The photo-quality printers will still be expensive, and the all-in-one printer/scanner/copier/fax combos, but if I was able to follow the not-very-specific PR well enough, the low-end office-quality printers will be about $49, with ~$10 print cartridges, so at least for printing mostly text and maps, at 1200x1200 resolutions, it'll be a lot cheaper. (The resolution is "optimized" to 4800x1200-equivalent, and it's hard to tell whether the second 1200 is really 1200, or is "1200-optimized" and really some lower resolution like 600, but for just about anything other than photos, it's still really enough.)
Sure, they'll have to junk the things they've said in the past, and say things that are almost the total opposite, but they can still chalk it up to improvements in technology, something like "We've improved print-head lifetimes to the point that we can package them separately, unlike the previous generation which got really high quality printing by being so thin they wore out rapidly" or whatever.
If they can integrate the printhead into the printer, rather than making it a separate replaceable component, at least you'll have to buy a new printer body when the head wears out - it's not quite as good as buying a new one when the ink runs out, but it's a good start.
OK, it was a laser printer, not an inkjet, but it did use liquid toner, a nasty mixture of carbon black and kerosine (or something pretty much like that) - you didn't want to spill it, and the machine did clog up a lot and occasionally overflow:-) The print engine was derived from a Canon photocopier engine of the same vintage, and got 240x240 pixels/inch. Unlike modern dry-toner laser printers, the ink smeared like an inkjet - I once had somebody ask me to send them an original of a document I'd printed on it, not a photocopy (Sorry, that *is* the original:-). The followon product was a 300x300dpi dry-process printer that was much much nicer.
The technology's a bit older, according to this very nice article on Stanford spinoffs. Canon loaned Stanford a small laser printer in 1979, and Stanford folks integrated a microprocessor with it and then spun off Imagen to commericialize it.
They're being offered free colo and bandwidth. If they were to rent that for $300/month, they'd have spent $3000 in 10 months (or 15 months, if they're paying $200/month.) That's not very long - it's much better financially to buy the hardware and take the free monthly. Also, it's probably much easier to raise funds for a one-time item like this than to beg for rent money every month.
Sorry. Eyeball recognition is right out. Besides, there are lots of problems with using it in practice that aren't that easy - databases still are never 100% correct, and once the Feds have put somebody with a similar name to yours on the Don't Fly List without a trial, and the airlines have put your eyeball print on the computer with your name and refused to put you on a flight because your name is similar to the guy on the list, your terrorist status is just going to propagate to anything that uses your eyeball print.
Since the report is closed-source, you can't see the real numbers, but suppose they're approximately correct. Some other articles were concerned about "5 years", but that's actually a reasonable timeframe for a technology project that's production-mode and not just a pilot - you want to amortize the costs over a reasonable period of time, even though most of the cash is upfront during the building phase, and in 3-5 years, either the project will miserable failure, or it'll be a raving success everybody loves and wants continued, but the equipment is by then antique and needs replacing.
The question is what are their assumptions about "what's a user" and "what's feasible"? I've seen several models for community wireless, and they've got much different definitions
Free service that makes the community more friendly, so that the citizens are happier and businesses make more money, similar to the ways streetlights and socialized baseball stadiums do. In this model, success means "everybody loves it, tourists find it easier to get around, people drink more coffee, eat more restaurant food, and use the subways and city museums more."
Quasi-commercial service with subscribers who pay by the month, or occasional users who pay per use/hour/day/etc. The city's basically competing with T-Mobile, Boingo, etc.etc. and Starbucks for roaming, plus competing against the cable modem and DSL companies for residential business. In this model, success is defined as "subscribers/users paid enough money to pay for the costs of the service", and "feasible" means "there's a reasonable chance that the service will succeed, given some pricing model." It's a much different concept.
Volunteer-hobbyist-run networks like BAWUG that provide free access - the city doesn't do the work of running it, and maybe it's less scalable, but the city provides access to streetlights and well-placed rooftops, and doesn't abuse and extort them the way they would treat a commercial provider.
"Public-Private Partnership" is usually that the city hires one of the mayor's buddies to implement the quasi-commercial model instead of having city workers do it, and ideally is friendly and cooperative and doesn't abuse and extort them the way they would treat a normal commercial provider.
$150K / 5 years is $30K/year or $2500/month. For the quasi-commercial models, yes, that means they need 100 subscribers/square mile at $25/month or 500 subscribers at $5, which is probably easier to get if the performance is ok, but there are cities where it won't work. For the public service model, $30K/year/square mile isn't a lot of money - it's certainly cheaper than paying police overtime for baseball games, and it's more likely to attract geek tourists and business tourists than a baseball stadium is, so at least in San Francisco, you'd probably justify paying for it from the hotel tax fund or something for downtown, but in residential neighborhoods it's a tossup of whether it's a win or not.
Obviously a private company is only going to use the commercial models, not the streetlight model. In a purely commercial model, if the costs look feasible, there will be a bunch of competitors; in a public-private partnership model, the costs are usually cheaper because the city government causes much less trouble to the service provider, but there's less likely to be competition unless they assign different territory to different providers.
