The parent post brings up some very important points, and touches some very important issues. It does not deserve a Flamebait mod. At best, it is offtopic to the SCO/GPL/FSF discussion.
\In*ex"o*ra*ble\, a. [L. inexorabilis: cf. F. inexorable. See In- not, and Exorable, Adore.] Not to be persuaded or moved by entreaty or prayer; firm; determined; unyielding; unchangeable; inflexible; relentless; as, an inexorable prince or tyrant; an inexorable judge. ``Inexorable equality of laws.'' --Gibbon. ``Death's inexorable doom.'' --Dryden.
The article didn't seem to say much about the difficulty of developing a system whose modules all run in the same address space, perhaps because the project it describes seems like a one-man deal. An advantage of a Linux system is the ease with which userland apps can be developed and debugged, and consequently delegated even to programmers who don't have kernel hacking experience. This is nice in multi-person projects.
As for the comments about disadvantages of userland (latency, different interface to hardware, etc.), much of this can be avoided through correct design. For example: the memory-mapped registers can be exposed through a block device; dealing with hardware interrupts on time can be solved by factoring out the latency-sensitive part into a kernel driver, and access it via sockets. I guess the author is right about the embedded developers' mindset which finds it alien to separate drivers from the application:)
You're right, the RFC does mandate this, and it's reasonable for small servers. But if I were a mail provider serving millions of users, more than half of whose incoming mail consisted of spam, I would skip the strict standard conformance in favor of what makes sense.
Hotmail seems to receive more spam than other free email providers. I believe this may be due to how they handle recipient verification in SMTP. When a mail client attempts to send a message to an unknown username, the hotmail mail server will reply with an error message, indicating that the user doesn't exist. As a result, it is possible for a single spammer to spend some time just once to brute-force user names, and then distribute the list of known-good user names.
Yahoo generates the same reply regardless of whether the recipient exists or not. Thus, to guess user names, spammers would have to brute-force every mailing, as opposed to just the initial one like in the hotmail case.
Why hotmail would do something like this is completely beyond me.
I can't begin to express how much it amuses me when people who whine about privacy loss use credit cards for their transactions. The entertainment value of the irony is... priceless.
RFID tags on stuff are no different than UPC codes, as far as traceability of purchases is concerned. Walmart could tie your purchase history with your CC# as it is-- why do you think this will change with RFID?
If you can design a portable reader for passive RFID tags that works at greater than 10 meter distances through obstacles, I encourage you to patent it and start a company immediately-- you will make out like a bandit.
Let me give some more context to the original post. By "complex protocol design" I mainly think of the way classic telcos run things: e.g. run SS7 signaling, use expensive 5ESS switches, ATM, circuit switching, etc. In comparison, an IP network is much simpler, and cheaper; there are only datagrams forwarded through stateless routers. (Though to run a voice network, there will have to be some signalling protocol like SIP, and the VoIP traffic separated from best effort traffic).
The question is, why does a telco like Sprint switch its backbone to IP? What's the payoff?
No need to run two parallel networks (SS7 + IP), run everything over IP
Eliminate expensive telco switching equipment, replace with cheaper IP routers + VoIP equipment
I don't know what the real costs are of the two sides. Since there are companies switching to VoIP, I'm guessing there is a cost advantage.
Copyright isn't ownership of ideas. It's the exclusive right to make copies.
No other party is permitted to distribute copies of the blueprint. However, they can copy it to their heart's content, privately, or build the object contained in the blueprints on their own property. Remember fair use?
The problem is that it's much more expensive to engineer complex protocols that provide guaranteed qualities of service-- both in startup cost and maintenance in the long run-- than to just expand the pipes until the link utilization is low enough to make latency problems of IP disappear. It is a simple and stupid solution, there is no sexy protocol design that gets papers published, but it works well enough and is cheap.
One step further: buy a new hard drive every spring. Their capacity doubles every year, so you don't even need to back anything up, just copy to the new hard drive and stash the old one somewhere.
I wouldn't hold my breath for it. Using Bluetooth will require some serious battery power (at least, compared to what is ordinarily put into watches). USB is nice because it can power the device from the host (and perhaps even recharge the watch).
Unbeknownst to Gorshkov and Ivanov, the agents had installed onto the "company's" computers a program that logged the young men's keystrokes as they were accessing the tech.net.ru systems in Russia. That allowed U.S. law enforcement to obtain the hackers' passwords.
