From observation to collection, the harvesting process takes about 9 seconds per berry. That's too slow.
This isn't the first strawberry-picking robot. Here's one from five years ago. But compare this with a commercial strawberry harvester that's just digging up the beds. (Note, incidentally, that the tractor is driverless. That's standard precision farming technology today; several GPS manufacturers make the gear for that.)
Automated fruit sorting using computer vision is a routine process, and it's really fast. Small-fruit sorting machines are strange to watch. Cameras watch the fruit go by, and air jets push it around. This is all happening in bulk, much faster than humans can even watch, as big conveyors pump a stream of mixed product through the machine and streams of sorted product come out.
Robotic tomato pickers have been built by several groups, but so far the machines are too slow and the cost is too high.
In practice, the way agricultural sorting works is that the good stuff is sold is fresh fruit, the not-so-good stuff goes off to make jellies, tomato paste, and such, and the rejected stuff becomes animal feed or fertilizer.
"Instructions: Based on the classic arcade game Pac-Man, the aim is to eat all the pills in the maze, while avoiding the four ghosts. There are also power-pills available in each corner which temporarily turn the ghosts blue, and more importantly, edible! Bonuses are awarded for eating ghosts. Fruit bonuses also appear and can be eaten for additional points."
Here's a typical break-in, at University of Oakland.. This has a good search position in Google for "64 bit Windows". This leads to a software-for-sale page with phony seals of approval from Microsoft, Verisign, etc. That's hosted at Starnet, in Moldovia. The payment site for the sales site is "payment8ltd.net", also hosted on Starnet in Moldovia. They're selling pirated copies of brand-name software at roughly half retail price.
That site has a TrustWave seal, which pops up a popup for Paym8, a real payment processor in Zaire. TrustWave's seal server doesn't check the referrer when displaying a seal popup, so it can be spoofed. Nor does the TrustWave seal even give the domains to which it applies. Verisign and BBBonline check this, but not TrustWave.
It looks like the actual payment processing occurs at "https://payment8ltd.net/shop/order/process/"; that's where the order goes on "Submit". The site has one of those worthless GoDaddy "Domain control only validated" SSL certs.
Starnet presents itself as an Internet and telecom service provider, offering the usual data, voice, colocation, and hosting.
Headquarters of Starnet seems to be at Vlaicu Parcalab, 63, Chisinau, Republic of Moldova. That's a property of Flexi Offices, one of those small-office rental places. Interestingly, Microsoft also has an office in that building.
There's actual Whois information for that site:
Registrant Contact:
Viktor Menshikov
Viktor Menshikov (loyal@yourisp.ru)
ul.V.Urdasha d.36 kv.1
Rakovo, Respublika Tatarstan, RU 422455
P: +7.8435122221 F: +7.8435122221
That location exists; it's a farm town about 500Km east of Moscow. Probably not a real address.
Searching for "yourisp.ru" brings up a large number of scam reports. The domain itself is registered but not in DNS.
Most of this recent batch of attacks seem to have similar underlying information.
That's puzzling. Why would Google need high-cost data center space in NYC? They're distributed enough that it doesn't matter. I could see Google buying an office building in Manhattan and filling it with advertising salespeople, but not much hardware needs to be there.
Even for Wall Street, many of the big data centers are elsewhere, usually in New Jersey.
Several cities once had sizable pneumatic tube systems. London, Paris, Berlin, Prague, and New York City all had extensive systems.
Tube diameters were small, though, in the 2" to 3" range. The Prague system was the last to shut down, in 2002. Prague is repairing their system, and it may come back up.
The London system had the ability to automatically transfer carriers to and from from public tubes to "house systems" within a building. So it could provide end to end service. Most of the other systems were post office to post office only.
The
Chicago tunnel system had almost full coverage of downtown Chicago a century ago, with small electric trains in freight tunnels under most of the downtown streets. Goods were transferred from full-sized rail cars to tunnel cars, which were then delivered to buildings in the city and carried upward in special elevators. That system ran until 1959.
Maintaining the infrastructure for such systems is expensive for the amount of traffic, though.
Its all about the FINANCING of high ticket items. GM will make you pay...