Sure, not everybody's Jerry Garcia or Karl Franzen. But music isn't just about listening to performances - it's about *playing*. You can have a lot of fun hanging out with other musicians jamming, and even if you're playing standard pieces, the group improvisation is a really pleasant experience, and sometimes even sounds good to people who are listening. I play dulcimer, and I'm only beginning to play guitars and ukuleles and such, but once you've acquired a few basic skills you can keep up with other players at least by playing background chords and occasional improv. The San Francisco Bay Area is rife with amateur music get-togethers - lots of Irish musicians, and a bunch of old-timey and various folkies, some French stuff, and presumably lots of other genres. Old-Timey has the advantage that it's usually at a moderate speed, and usually only plays in a couple of different keys that work well for fiddlers and don't require the banjo players to retune, so three or four chords will get you started and it's fun music. (And learning a few guitar chords means that you can often look at a guitar player and figure out what keys you're in and where you are in the music, which can be immensely helpful.) Good musicians are usually very helpful and patient with people sitting in on jam sessions, though you may not be able to play every note or every measure depending on how fast the group is going.
Guitar playing isn't just about producing good music - it's about having fun doing it, creating as you go along, interacting with other players, mastering the basics of a tune and then adding depth and fooling around.
Also, back when I was in high school, one classic reason to play guitar was that girls liked it when you did that, and this robot isn't going to pick up that many girls... (I didn't play guitar back then, did find other ways to meet girls, though.)
Postscript-Based Windowing was Great!
on
Longhorn Preview
·
· Score: 1
Back in the mid-80s, James Gosling's NeWS Network Extensible Windowing System was a Postscript-based windowing system. Originally written on Suns, and ported to a variety of things by the Grasshopper Group and maybe some other people (including Macs.) What you saw was what you got, what you saw on the screen was what you saw on the printer, whatever Postscript font you wanted, you could get (so we gave my supervisor a 24-point Palatino on his Sun to make it easy to read.)
Postscript wasn't just used for display - this was an object-based system that let clients applications hand Postscript programs to the window server, so unlike X Windows, you could do work whereever it made the most sense, things like doing mouse tracking at the server and only passing significant events to the client, which made it far more responsive. Some of the things Gosling learned from this showed up later in Java - Security, for instance, was pretty dodgy... And when you iconized the terminal client, it wasn't just a dead icon - it was the same psterm window squashed down to a 1-point font (which means one pixel per character on a typical monitor), so you could still watch for activity on your terminal even though it was iconified.
Display Postscript on NeXt machines wasn't quite as powerful or flexible, but it still let you run your windows in Postscript, getting good-looking displays and a certain amount of power, and some people really liked it.
One Niche OS I'd happily run on something if it were vaguely finished is EROS-OS, the Extremely Reliable Operating System, a capability-based operating system that Jon Shapiro worked on. The security possibilities make it highly interesting, and it's designed so you can do things like unplug the machine in the middle of a calculation, plug it in again, and have it start up where it left off. And Plan 9 and its successors were designed for scalability and resource-location transparency.
Both of these OS's were designed in a deep academic environment to be able to do really interesting things, and they're fundamentally different from just building Yet Another Unix-like thing with a window system on it (ok, Plan 9 did evolve from Unix, and does have an aggressively different window system, but it's not just random me-too-ism.)
I've used a couple of niche OS's - PalmOS was clunky but had good applications on it, the Psion 3A's OS was a lot smoother and everybody really raved about the 32-bit version on the Psion 5, but alas, while the hardware was nearly bulletproof, after about the tenth time you drop it onto concrete the hinges eventually die. I'm not a gamer, so I don't have game-console OS's. MacOS? Sure, if I wanted everything to be pretty-looking and Just Work.
But why would I put a niche OS on PC hardware? Niche Linux distributions like MythTV, maybe, or LTSP lightweight distros designed to use old hardware as a thin client, or LiveCD OpenBSD firewall things or whatever.) Emulators for other hardware environments, maybe (one of the Psion development environments booted from PC MS-DOS mode, and I gather there are some gamer emulators that do similar things, and you used to need to run DOOM in MS-DOS instead of Windows to get native hardware access or something.)
Pen-based OS's were the last niche OS I saw that looked really interesting as a user - though they could just as well be a user interface on top of a full-featured operating system, and of course they choked and died and were replaced by PalmOS and Wince. QNX has always been somewhat interesting as hacker environment, because it's real-time, blazingly fast, and fits inside the Level 1 cache on your older CPU, though the last time I tried it it didn't have a driver for my Ethernet cards and was therefore pretty useless.
