-purpose devices, but they have very different software requirements.
The little 8-bit devices were typically used for real-time control of local hardware and were programmed, often in assembly language, at the "bare metal". Big process control stuff often has significant database and display requirements so I tend to suggest running PostgreSQL and X11 under Linux. (I'm also involved in a system that makes extensive use of Windows NT Embedded, so it's not just something Linux fanatics do.)
What's interesting is that as the price on more powerful processors and memory drops, the embedded systems built from those parts get faster and more capable of running software built with techniques that weren't efficient enough before. One example is the fact that higher-level languages are now commonly used on new embedded designs. Ten years ago, that didn't happen all that often. The use of operating systems instead of programming to the "bare metal" is another example. The more advanced techniques allow greater programmer productivity, enhanced reliability, and lower maintenance costs and the reduced costs mean that it's economical enough to do on smaller and smaller systems.
With the wide use of so-called "embedded PC" devices, that trend has accelerated because Wintel PC's have to be both immensely powerful and quite inexpensive to compete with the market and the hardware manufacters take advantage of the technology developed for desktops to reduce the prices of their own products. The fact that many embedded systems only have to be "real time enough" and the fact that embedded systems tend to be very sensitive to marginal costs bodes well for the use of Linux for embedded systems into the long term.
It can also be good for the creditors, too. You can't get money from a company that's filed for Chapter 7, as that means the company has gone out of business, but generally the payments are rescheduled in a Chapter 11 bankruptcy so the creditors at least have a shot at getting all of what they're owed.
Of course, the folks at Loki probably would rather have not filed for protection, as it is a very public admission of being in over your head, but this doesn't mean that they're going to go under no matter what.
Most people, even heavy computer users, have never seen an EISA based computer. I've never seen an EISA peripheral card. I have seen EISA computers, but then I run some of my software on certain models of Dell servers. It certainly wasn't a major selling point for most Dell customers which is no doubt why more recent models in the same line have dropped EISA and use PCI. The MCA was released around 1986 or 1987, and the EISA came out about then. The ISA dominated from before then until VLB and PCI came out and a few years after, in fact. PCI didn't start being popular until 1995 or so, after a brief fight with VLB, and has become dominant only in the last three years or so.
What really happened is that the PC buyers looked at machines based on MCA, EISA, and ISA, and decided that the benefits of the more powerful busses weren't worth the additional cost and stuck with ISA. I don't know if the perceived benefits were higher with VLB and PCI or if the costs were lower, but the situation was obviously different when they came on the scene. (I suspect that the costs were lower because Intel gave away the rights to use the PCI with the expectation that they would be in the best position to produce motherboards for PCI-based computers and so would make their money from hardware sales, but that's not based on any hard knowledge, just on my observation of what actually happened.)
I, too, have spent many hours playing VAX Empire. For addictive, though, it's tough to beat BSD Empire"
With respect to the conversion to color television from black-and-white, it really was something that the "grass roots" end users wanted. The most certainly did not need to be "pushed" to accept it. When the only televisions were B&W, everyone expected that there would be an eventual conversion to color and that expectation drove the search for and the eventual widespread acceptance of a color television standard. That situation really isn't comparable to HDTV. HDTV isn't something that very many end users want. Instead, it was something that some electronics manufacturers sold to the US government as something necessary to keep the US up technically with the rest of the world. The end users didn't get any choice in the matter except the one they're taking: They simply don't buy the televisions or watch the programming.
Barring a radical change in display technology, I don't think the prices for HDTV's are going to decline quickly any time soon. While the technology to build CRT's for computer monitors may be similar to that needed to make CRT's for HDTV use, the tooling is different and the details are different (there are significantly more holes in the shadow mask and each one must be very precisely drilled) and the production lines are not optimized for the 16:9 aspect ratio CRT's.
That means that it is not necessarily just a matter of time before $500 HDTV displays are available, and the cost per unit probably has to get down to about half that before it'll truly be mass-market. You'll have to wait for the economies of scale to kick in and that means that there has to be substantial demand for the product, which simply isn't materializing, for the cost to come down to the point where it'll be competitive with analog TV's.
Go ahead and be confident, but if the people who are making budgeting decisions for their households don't view HDTV as "progress" then they won't spend the money to buy the equipment and what I have predicted will actually happen. I guess we'll see in a decade or so. I can wait.