Historically, yes. But not currently. GM sold off General Motors Acceptance Corporation (which is now Ally Financial) between 2006 and 2010, because they needed the cash. GM wanted to bring that operation back under their control, but couldn't afford it. They bought AmeriCredit, a small Texas bank active in "subprime auto lending" last October 1, and renamed it "GM Financial", but so far, it's not a big player in GM auto financing. The dealers are mostly using Ally.
The bailout, though, worked; GM is alive and well and rapidly paying off the U.S. Government.
Wikileaks is actually hosted in a data center in an underground bunker in a Swedish mountain. That was a good move. They actually need that level of protection.
The data center operator, Bahnhof, is fully behind Wikileaks in this. "The company's data center is "a kind of metaphor" for Bahnhof's commitment to resist any sort of intrusion, physical or legal. We're proud to have clients like these," he says. The Internet should be an open source for freedom of speech, and the role of an ISP is to be a neutral technological tool of access, not an instrument for collecting information from customers."
This is the fundamental problem with "crowdsourcing" reviews. Where the number of reviewers is large compared to the number of items being reviewed, as with movies, it works fine. Where the ratio is small, it doesn't. It's far too easy to game the system. There are automated tools for that.
This problem has become worse since the October 27th change to Google, when Google Places/Maps results were merged into web search. This made "local" results much more prominent. Look at the first screen of Google search results for a local product or service. Most of what you see are Google Places results, maps, or ads. The organic results are so far down they don't matter.
As a result, the "black hat" SEO companies are now aggressively targeting Google's places and maps system. "Convert Offline" is quite open about this, with their article
Dominating Google Maps- The Most Effective Spam Ever And What You Can Learn From It" In some ways, Google Places is more vulnerable to attack than organic search. The number of web mentions of a local business tends to be small, so the amount of phony material that has to be generated to make a business look good is also small. Each mention carries a lot of weight.
The most significant disclosure so far is that China's leadership is fed up with North Korea acting like a "spoiled child". Previously, China was considered to be a supporter of North Korea. Now, confirming the info from Wikileaks, Chinese officials are admitting that China's leadership is fed up with the drama. This leak was a win for both the US and China. It gets the word out that China isn't going to back any stupid actions by Kim Jong-il. without China's leadership having to say so publicly. This helps calm the situation down. That one item outweighs any harm Wikileaks may possibly have done.)
Can anyone name one Microsoft Research project that has significantly affected the computer industry?
Yes.
Microsoft's natural language work resulted in the grammar checker in Word, which really is parsing sentences, not just looking for common errors. Microsoft Research used to give out a program you could plug into Word which let you see the sentence diagrams.
Microsoft has for years been doing serious work in automated proof of correctness for programs. "Spec#", the proof system for C#, was a research result. Another effort in that area involved automated verification of Windows drivers to determine if they could crash the rest of the OS. That paid off.
In Windows 7, every driver has to pass the static verifier before it gets signed. Verified drivers may not drive the device correctly, but they don't crash the rest of the OS. (Yes, there's a formal undecidability problem. In practice, the system can either provide a proof or a counterexample for 97% of drivers submitted. The remaining 3% are typically flaky anyway; if your kernel driver has formally undecidable semantics, it needs a rewrite.)
There's more, but that's enough for now. Microsoft really does have one of the very few pure research groups left in computer science.
The "Singularity" guys have been around for a while. I've met many of them, from Eric Drexler to Bill Joy Rod Brooks to Ed Feigenbaum. All of them talked about strong AI Real Soon Now. We're not even close.
There's steady progress today, though. The "expert systems" guys were full of shit, and we had 15 years of "AI Winter" once people figured that out. Now, the machine learning guys are in charge, and making progress.
AI, as a field, has the problem that too many people think we're one good idea away from strong AI. Each time somebody has a reasonably good idea, like tree search, the General Problem Solver, hill climbing, expert systems, genetic algorithms, or support vector machines, it's hyped as being the step that will take us to strong AI. Each time, after a few years, most of the things that can be done with the new idea have been done, and we're stuck again. At a higher level, though. Each time around, a few more things that used to require humans are now done by machines.
The encouraging thing about the current state of the art in AI is that there are useful, shipping products in volume production. That wasn't really true until a decade ago. The earlier technologies never resulted in successful products. Because of this, the field is now economically self-supporting; more money is put into it by the successful manufacturers.
The problem with Kurtzweil is that he's pre AI winter. His real work was in the 1980s. Since then, he's been a pundit, which becomes embarrassing if you do it for too long.