Any OS that wants me to spend time installing it had better have a lot of interesting features, or a few VERY interesting features, and it needs to run on a LiveCD (or floppy) on an older PC like a Pentium133 with 64MB RAM, because I'm not going to scrag my main machine to play with it. Neither of these includes a Reality Distortion Field, so their web pages need to actually say why they're interesting - and they don't. Syllable provides no obvious value - its web page says it's a fork off a 3-year-old PersonalEgoOS and doesn't say why it's more interesting than a well-supported OS. SkyOS looks like it has a screenshot tour and an 18MB AVI video tour, but it's too slashdotted to actually display those things, and screenshots might tell me why I want a new wallpaper or window manager but aren't the same as telling me what the OS *does* that's interesting - telling me that they'd like to offer a bounty for getting somebody to port OpenOffice just means they're running behind Linux and the BSDs - ZZZZ.
Yes, the Mounties don't massacre large bunches of students in big city squares, not that they're especially friendly to large anti-government protests. But the relationships with the First Nations have not always been good - you may remember the standoff in 1990 in Oka between the Akwesasne Mohawks and the various Canadian army, etc. forces.
A University is one of the last places that you'd want to block all email from China, because you're almost guaranteed to have lots of actual legitimate mail to your students and faculty. But per-user country blocklists can be quite practical. I use pobox.com, a mail forwarding service, and they let users check off which country blocklists to use - so I blacklist China, Korea, Nigeria, Brazil, and Argentina, and have Japan set for allow-with-extra-filtering. I don't know how they implement the back end (UI is a web page with checkboxes), but it seems to scale adequately for them.
You can also use this to help reduce spam sent from China to your Chinese students - if somebody who doesn't want mail from China receives mail from a given Chinese IP address, greylist that address for everybody and/or direct mail from it to your Spamassassin server instead of your regular mail server.
Besides, with China, the big problem isn't email that they send - it's web sites that get used for responses. It's much harder to filter through email message bodies looking for websites, resolving any obfuscation and DNS queries, and then deciding if they're in China before you deliver it to the user. It's much more scalable if the user's PC does that themselves, though you can still provide the country lists.
The current firewall, AFAICT, doesn't seem to stop viruses or detect zombie sites, and the China Telecom and China Netcom ISPs don't seem to do anything about spammer websites, at least not about ones in English marketing to the foreign-barbarian market that are profitable for the quasi-monopoly internet services. Perhaps they do scan for Falun-Gong propaganda, and maybe even pr0n, but not spam.
But the scalability problems are much different when you're a government bureaucracy looking for politically incorrect material to censor, when the people who want to make it available are trying to hide it from you, than when hundreds of millions of people are forwarding complaints to you about messages they've received. You're outsourcing the detection problem to people willing to do it for free. Even if you just do a crude filtering process to require N complaints from N different IP addresses before you respond to a complaint, to keep your signal-to-noise ratio down, it's easy to find the big offenders.
Most spam that I receive either wants a response sent to a disposable free email address (Yahoo/netscape/teenmail.co.za/whatever), or else wants you to connect to a website. The websites are usually in China. Occasionally they're in Korea or Brazil or elsewhere, but China's by far the dominant player.
The Great Firewall of China has a lot of hype about preventing Chinese people from receiving politically incorrect information, but as far as I can tell, it puts entirely no effort into policing English-language spammer sites, machines infected with viruses, consumer-broadband zombies, or anything like that. It's possible that they police English-language pr0n in China, and probable that they police English-Language Falun Gong sites, but they certainly don't have a problem with mortgage spammers, medicine spammers, or other scams that are directed toward outside barbarians.
Sending complaint email to abuse@ at the major Chinese ISPs has no discernable effect. If this new agreement includes a spam-complaint email address, cool, I'll be happy to Cc: them.
Brazil's not a coca producer because it's mostly lowlands, and coca's a mountain crop. Peru, Colombia, and Bolivia are the best places to grow it. I've seen an estimate that 50 square miles was enough to supply the entire world market of the stuff - no need to grow it in Brazil, especially if Brazil's government aren't rabid prohibitionists like the US government.
And no, growing legal products doesn't compete economically with growing illegal drugs for export to the US, as long as the costs of paying off the police and army are managable, which they normally are, and as long as the major product processing/distribution companies aren't having turf wars with each other, which sometimes happens. Very little of the cost or price is actually related to production - it's all about the Black Market that's caused by US domestic and exported Prohibition. The farmers are the inexpensive part of the business, and the violence, bribery, and losses to seizure are the expensive parts, and the sellers can afford to pay the farmers whatever they need to supply current demands.
If the US were to legalize drugs, the prices drop radically - the estimates I've seen from the US government anti-drug propagandists are that a non-black-market price would be about $3/gram for powder cocaine, and about $1/day for typical heroin-addict quantities of opiates, which grow easily almost anywhere, and they estimated that the black market and associated violence and corruption wouldn't be sustainable competing with legal products at less than ~$10/gram. (That particular document didn't give a price estimate for legal marijuana, but since you can grow it anywhere you can grow tomatoes, it'd be about $1/pound in the summer....) This was before the current emphasis on methamphetamines as a popular drug, but they're an easily-synthesized industrial product that could be made for about the cost of Sudafed if they were legal.
Dude, there's no need to be insulting or condescending.
Actual production systems are typically much more complex than a raw Debian installation. It takes a *lot* of discipline to run a good configuration management environment at all, something that far too many companies with web sites don't do well, much less to run it in a way that gives you a trusted base for every application on your system.