I think you miss his point. I've seen HDTV up close and in person, and I won't argue that the differences aren't apparent, but I will argue that my TV-watching experience isn't significantly enhanced by the extra resolution. I mean, how many of Jay Leno's jokes am I going to not get because I'm still watching him on an analog TV? I can tell you that my enjoyment of "Mission Impossible 2" wasn't significantly reduced because I watched the opening on HDTV and the end on regular analog.
The point is that most people think that their televisions work well enough and don't see any crying need to make any significant changes. This is not the same as color television because it was generally acknowledged by the great unwashed masses that color was enough better than black and white to justify paying a hefty (at one point, on the order of 3-5 times the cost of the B&W TV) premium.
Anyway, since there is no general acknowledgement, especially among the "aluminum foil-enhanced rabbit-ears" crowd that makes up the bulk of television audiences, that the improved display significantly enhances what people watch television for, it is not clear to me that the "progression is inevitable" or that the economies of scale will ever kick in and drive the costs of HDTV lower before HDTV itself is abandoned and the equipment can be bought on closeout for ten cents or so on the dollar.
Yes, it looks better (when it works at all, and HDTV is significantly less robust than analog television) but it's a quantitative difference rather than qualitative. That's why HDTV is such a hard sell and why it's adoption is way slower than most HDTV advocates expected.
However, there are some things about photovoltaics that you have to understand. First, they're not zero-maintenance. For example, if a solar panel gets covered by dust or other debris, it has to be cleaned. Second, they don't have an infinite lifespan. When I last paid attention to this sort of thing, the figures I heard was that any given solar panel was expected to last 5 years, but the efficiency of any photovoltaic cell decreases over time. When you add to those two problems the fact that there is absolutely no way to efficiently store the energy that the cell produces, you find that there's a reason that almost nobody uses it if other sources of power are available. The total installed cost of the system has to be less than the cost of the electricity that would otherwise have to be paid for over the life of that system or it makes no sense to use solar power.
It's also interesting to note that the early demand for solar power equipment came, in large part, from those mean old nasty oilmen who, according to some people, are the only thing standing between the world and clean, efficient power. If it weren't for the solar cells used to power offshore oil production platforms and similar on-shore equipment in the often undeveloped places where petroleum is found, the modern high-efficiency amorphous silicon photovoltaics likely wouldn't exist.
The problem that "alternative" energy forms have is that they are difficult to store or it is difficult to convert their energy into something useful. Also, every form of energy production has some sort of negative environmental impact whether it be lakes whose level goes up and down with the energy demand, forests of windmills, or acres of shade where there was none before.
The primary reason to use sid is to get the up-to-date software. Among other things, I got 3d accelerated video with DRI well before it was available in stable. The primary burden with using sid (other than the obvious---things break sometimes) is the fact that you have to keep updating it. Of course, even if you're running stable you should add the security patch source to your apt-get and upgrade your system periodically, but with stable you don't see enhancements coming at the fast and furious pace of sid.
On the other hand, it's difficult to distribute a meaningful distribution from a collection that's continuously changing without occasional freezes. If you're not willing to keep up with the bleeding edge, you want a single version number to work from to let you know where you are in relation to what's available. Having a collection revision makes it easier to know where you stand. That's why most of my systems are running stable. Also, I prefer to install from CD-ROM, which is another thing that is difficult to do without freezes.
What you are asking is if Linux is compatable with all the partitioning methods and file systems that those kinds of computers use. The answer is "yes". Of course, the PC is well supported because even early Linux systems had to coexist with and use the filesystems for the other PC operating systems installed on the same hardware, but the Macintosh is well supported as well. In fact, recently my office received a Mac ZIP disk containing works in progress from our advertising agency and of all the computers in the office, only those running Linux could read it.
Whether or not a kernel from one distribution (Mandrake) can read the file system used by another (RedHat) depends upon the options used to build the kernel. Some older kernels won't be able to read some file systems. However, most all kernels distributed understand the second extended file system (or ext2fs) as it's been around for a number of years now and is widely used.
BassGuy23 also asked:
On the Internet, there is no such thing as a "Windows network". Everybody uses the same TCP/IP protocols, including Windows computers. Unless the other Windows computers on the network are running special software as well, you shouldn't have a problem with a Linux computer.
I would also specify the means by which people would get their data back in the event of the business going under or being sold. Again, this is essential part of the ASP business because what an ASP is really selling is security and selling security requires that people trust you.
One example of a radar-absorptive coating is a layer of conductive paint a quarter wavelength above the metal surface of the aircraft. This is very simple to do, doesn't require any particular care, and doesn't require any exotic materials. If, however, the radar is at twice the frequency that the coating is designed to absorb, that particular coating enhances the return rather than attenuates it.