Neither Windows nor Linux has per-application compartmentalized security. In theory, you could use something like SELinux to give each vendor their own compartment, preventing an install from vendor A from affecting an install from vendor B. But the installers would have to be aware of this, and carefully stay in their own spaces, or installations would fail. Nobody does that.
(Someday, somebody is going to crack the signing key for Windows update, hijack a router to reroute Microsoft's IP address, and take over every Windows machine in the world.)
One of the more embarrassing items is this: American officials sharply warned Germany in 2007 not to enforce arrest warrants for Central Intelligence Agency officers involved in a bungled operation in which an innocent German citizen with the same name as a suspected militant was mistakenly kidnapped and held for months in Afghanistan. A senior American diplomat told a German official "that our intention was not to threaten Germany, but rather to urge that the German government weigh carefully at every step of the way the implications for relations with the U.S."
What seems to be driving the copyright enforcers nuts is simply that Syfert's package of boilerplate letters contains one raising a "personal jurisdiction" issue. The copyright enforcers have been filing all their suits in one Federal court, in the District of Columbia, regardless of where the defendant is. In such cases, where the defendant has no connection to the district where the court is, it's routine to object, and force the plaintiff to refile in the defendant's district court. The lawyers for the US Copyright Group then have to sue in dozens of different district courts. They hate that.
These days I'd wager that the vast (VAST!) majority of packets are part of ongoing streams - streaming Netflix over the net, torrenting the collected porn of the 80ies, that kind of thing.
True. However, many of those streams are bandwidth-adaptive and heavily buffered.
That's been known in the TCP community for decades.
I looked at this back in my RFC 896 days, when TCP was in initial development and I was working on congestion. I introduced the "congestion window" concept and put it in a TCP implementation (3COM's UNET, which predated Berkeley BSD). The question was, what should be the initial size of the congestion window? If it's small, you get "slow start"; if it's large, the sender can blast a big chunk of data at the receiver at start, up to the amount of buffering the receiver is advertising.
I decided back then to start with a big congestion window, because starting with a small one would slow down traffic even when bandwidth was available. One of the big performance issues back then was the time required to FTP a directory across a LAN, where TCP connections were being set up and torn down at a high rate. So startup time mattered. The decision to go with a smaller initial congestion window size came years later, from others. This reflected trends in router design. I wanted routers to have "fair queuing", so that sending lots of packets from one source didn't gain the sender any bandwidth over sending few packets. But routers gained speed faster than RAM costs dropped, and so faster routers couldn't have
enough RAM for fair queuing. Today, your "last mile" CISCO router might have fair queuing. Some DOCSIS cable modem termination units have it. But many routers are running Random Early Drop, which is a simple but mediocre approach. (The backbone routers barely queue at all; if they can't forward something fast, they drop it. Network design tries to keep the congestion near the edges, where it can be dealt with.)
Remember, every dropped packet has to be retransmitted. (Too much of that leads to congestion collapse, a term I coined in 1984. That's what the "Nagle algorithm" is about.) In a world with packet-dropping routers, "slow start" makes sense. So that was put into TCP in the late 1980s (by which time I was out of networking.)
However, the RFC-documented slow start algorithm is rather conservative. RFC 2001 says to start at one maximum segment size. Microsoft's implementations in Win95 and later start at two maximum segment sizes.
In RFC 3390, from 2002, the limit was raised to 3 or 4 maximum segment sizes. (We used to worry about delaying keystroke echo too much because big FTP packets were tying up the 9600 baud lines too long. We're past that.)
But Google is sending at least 8 segments at start, and Microsoft was observed to be sending 43. Sending 43 packets blind is definitely overdoing it.
I wonder whether they're doing this blindly, or if there's more smarts behind the scenes. If their TCP implementation kept a cache of recent final congestion window sizes by IP address, they could legitimately start off the next connection with the value from the last one. So, having discovered a path that's not dropping big bursts of packets, they could legitimately start fast. If they're just doing it the dumb way, starting fast every time, that's going to choke some part of the net under heavy load.
Why run this through Twitter? If the server wants to send an SMS message, it should just send an SMS message using an SMS gateway. Why package it as a "tweet?"
(I suspect why. So they can spam you. It's illegal to send unsolicited commercial SMS messages in the US. If PayPal makes you "follow" them on Twitter to get transaction confirmations, they can then send you ads, too.)