You're probably not just going to have raw Debian and some data files. There'll be databases (typically Oracle or something of similar power, for which you don't have source code), home-grown applications, for which the source code lives on your development environment, testing and instrumentation to model what's really happening on the production system, little hackish shell/perl/etc. scripts to glue things together, detritus that's accumulated over several years that nobody's really sure what it does, etc.
Furthermore, unless your CM people are extremely good, it's hard to build an adequate development environment without some connection to the production systems, and that means that there are going to be some crossover connections between the two, possibly some developer logins on the production system (you *do* check all your SSH permission files into CVS, and your DNS resolvers, and all the proxy settings for Mozilla that each user uses, don't you? It'd be a real shame if the Debian you thought you were connecting to wasn't the *real* Debian.), and that means there's a risk that your developers will get *their* environments contaminated before you've noticed the breakin, just as there's a risk that your backups will get contaminated. And I've almost *never* seen a clean development shop environment - it's tough enough to get the developers to make sure the QA department is testing on a properly-built-from-scratch environment for each subrelease, as opposed to upgrading from the previous version.
Yes, SP2 is much better than previous MS versions. It's still Windows, it's still the Internet, and you're still probably running vulnerable applications. So you still need defense in depth. If you want to open holes in the firewall for specific applications, most cheap hardware firewalls will let you do this, but don't just leave the thing wide open handing packets to Windows hoping that something hasn't opened holes in Windows's firewall, either maliciously or just to implement their own application.
And if you've got a network at home, with shared printers or file sharing or anything at all that lets your separate PCs talk to each other, you especially need the firewall.
No, it doesn't, because the parts of the network with the most congestion are your upstream and downstream links, so if you're penalizing TCP packets, it's your own TCP packets, and TCP is designed so that this makes it run slower, but still reliably. That means that email, HTTP, and BitTorrent will back off and let the more latency-sensitive predictable-data-rate VOIP and maybe video applications win, and a few applications like DNS get some Extra Slack they didn't really need, and sometimes your email arrives a few milliseconds late and your web pages take a few percent longer to download. And that's almost always what you really wanted to happen (especially with BitTorrent.) A real QoS-aware network would let your applications mark packets in ways the network would pay attention to, and some ISPs offer that; this is a partial solution that may be more generally applicable.
The backbone of the public network doesn't get congested unless your ISP has serious problems, and it isn't where you'd implement something like this because of scalability concerns. The part of the public network that you really are sharing that could get congested is the downlink from your carrier's backbone to their DSLAMs, and yes, UDP would get extra priority there. For VOIP, that's seriously not a problem - if it's compressed voice, it's typically about 22-26kbps, though some uncompressed voice can get up to 80kbps, which is annoying and unnecessary for WAN voice (I mean, at least run 32kbps ADPCM if you don't want to use tighter compression...) If that's the bottleneck for enough people, some applications may try to abuse it, but the major applications that make up most people's traffic won't, because it won't help most of them.
Could this cause you problems if you're running UDP-based remote file system programs that shouldn't be prioritized, like some hypothetical public NFS server? Possibly, yes, but you usually shouldn't do that anyway.
Upon further reflection, one thing I did miss was the question of IPSEC - do you want to prioritize it, or not? Probably yes, for most people.
Trademark law is designed to protect against consumers confusing products - a floor wax and a dessert topping could both use the same name without risk of trademark dilution, but two floor waxes can't have similar names without the newer one interfering with the older one. (So the other day's article about the guy who trademarked Stealth and keeps trying to get people to pay him off is mostly abuse of the process but occasionally reasonable.)
So can you tell Moxie the soda from ink? If you spill it on paper, can you tell it from a Moxi-drawn painting? Does Moxi the software package contain enough caffeine to keep the user awake?
"No War for Hardwood!"
on
Real Wood iPod
·
· Score: 5, Informative
Actually, deforestation really is a serious problem, and decorative hardwoods are often from much more environmentally sensitive areas than pine and some of the other softwoods that grow efficiently in tree-farms. People who like fancy guitars and other musical instruments have to deal with this issue also.
Some softwood types are also sensitive - old-growth redwoods forests and high mountain areas. Forest Service roadbuilding typically costs about 10 times as much as the value of the wood that gets logged using those roads, so it's essentially subsidizing the destruction of old-growth forests; the Clinton administration belatedly got around to banning it in many areas, and the Bush Administration rapidly re-authorized it.
Disclaimer: I work for a telecom company, but this is my personal opinion only, except on the rare occasions when they have the good sense to take my advice. The question isn't whether somebody's going to eat our lunch - it's just who, and whether we're going to help them. Moore's Law has been trashing the whole computer industry's infrastructure for years; why should the telcos be any different, just because we used to be able to design for 40-year equipment lifetimes instead of 4-year?