Please note that even an absolutely stealthy aircraft can't necessarily escape detection. There are techniques for using backscatter from the atmosphere (developed primarily for wind shear advisories around airports) that can detect an aircraft's passage from the air that is disturbed around it. Even flying slow (which defeats a doppler-based system for low-observability aircraft) won't defeat that because it doesn't look at the aircraft, but at the air and an airplane is going to move a lot more air than a bird and is going to move it a lot faster than the bird would.
A second problem is that DNS servers can be a major hassle and a misconfigured DNS server can cause things to stop working. An admin tends to not be real comfortable delegating domains to the typical customer because the typical customer hasn't proven he knows what he's doing. We often run both the forward and reverse DNS for our fixed-IP customers for this very reason.
On the other hand, as someone else pointed out, it should be enough for tcpd if there's a forward DNS entry that matches each reverse-DNS entry no matter what other DNS entries also map to that address. If each address has a default name and each address maps to that default name, then everything should work even if other names map to any given address. I consider that sort of thing to be basic to good network design.
The fact that none of you have ever heard of QuickPascal shows how well that plan worked. Borland ran into their own problems, but they had more to do with the mistakes made by the Borland people than anything Microsoft did and Turbo Pascal is still around, it's just been renamed "Delphi".
Sometimes, Microsoft persists with a poor seller and makes it into a market power. With QuickPascal, they decided to go in a different direction (to wit, emphasize Windows development, where they have a natural advantage) and to bury the loser deep.
The moral of the story: Microsoft is not a magical company. They make mistakes.
The supposed difference in the amount you have to pay a Linux administrator as opposed to having one of your nontechnical employees manage your W2K server part-time is buried by the fact that it takes a real admin to keep W2K running properly. Just like any other server system, W2K is reliable and secure only if properly administered. A real admin costs about the same whether he knows W2K or Linux or BSD or whatever.
If the cost of the system is dominated by the software (which it will be, in the typical case) and the cost of operating the system is dominated by administration (which it is) and administrating the system costs the same whether you run Linux or W2K, then just how does Microsoft win on TCO?
Finally, the test results, as given, say absolutely nothing about the performance of Linux on the same hardware as W2K was using. Neither do they say anything about the performance of W2K on the hardware that Linux was using. It is entirely possible that W2K would lose on the four-computer cluster just as it is possible that W2K would still win if Linux was running on the 8-way SMP box. We simply have no information either way.
And what if SQL Server under W2K pulls in a few (10, say) percent more performance than Linux on the same hardware? What difference does it make? I'll admit that I'm a free software (I run Linux, but the *BSD projects have some really cool stuff going on, too) advocate and I work really hard to get Linux accepted by the guys in suits, and I am encouraged by these results.
Why? Well among people who know what they're talking about, it is generally acknowledged that benchmarks are mostly bogus, and this benchmark especially so because it's more for the hardware manufacturers to strut their stuff than to show off the performance of the software, but it's tougher to convince someone with the argument that a benchmark isn't a "real world" test if the system you prefer doesn't even appear near the top of the benchmark results than if it does. Being at the top is nice, but being in the top 10 is almost as good. Linux doesn't have to be the best, it just has to be good enough.
With all due respect to that person of questionable intelligence who posted that rant about asking Slashdot about this as opposed to doing "research", well, I am the "pop" half of a "mom-and-pop" ISP in Houston, so I can probably give some sort of useful advice. In fact, in my opinion, and I've been doing this for most of the last decade, this sort of question asked in this venue is likely to produce more useful information than going to some library and trying to find a book that describes how to run an ISP. I really pity someone who tries to figure out how to start an ISP by reading back issues of "Boardwatch".
The first thing that I recommend to people who want to start their own ISP is professional psychiatric help. (In fact, I have the names of several very good psychiatrists and even a psychologist or two in and around Houston, TX, US.) If that doesn't convince them that it's more fun to go broke on a trip to Vegas than to accomplish the same task by starting an Internet business, then I can get down to brass tacks.
So, the advice (some of it contradictory) observations, and opinions in no particular order:
I'm sure there's more, but that's enough for now.
I should point out that at the company I own, as opposed to the one I work for, we actually deployed Linux on the secretary's desktop and it did, indeed, work exactly this way. I set it up and managed it, and she used it and she had no problems at all, except she never did figure out how to run the games that were installed. (I don't know what was so hard about "stepping on the foot" and selecting "Games", but she's never managed it.)