The opposition is only entropy.
on
Anxiety and IT?
·
· Score: 2, Insightful
If the opposition is just entropy, it's not too bad. Active opposition is much more stressful. Lifeguards, firefighters, and EMTs tend not to be overly stressed. Cops and soldiers, though, routinely get stressed out.
What they actually did was to store about 100 bytes.
This may be useful for putting copyright information into genetically engineered organisms. As a method of bulk data storage, though, it leaves much to be desired.
DNA synthesis costs about $0.29 per base pair. Sequencing is a bit cheaper, but you currently get less than 1000 base pairs sequenced per run. Reading and writing takes a room of expensive wet lab gear, and hours to days.
This predates MMO games. Pokemon ("Gotta Catch 'Em All") produced the same mindset. There were predecessors to Pokemon, but it was the first one to get huge. Arguably, Wizards of the Coast introduced this genre with "Magic, the Gathering" in 1990. Collectable cards have been around for about a century, but they were usually tied to real-world sports. Wizards of the Coast detached them from that world and made then stand alone.
Wizards of the Coast, however, managed not to be slimeballs. Zynga (Farmville, Mafia Wars, etc.) has a slimy history.
Their problem is that they've lost indexing data, not the underlying documents. So just make the documents, which are public records, visible to Google. Google will index them and anyone can then search.
Bear in mind that a spacecraft launch and an ICBM launch look very similar, and a re-entry looks like an incoming missile. It's best if everybody knows where and when to expect such events, so that various military forces don't overreact. Both the normal scenarios and the abort plans need to be reviewed.
Slashdot Mirror
Believe it or not, clones are not illegal.
Nope. That's been litigated, for a PacMan clone, no less. See Munchkin, 1981, the PacMan clone for the Philips Videopac. Atari sued Phillips and won.
From observation to collection, the harvesting process takes about 9 seconds per berry. That's too slow.
This isn't the first strawberry-picking robot. Here's one from five years ago. But compare this with a commercial strawberry harvester that's just digging up the beds. (Note, incidentally, that the tractor is driverless. That's standard precision farming technology today; several GPS manufacturers make the gear for that.)
Automated fruit sorting using computer vision is a routine process, and it's really fast. Small-fruit sorting machines are strange to watch. Cameras watch the fruit go by, and air jets push it around. This is all happening in bulk, much faster than humans can even watch, as big conveyors pump a stream of mixed product through the machine and streams of sorted product come out.
Robotic tomato pickers have been built by several groups, but so far the machines are too slow and the cost is too high.
In practice, the way agricultural sorting works is that the good stuff is sold is fresh fruit, the not-so-good stuff goes off to make jellies, tomato paste, and such, and the rejected stuff becomes animal feed or fertilizer.
"Instructions: Based on the classic arcade game Pac-Man, the aim is to eat all the pills in the maze, while avoiding the four ghosts. There are also power-pills available in each corner which temporarily turn the ghosts blue, and more importantly, edible! Bonuses are awarded for eating ghosts. Fruit bonuses also appear and can be eaten for additional points."
Of course you got an infringement notice.
Press coverage today is more favorable to Wikileaks.
There's even talk that Assange might be Time's "Man of the Year".
Also, there are now 74 mirrors of Wikileaks.
Here's a typical break-in, at University of Oakland.. This has a good search position in Google for "64 bit Windows". This leads to a software-for-sale page with phony seals of approval from Microsoft, Verisign, etc. That's hosted at Starnet, in Moldovia. The payment site for the sales site is "payment8ltd.net", also hosted on Starnet in Moldovia. They're selling pirated copies of brand-name software at roughly half retail price.
That site has a TrustWave seal, which pops up a popup for Paym8, a real payment processor in Zaire. TrustWave's seal server doesn't check the referrer when displaying a seal popup, so it can be spoofed. Nor does the TrustWave seal even give the domains to which it applies. Verisign and BBBonline check this, but not TrustWave.
It looks like the actual payment processing occurs at "https://payment8ltd.net/shop/order/process/"; that's where the order goes on "Submit". The site has one of those worthless GoDaddy "Domain control only validated" SSL certs.
Starnet presents itself as an Internet and telecom service provider, offering the usual data, voice, colocation, and hosting. Headquarters of Starnet seems to be at Vlaicu Parcalab, 63, Chisinau, Republic of Moldova. That's a property of Flexi Offices, one of those small-office rental places. Interestingly, Microsoft also has an office in that building.