VOIP could replace the telco infrastructure with something that would be far cheaper, simpler, and almost as reliable, if only you could build it from scratch without needing to talk to traditional phones. Doing a smooth transition is much much harder, and of course it's not clear that the traditional telcos would make the money, so they're desperately trying to find ways to do that, such as becoming Internet carriers and buying cellphone companies.
The big things driving the current telco infrastructure are
Historical evolution from far older generations of hardware with a lot less CPU power than a good digital watch (My first draft had said "horsepower", but electromechanical switches really did have a lot of horsepower...)
Regulatory infrastructures of who pays what money to whom which made sense in the old architectures and don't make much sense today
Reliability and scalability that have been high priorities and have a lot of effort put into them. The Internet today carries a lot more bits than the phone network - but that doesn't mean that it's easy to scale a large call center, or that it's easy to scale a SIP server network to support a continent, or that ENUM's model of using DNS is the right way to handle directory lookups (though it's certainly more scalable than a big honking LDAP server.) And gateways between the two sides are harder to scale than either side's internal connections.
Evolution of business data communications, because that's where much of the money is for infrastructure change. To some extent, consumer Internet matters also, because that puts some kinds of wires to people's houses.
Lack of decent radio support for traditional telephony, because the FCC gave out separate quasi-monopolies to the telcos and the radio broadcasting companies back when FDR knew that government was the best way to manage industry. It prevented development of good crossover solutions for decades, though Moore's Law was catching up with us by the time the FCC started to get out of the way.
Difficulty of scaling radio-based Internet access as a telco access line replacement. 802.11 is a fun solution for hobbyists, and a really nice solution for wandering laptops in coffee-shops, but it's harder than you'd think to build a cost-effective scalable replacement for telco access. 802.16 WiMAX will help some of the problems, but people will suck down bandwidth as fast as it's built - you're not limited to the amount of VOIP bits a cellphone replacement would use. Wireless last-500-meter solutions with a cable TV sized infrastructure feeding them could be an interesting replacement, though.
Creativity. There's never enough of that stuff to go around.
If you've got a consumer VOIP box, you've probably noticed that it wants to be placed between your cable/DSL modem and any hubs/firewalls/computers you have. That's because it *is* doing QoS on the upstream connection. It's not doing packet marking, just prioritizing packets from its VOIP hardware, and it really does make a difference.
The most constrained bandwidth in the network is usually your upstream, and that's the easiest place to prioritize. The second most constrained bandwidth is usually your downstream - that's harder to fix, but not usually full. Sometimes the connection from your DSLAM to your ISP's backbone routers will be too heavily oversubscribed, and that can be a performance problem too, but once you've hit your ISP's backbone, you're in safe territory, even though your packets aren't marked for priority.
If your ISP *wanted* to get fancy, prioritizing UDP packets over TCP packets would do the job, with one exception - some P2P programs. Fortunately, BitTorrent uses TCP.
Telco services are designed for high reliability, which does cost money to design for and build, but a big difference is how many workers and repair trucks they have per customer. Cable TV's reliability model is that it's just television, and if your TV cable goes out for the weekend you can watch videotapes or read a book until they can get somebody there to fix it, which isn't going to happen during a hurricane. As long as it's sufficiently reliable that the customers don't dump your service for satellite or rabbit-ears, and you don't have to pay too many refunds for bad service, it's good enough. Adding cable modem service doesn't change the fundamental model, even though the users get a lot grumpier about losing connectivity.
DSL service, on the other hand, doesn't get the same repair priorities that POTS service gets, especially if it's from a CLEC that's renting copper as opposed to an ISP using a telco-managed DSLAM. If your DSL goes out, it'll often take a while for that to get fixed also.
And the backup for residential VOIP service is cell phones (do you know anybody who has VOIP who doesn't have a cell phone? Probably 95-99% of cable modem and DSL users also have cell phones...)
Really Cheap Global Calling - Nigerians can call you for nearly free. Telemarketers who aren't total scammers can also call you for nearly free (the telco costs are already much less than US minimum wage, but foreign workers can be cheaper.) US Don't-Call-List laws don't have jurisdiction over non-US call centers, though they do cut down on the products that can be sold that way. A couple of years ago I got a call from a Nigerian Scammer using the Deaf Relay Operator services (which are free, and have Internet support) - apparently those services have cut back on fraudulent abuse, but VOIP is still dirt cheap and getting cheaper.
Hiding call origins - CallerID is already *so* easy to fake out, and VOIP is likely to mean that calls appear to come from some local number near you instead of from wherever the spammer really lives. Sure, there's a bill, but even if it's not just a bulk bill from the telco to the VOIP company, it'll just show the equivalent of a bought-for-cash phone card, essentially untraceable.
Direct VOIP-to-VOIP calls - As VOIP systems become more interoperable, direct calls will become increasingly possible, and we'll have to build the equivalent of spam filters and anti-spammer-block-lists for VOIP to deal with it because it will be easier to automate things than it is in the current telco network. Business systems like your PBX may take a bit longer to be compatible, but consumer-oriented systems like Skype and Vonage and such already support user-to-user calls for free. Just wait until Skype gets reverse-engineered a bit more... and the SIP-based systems are already designed for openness.