There's actual Whois information for that site:
Registrant Contact: Viktor Menshikov
Viktor Menshikov (loyal@yourisp.ru)
ul.V.Urdasha d.36 kv.1
Rakovo, Respublika Tatarstan, RU 422455
P: +7.8435122221 F: +7.8435122221
That location exists; it's a farm town about 500Km east of Moscow. Probably not a real address.
Searching for "yourisp.ru" brings up a large number of scam reports. The domain itself is registered but not in DNS.
Most of this recent batch of attacks seem to have similar underlying information.
That's puzzling. Why would Google need high-cost data center space in NYC? They're distributed enough that it doesn't matter. I could see Google buying an office building in Manhattan and filling it with advertising salespeople, but not much hardware needs to be there.
Even for Wall Street, many of the big data centers are elsewhere, usually in New Jersey.
Several cities once had sizable pneumatic tube systems. London, Paris, Berlin, Prague, and New York City all had extensive systems. Tube diameters were small, though, in the 2" to 3" range. The Prague system was the last to shut down, in 2002. Prague is repairing their system, and it may come back up.
The London system had the ability to automatically transfer carriers to and from from public tubes to "house systems" within a building. So it could provide end to end service. Most of the other systems were post office to post office only.
The Chicago tunnel system had almost full coverage of downtown Chicago a century ago, with small electric trains in freight tunnels under most of the downtown streets. Goods were transferred from full-sized rail cars to tunnel cars, which were then delivered to buildings in the city and carried upward in special elevators. That system ran until 1959.
Maintaining the infrastructure for such systems is expensive for the amount of traffic, though.
Its all about the FINANCING of high ticket items. GM will make you pay ...
Historically, yes. But not currently. GM sold off General Motors Acceptance Corporation (which is now Ally Financial) between 2006 and 2010, because they needed the cash. GM wanted to bring that operation back under their control, but couldn't afford it. They bought AmeriCredit, a small Texas bank active in "subprime auto lending" last October 1, and renamed it "GM Financial", but so far, it's not a big player in GM auto financing. The dealers are mostly using Ally.
The bailout, though, worked; GM is alive and well and rapidly paying off the U.S. Government.
Wikileaks is actually hosted in a data center in an underground bunker in a Swedish mountain. That was a good move. They actually need that level of protection.
The data center operator, Bahnhof, is fully behind Wikileaks in this. "The company's data center is "a kind of metaphor" for Bahnhof's commitment to resist any sort of intrusion, physical or legal. We're proud to have clients like these," he says. The Internet should be an open source for freedom of speech, and the role of an ISP is to be a neutral technological tool of access, not an instrument for collecting information from customers."
This is the fundamental problem with "crowdsourcing" reviews. Where the number of reviewers is large compared to the number of items being reviewed, as with movies, it works fine. Where the ratio is small, it doesn't. It's far too easy to game the system. There are automated tools for that.
This problem has become worse since the October 27th change to Google, when Google Places/Maps results were merged into web search. This made "local" results much more prominent. Look at the first screen of Google search results for a local product or service. Most of what you see are Google Places results, maps, or ads. The organic results are so far down they don't matter.
As a result, the "black hat" SEO companies are now aggressively targeting Google's places and maps system. "Convert Offline" is quite open about this, with their article Dominating Google Maps- The Most Effective Spam Ever And What You Can Learn From It" In some ways, Google Places is more vulnerable to attack than organic search. The number of web mentions of a local business tends to be small, so the amount of phony material that has to be generated to make a business look good is also small. Each mention carries a lot of weight.
Google might lose this battle. Craigslist did. Back in 2008, Cory Doctorow wrote about "Spammers discuss breaking Craigslist verification system". It's become much worse since then. Personals were the first to go, and are now over 90% spam. Then Computer Services and Self Employment fell to the spammers. Jobs and Real Estate are under attack. Along the way, Gmail became a spam haven, especially after Jiffy Gmail Email Creator became widely used.
The fundamental design assumption of Google is that important stuff has lots of links to it. That's not a valid assumption in local search.
The most significant disclosure so far is that China's leadership is fed up with North Korea acting like a "spoiled child". Previously, China was considered to be a supporter of North Korea. Now, confirming the info from Wikileaks, Chinese officials are admitting that China's leadership is fed up with the drama. This leak was a win for both the US and China. It gets the word out that China isn't going to back any stupid actions by Kim Jong-il. without China's leadership having to say so publicly. This helps calm the situation down. That one item outweighs any harm Wikileaks may possibly have done.)
(Here's the best analysis of the Korean situation I've seen in print.)
Can anyone name one Microsoft Research project that has significantly affected the computer industry?
Yes.
Microsoft's natural language work resulted in the grammar checker in Word, which really is parsing sentences, not just looking for common errors. Microsoft Research used to give out a program you could plug into Word which let you see the sentence diagrams.
Microsoft has for years been doing serious work in automated proof of correctness for programs. "Spec#", the proof system for C#, was a research result. Another effort in that area involved automated verification of Windows drivers to determine if they could crash the rest of the OS. That paid off. In Windows 7, every driver has to pass the static verifier before it gets signed. Verified drivers may not drive the device correctly, but they don't crash the rest of the OS. (Yes, there's a formal undecidability problem. In practice, the system can either provide a proof or a counterexample for 97% of drivers submitted. The remaining 3% are typically flaky anyway; if your kernel driver has formally undecidable semantics, it needs a rewrite.)
There's more, but that's enough for now. Microsoft really does have one of the very few pure research groups left in computer science.
The "Singularity" guys have been around for a while. I've met many of them, from Eric Drexler to Bill Joy Rod Brooks to Ed Feigenbaum. All of them talked about strong AI Real Soon Now. We're not even close.
There's steady progress today, though. The "expert systems" guys were full of shit, and we had 15 years of "AI Winter" once people figured that out. Now, the machine learning guys are in charge, and making progress.
AI, as a field, has the problem that too many people think we're one good idea away from strong AI. Each time somebody has a reasonably good idea, like tree search, the General Problem Solver, hill climbing, expert systems, genetic algorithms, or support vector machines, it's hyped as being the step that will take us to strong AI. Each time, after a few years, most of the things that can be done with the new idea have been done, and we're stuck again. At a higher level, though. Each time around, a few more things that used to require humans are now done by machines.
The encouraging thing about the current state of the art in AI is that there are useful, shipping products in volume production. That wasn't really true until a decade ago. The earlier technologies never resulted in successful products. Because of this, the field is now economically self-supporting; more money is put into it by the successful manufacturers.
The problem with Kurtzweil is that he's pre AI winter. His real work was in the 1980s. Since then, he's been a pundit, which becomes embarrassing if you do it for too long.
It's hard to fight Windows Update.
Neither Windows nor Linux has per-application compartmentalized security. In theory, you could use something like SELinux to give each vendor their own compartment, preventing an install from vendor A from affecting an install from vendor B. But the installers would have to be aware of this, and carefully stay in their own spaces, or installations would fail. Nobody does that.
(Someday, somebody is going to crack the signing key for Windows update, hijack a router to reroute Microsoft's IP address, and take over every Windows machine in the world.)
The document are out, and The New York Times is already reporting on the good stuff.
One of the more embarrassing items is this: American officials sharply warned Germany in 2007 not to enforce arrest warrants for Central Intelligence Agency officers involved in a bungled operation in which an innocent German citizen with the same name as a suspected militant was mistakenly kidnapped and held for months in Afghanistan. A senior American diplomat told a German official "that our intention was not to threaten Germany, but rather to urge that the German government weigh carefully at every step of the way the implications for relations with the U.S."
What seems to be driving the copyright enforcers nuts is simply that Syfert's package of boilerplate letters contains one raising a "personal jurisdiction" issue. The copyright enforcers have been filing all their suits in one Federal court, in the District of Columbia, regardless of where the defendant is. In such cases, where the defendant has no connection to the district where the court is, it's routine to object, and force the plaintiff to refile in the defendant's district court. The lawyers for the US Copyright Group then have to sue in dozens of different district courts. They hate that.
Here's an example of such a motion. Such motions are usually granted. The EFF has a long filing on this.
These days I'd wager that the vast (VAST!) majority of packets are part of ongoing streams - streaming Netflix over the net, torrenting the collected porn of the 80ies, that kind of thing.
True. However, many of those streams are bandwidth-adaptive and heavily buffered.
That's been known in the TCP community for decades.