Slashdot Mirror
The photo-quality printers will still be expensive, and the all-in-one printer/scanner/copier/fax combos, but if I was able to follow the not-very-specific PR well enough, the low-end office-quality printers will be about $49, with ~$10 print cartridges, so at least for printing mostly text and maps, at 1200x1200 resolutions, it'll be a lot cheaper. (The resolution is "optimized" to 4800x1200-equivalent, and it's hard to tell whether the second 1200 is really 1200, or is "1200-optimized" and really some lower resolution like 600, but for just about anything other than photos, it's still really enough.)
If they can integrate the printhead into the printer, rather than making it a separate replaceable component, at least you'll have to buy a new printer body when the head wears out - it's not quite as good as buying a new one when the ink runs out, but it's a good start.
The technology's a bit older, according to this very nice article on Stanford spinoffs. Canon loaned Stanford a small laser printer in 1979, and Stanford folks integrated a microprocessor with it and then spun off Imagen to commericialize it.
They're being offered free colo and bandwidth. If they were to rent that for $300/month, they'd have spent $3000 in 10 months (or 15 months, if they're paying $200/month.) That's not very long - it's much better financially to buy the hardware and take the free monthly. Also, it's probably much easier to raise funds for a one-time item like this than to beg for rent money every month.
Sorry. Eyeball recognition is right out. Besides, there are lots of problems with using it in practice that aren't that easy - databases still are never 100% correct, and once the Feds have put somebody with a similar name to yours on the Don't Fly List without a trial, and the airlines have put your eyeball print on the computer with your name and refused to put you on a flight because your name is similar to the guy on the list, your terrorist status is just going to propagate to anything that uses your eyeball print.
The question is what are their assumptions about "what's a user" and "what's feasible"? I've seen several models for community wireless, and they've got much different definitions
- Free service that makes the community more friendly, so that the citizens are happier and businesses make more money, similar to the ways streetlights and socialized baseball stadiums do. In this model, success means "everybody loves it, tourists find it easier to get around, people drink more coffee, eat more restaurant food, and use the subways and city museums more."
- Quasi-commercial service with subscribers who pay by the month, or occasional users who pay per use/hour/day/etc. The city's basically competing with T-Mobile, Boingo, etc.etc. and Starbucks for roaming, plus competing against the cable modem and DSL companies for residential business. In this model, success is defined as "subscribers/users paid enough money to pay for the costs of the service", and "feasible" means "there's a reasonable chance that the service will succeed, given some pricing model." It's a much different concept.
- Volunteer-hobbyist-run networks like BAWUG that provide free access - the city doesn't do the work of running it, and maybe it's less scalable, but the city provides access to streetlights and well-placed rooftops, and doesn't abuse and extort them the way they would treat a commercial provider.
- "Public-Private Partnership" is usually that the city hires one of the mayor's buddies to implement the quasi-commercial model instead of having city workers do it, and ideally is friendly and cooperative and doesn't abuse and extort them the way they would treat a normal commercial provider.
$150K / 5 years is $30K/year or $2500/month. For the quasi-commercial models, yes, that means they need 100 subscribers/square mile at $25/month or 500 subscribers at $5, which is probably easier to get if the performance is ok, but there are cities where it won't work. For the public service model, $30K/year/square mile isn't a lot of money - it's certainly cheaper than paying police overtime for baseball games, and it's more likely to attract geek tourists and business tourists than a baseball stadium is, so at least in San Francisco, you'd probably justify paying for it from the hotel tax fund or something for downtown, but in residential neighborhoods it's a tossup of whether it's a win or not.Obviously a private company is only going to use the commercial models, not the streetlight model. In a purely commercial model, if the costs look feasible, there will be a bunch of competitors; in a public-private partnership model, the costs are usually cheaper because the city government causes much less trouble to the service provider, but there's less likely to be competition unless they assign different territory to different providers.
Sure, not everybody's Jerry Garcia or Karl Franzen. But music isn't just about listening to performances - it's about *playing*. You can have a lot of fun hanging out with other musicians jamming, and even if you're playing standard pieces, the group improvisation is a really pleasant experience, and sometimes even sounds good to people who are listening. I play dulcimer, and I'm only beginning to play guitars and ukuleles and such, but once you've acquired a few basic skills you can keep up with other players at least by playing background chords and occasional improv. The San Francisco Bay Area is rife with amateur music get-togethers - lots of Irish musicians, and a bunch of old-timey and various folkies, some French stuff, and presumably lots of other genres. Old-Timey has the advantage that it's usually at a moderate speed, and usually only plays in a couple of different keys that work well for fiddlers and don't require the banjo players to retune, so three or four chords will get you started and it's fun music. (And learning a few guitar chords means that you can often look at a guitar player and figure out what keys you're in and where you are in the music, which can be immensely helpful.) Good musicians are usually very helpful and patient with people sitting in on jam sessions, though you may not be able to play every note or every measure depending on how fast the group is going.