I looked at this back in my RFC 896 days, when TCP was in initial development and I was working on congestion. I introduced the "congestion window" concept and put it in a TCP implementation (3COM's UNET, which predated Berkeley BSD). The question was, what should be the initial size of the congestion window? If it's small, you get "slow start"; if it's large, the sender can blast a big chunk of data at the receiver at start, up to the amount of buffering the receiver is advertising.
I decided back then to start with a big congestion window, because starting with a small one would slow down traffic even when bandwidth was available. One of the big performance issues back then was the time required to FTP a directory across a LAN, where TCP connections were being set up and torn down at a high rate. So startup time mattered. The decision to go with a smaller initial congestion window size came years later, from others. This reflected trends in router design. I wanted routers to have "fair queuing", so that sending lots of packets from one source didn't gain the sender any bandwidth over sending few packets. But routers gained speed faster than RAM costs dropped, and so faster routers couldn't have enough RAM for fair queuing. Today, your "last mile" CISCO router might have fair queuing. Some DOCSIS cable modem termination units have it. But many routers are running Random Early Drop, which is a simple but mediocre approach. (The backbone routers barely queue at all; if they can't forward something fast, they drop it. Network design tries to keep the congestion near the edges, where it can be dealt with.)
Remember, every dropped packet has to be retransmitted. (Too much of that leads to congestion collapse, a term I coined in 1984. That's what the "Nagle algorithm" is about.) In a world with packet-dropping routers, "slow start" makes sense. So that was put into TCP in the late 1980s (by which time I was out of networking.)
However, the RFC-documented slow start algorithm is rather conservative. RFC 2001 says to start at one maximum segment size. Microsoft's implementations in Win95 and later start at two maximum segment sizes. In RFC 3390, from 2002, the limit was raised to 3 or 4 maximum segment sizes. (We used to worry about delaying keystroke echo too much because big FTP packets were tying up the 9600 baud lines too long. We're past that.)
But Google is sending at least 8 segments at start, and Microsoft was observed to be sending 43. Sending 43 packets blind is definitely overdoing it.
I wonder whether they're doing this blindly, or if there's more smarts behind the scenes. If their TCP implementation kept a cache of recent final congestion window sizes by IP address, they could legitimately start off the next connection with the value from the last one. So, having discovered a path that's not dropping big bursts of packets, they could legitimately start fast. If they're just doing it the dumb way, starting fast every time, that's going to choke some part of the net under heavy load.
Why run this through Twitter? If the server wants to send an SMS message, it should just send an SMS message using an SMS gateway. Why package it as a "tweet?"
(I suspect why. So they can spam you. It's illegal to send unsolicited commercial SMS messages in the US. If PayPal makes you "follow" them on Twitter to get transaction confirmations, they can then send you ads, too.)
If the opposition is just entropy, it's not too bad. Active opposition is much more stressful. Lifeguards, firefighters, and EMTs tend not to be overly stressed. Cops and soldiers, though, routinely get stressed out.
What they actually did was to store about 100 bytes. This may be useful for putting copyright information into genetically engineered organisms. As a method of bulk data storage, though, it leaves much to be desired.
DNA synthesis costs about $0.29 per base pair. Sequencing is a bit cheaper, but you currently get less than 1000 base pairs sequenced per run. Reading and writing takes a room of expensive wet lab gear, and hours to days.
This predates MMO games. Pokemon ("Gotta Catch 'Em All") produced the same mindset. There were predecessors to Pokemon, but it was the first one to get huge. Arguably, Wizards of the Coast introduced this genre with "Magic, the Gathering" in 1990. Collectable cards have been around for about a century, but they were usually tied to real-world sports. Wizards of the Coast detached them from that world and made then stand alone.
Wizards of the Coast, however, managed not to be slimeballs. Zynga (Farmville, Mafia Wars, etc.) has a slimy history.
Their problem is that they've lost indexing data, not the underlying documents. So just make the documents, which are public records, visible to Google. Google will index them and anyone can then search.
is "Fully Loaded", by Bruce McCall.
(This is one of Bruce McCall's many drawings of dream cars of the 1959s that should have been.)
Bear in mind that a spacecraft launch and an ICBM launch look very similar, and a re-entry looks like an incoming missile. It's best if everybody knows where and when to expect such events, so that various military forces don't overreact. Both the normal scenarios and the abort plans need to be reviewed.