Also, back when I was in high school, one classic reason to play guitar was that girls liked it when you did that, and this robot isn't going to pick up that many girls... (I didn't play guitar back then, did find other ways to meet girls, though.)
Postscript wasn't just used for display - this was an object-based system that let clients applications hand Postscript programs to the window server, so unlike X Windows, you could do work whereever it made the most sense, things like doing mouse tracking at the server and only passing significant events to the client, which made it far more responsive. Some of the things Gosling learned from this showed up later in Java - Security, for instance, was pretty dodgy... And when you iconized the terminal client, it wasn't just a dead icon - it was the same psterm window squashed down to a 1-point font (which means one pixel per character on a typical monitor), so you could still watch for activity on your terminal even though it was iconified.
Display Postscript on NeXt machines wasn't quite as powerful or flexible, but it still let you run your windows in Postscript, getting good-looking displays and a certain amount of power, and some people really liked it.
Both of these OS's were designed in a deep academic environment to be able to do really interesting things, and they're fundamentally different from just building Yet Another Unix-like thing with a window system on it (ok, Plan 9 did evolve from Unix, and does have an aggressively different window system, but it's not just random me-too-ism.)
But why would I put a niche OS on PC hardware? Niche Linux distributions like MythTV, maybe, or LTSP lightweight distros designed to use old hardware as a thin client, or LiveCD OpenBSD firewall things or whatever.) Emulators for other hardware environments, maybe (one of the Psion development environments booted from PC MS-DOS mode, and I gather there are some gamer emulators that do similar things, and you used to need to run DOOM in MS-DOS instead of Windows to get native hardware access or something.)
Pen-based OS's were the last niche OS I saw that looked really interesting as a user - though they could just as well be a user interface on top of a full-featured operating system, and of course they choked and died and were replaced by PalmOS and Wince. QNX has always been somewhat interesting as hacker environment, because it's real-time, blazingly fast, and fits inside the Level 1 cache on your older CPU, though the last time I tried it it didn't have a driver for my Ethernet cards and was therefore pretty useless.
Any OS that wants me to spend time installing it had better have a lot of interesting features, or a few VERY interesting features, and it needs to run on a LiveCD (or floppy) on an older PC like a Pentium133 with 64MB RAM, because I'm not going to scrag my main machine to play with it. Neither of these includes a Reality Distortion Field, so their web pages need to actually say why they're interesting - and they don't. Syllable provides no obvious value - its web page says it's a fork off a 3-year-old PersonalEgoOS and doesn't say why it's more interesting than a well-supported OS. SkyOS looks like it has a screenshot tour and an 18MB AVI video tour, but it's too slashdotted to actually display those things, and screenshots might tell me why I want a new wallpaper or window manager but aren't the same as telling me what the OS *does* that's interesting - telling me that they'd like to offer a bounty for getting somebody to port OpenOffice just means they're running behind Linux and the BSDs - ZZZZ.
Yes, the Mounties don't massacre large bunches of students in big city squares, not that they're especially friendly to large anti-government protests. But the relationships with the First Nations have not always been good - you may remember the standoff in 1990 in Oka between the Akwesasne Mohawks and the various Canadian army, etc. forces.
You can also use this to help reduce spam sent from China to your Chinese students - if somebody who doesn't want mail from China receives mail from a given Chinese IP address, greylist that address for everybody and/or direct mail from it to your Spamassassin server instead of your regular mail server.
Besides, with China, the big problem isn't email that they send - it's web sites that get used for responses. It's much harder to filter through email message bodies looking for websites, resolving any obfuscation and DNS queries, and then deciding if they're in China before you deliver it to the user. It's much more scalable if the user's PC does that themselves, though you can still provide the country lists.
But the scalability problems are much different when you're a government bureaucracy looking for politically incorrect material to censor, when the people who want to make it available are trying to hide it from you, than when hundreds of millions of people are forwarding complaints to you about messages they've received. You're outsourcing the detection problem to people willing to do it for free. Even if you just do a crude filtering process to require N complaints from N different IP addresses before you respond to a complaint, to keep your signal-to-noise ratio down, it's easy to find the big offenders.
The Great Firewall of China has a lot of hype about preventing Chinese people from receiving politically incorrect information, but as far as I can tell, it puts entirely no effort into policing English-language spammer sites, machines infected with viruses, consumer-broadband zombies, or anything like that. It's possible that they police English-language pr0n in China, and probable that they police English-Language Falun Gong sites, but they certainly don't have a problem with mortgage spammers, medicine spammers, or other scams that are directed toward outside barbarians.
Sending complaint email to abuse@ at the major Chinese ISPs has no discernable effect. If this new agreement includes a spam-complaint email address, cool, I'll be happy to Cc: them.
And no, growing legal products doesn't compete economically with growing illegal drugs for export to the US, as long as the costs of paying off the police and army are managable, which they normally are, and as long as the major product processing/distribution companies aren't having turf wars with each other, which sometimes happens. Very little of the cost or price is actually related to production - it's all about the Black Market that's caused by US domestic and exported Prohibition. The farmers are the inexpensive part of the business, and the violence, bribery, and losses to seizure are the expensive parts, and the sellers can afford to pay the farmers whatever they need to supply current demands.
If the US were to legalize drugs, the prices drop radically - the estimates I've seen from the US government anti-drug propagandists are that a non-black-market price would be about $3/gram for powder cocaine, and about $1/day for typical heroin-addict quantities of opiates, which grow easily almost anywhere, and they estimated that the black market and associated violence and corruption wouldn't be sustainable competing with legal products at less than ~$10/gram. (That particular document didn't give a price estimate for legal marijuana, but since you can grow it anywhere you can grow tomatoes, it'd be about $1/pound in the summer... .) This was before the current emphasis on methamphetamines as a popular drug, but they're an easily-synthesized industrial product that could be made for about the cost of Sudafed if they were legal.
Actual production systems are typically much more complex than a raw Debian installation. It takes a *lot* of discipline to run a good configuration management environment at all, something that far too many companies with web sites don't do well, much less to run it in a way that gives you a trusted base for every application on your system.
You're probably not just going to have raw Debian and some data files. There'll be databases (typically Oracle or something of similar power, for which you don't have source code), home-grown applications, for which the source code lives on your development environment, testing and instrumentation to model what's really happening on the production system, little hackish shell/perl/etc. scripts to glue things together, detritus that's accumulated over several years that nobody's really sure what it does, etc.
Furthermore, unless your CM people are extremely good, it's hard to build an adequate development environment without some connection to the production systems, and that means that there are going to be some crossover connections between the two, possibly some developer logins on the production system (you *do* check all your SSH permission files into CVS, and your DNS resolvers, and all the proxy settings for Mozilla that each user uses, don't you? It'd be a real shame if the Debian you thought you were connecting to wasn't the *real* Debian.), and that means there's a risk that your developers will get *their* environments contaminated before you've noticed the breakin, just as there's a risk that your backups will get contaminated. And I've almost *never* seen a clean development shop environment - it's tough enough to get the developers to make sure the QA department is testing on a properly-built-from-scratch environment for each subrelease, as opposed to upgrading from the previous version.
And if you've got a network at home, with shared printers or file sharing or anything at all that lets your separate PCs talk to each other, you especially need the firewall.
The backbone of the public network doesn't get congested unless your ISP has serious problems, and it isn't where you'd implement something like this because of scalability concerns. The part of the public network that you really are sharing that could get congested is the downlink from your carrier's backbone to their DSLAMs, and yes, UDP would get extra priority there. For VOIP, that's seriously not a problem - if it's compressed voice, it's typically about 22-26kbps, though some uncompressed voice can get up to 80kbps, which is annoying and unnecessary for WAN voice (I mean, at least run 32kbps ADPCM if you don't want to use tighter compression...) If that's the bottleneck for enough people, some applications may try to abuse it, but the major applications that make up most people's traffic won't, because it won't help most of them.
Could this cause you problems if you're running UDP-based remote file system programs that shouldn't be prioritized, like some hypothetical public NFS server? Possibly, yes, but you usually shouldn't do that anyway.
Upon further reflection, one thing I did miss was the question of IPSEC - do you want to prioritize it, or not? Probably yes, for most people.
So can you tell Moxie the soda from ink? If you spill it on paper, can you tell it from a Moxi-drawn painting? Does Moxi the software package contain enough caffeine to keep the user awake?
Some softwood types are also sensitive - old-growth redwoods forests and high mountain areas. Forest Service roadbuilding typically costs about 10 times as much as the value of the wood that gets logged using those roads, so it's essentially subsidizing the destruction of old-growth forests; the Clinton administration belatedly got around to banning it in many areas, and the Bush Administration rapidly re-authorized it.
VOIP could replace the telco infrastructure with something that would be far cheaper, simpler, and almost as reliable, if only you could build it from scratch without needing to talk to traditional phones. Doing a smooth transition is much much harder, and of course it's not clear that the traditional telcos would make the money, so they're desperately trying to find ways to do that, such as becoming Internet carriers and buying cellphone companies.
The big things driving the current telco infrastructure are
The most constrained bandwidth in the network is usually your upstream, and that's the easiest place to prioritize. The second most constrained bandwidth is usually your downstream - that's harder to fix, but not usually full. Sometimes the connection from your DSLAM to your ISP's backbone routers will be too heavily oversubscribed, and that can be a performance problem too, but once you've hit your ISP's backbone, you're in safe territory, even though your packets aren't marked for priority.
If your ISP *wanted* to get fancy, prioritizing UDP packets over TCP packets would do the job, with one exception - some P2P programs. Fortunately, BitTorrent uses TCP.
DSL service, on the other hand, doesn't get the same repair priorities that POTS service gets, especially if it's from a CLEC that's renting copper as opposed to an ISP using a telco-managed DSLAM. If your DSL goes out, it'll often take a while for that to get fixed also.
And the backup for residential VOIP service is cell phones (do you know anybody who has VOIP who doesn't have a cell phone? Probably 95-99% of cable modem and DSL users also have cell phones...)