Um, no. TCP's congestion control doesn't get used in single-hop connections such as between nodes on a LAN. If there is overlap in the coverage of the 802.11 access points, they will share the air by nature of the collision avoidance mechanism in the MAC layer. If everyone is blasting UDP at full speed, they should get equal shares of the medium (for some definition of "equal," har har).
Recall the first Matrix, where Agent Smith rants about the humans' deficiencies, and mentions the failure of an early beta version of the Matrix. It failed because they made the simulation too good, and people were missing all the pain and suffering.
Slashdot Mirror
He reminds me of Rowan Atkinson in "Ratrace": "I'm weening, I'm weening!"
The parent post brings up some very important points, and touches some very important issues. It does not deserve a Flamebait mod. At best, it is offtopic to the SCO/GPL/FSF discussion.
Actually for audio discs, he found two car CD players that could play the entire 1GB disc flawlessly. One an OEM Nisan player, one a fancy AIWA.
man xwd
Not disputing the use of the word-- just pointing out what it means :D
inexorable
\In*ex"o*ra*ble\, a. [L. inexorabilis: cf. F. inexorable. See In- not, and Exorable, Adore.] Not to be persuaded or moved by entreaty or prayer; firm; determined; unyielding; unchangeable; inflexible; relentless; as, an inexorable prince or tyrant; an inexorable judge. ``Inexorable equality of laws.'' --Gibbon. ``Death's inexorable doom.'' --Dryden.
(courtesy dictionary.com)
Yes, but capuccinos have fans, and are damn noisy.
The article didn't seem to say much about the difficulty of developing a system whose modules all run in the same address space, perhaps because the project it describes seems like a one-man deal. An advantage of a Linux system is the ease with which userland apps can be developed and debugged, and consequently delegated even to programmers who don't have kernel hacking experience. This is nice in multi-person projects.
:)
As for the comments about disadvantages of userland (latency, different interface to hardware, etc.), much of this can be avoided through correct design. For example: the memory-mapped registers can be exposed through a block device; dealing with hardware interrupts on time can be solved by factoring out the latency-sensitive part into a kernel driver, and access it via sockets. I guess the author is right about the embedded developers' mindset which finds it alien to separate drivers from the application
No, no, no, that would help their company. Instead, we must all sell their stock short! ;)
You're right, the RFC does mandate this, and it's reasonable for small servers. But if I were a mail provider serving millions of users, more than half of whose incoming mail consisted of spam, I would skip the strict standard conformance in favor of what makes sense.
Hotmail seems to receive more spam than other free email providers. I believe this may be due to how they handle recipient verification in SMTP. When a mail client attempts to send a message to an unknown username, the hotmail mail server will reply with an error message, indicating that the user doesn't exist. As a result, it is possible for a single spammer to spend some time just once to brute-force user names, and then distribute the list of known-good user names.
Yahoo generates the same reply regardless of whether the recipient exists or not. Thus, to guess user names, spammers would have to brute-force every mailing, as opposed to just the initial one like in the hotmail case.
Why hotmail would do something like this is completely beyond me.
Let me give some more context to the original post. By "complex protocol design" I mainly think of the way classic telcos run things: e.g. run SS7 signaling, use expensive 5ESS switches, ATM, circuit switching, etc. In comparison, an IP network is much simpler, and cheaper; there are only datagrams forwarded through stateless routers. (Though to run a voice network, there will have to be some signalling protocol like SIP, and the VoIP traffic separated from best effort traffic).
The question is, why does a telco like Sprint switch its backbone to IP? What's the payoff?
I don't know what the real costs are of the two sides. Since there are companies switching to VoIP, I'm guessing there is a cost advantage.
Copyright isn't ownership of ideas. It's the exclusive right to make copies.
No other party is permitted to distribute copies of the blueprint. However, they can copy it to their heart's content, privately, or build the object contained in the blueprints on their own property. Remember fair use?
The problem is that it's much more expensive to engineer complex protocols that provide guaranteed qualities of service-- both in startup cost and maintenance in the long run-- than to just expand the pipes until the link utilization is low enough to make latency problems of IP disappear. It is a simple and stupid solution, there is no sexy protocol design that gets papers published, but it works well enough and is cheap.
Raiding Western Russia would be pure suicide; it's in Europe. Let's hope they don't miss the Moon.
See original article: Egg Troll's Guide to FPS Games
What fnord red colors are fnord you talking about? fnord I don't see fnord anything unusual with the story.
One step further: buy a new hard drive every spring. Their capacity doubles every year, so you don't even need to back anything up, just copy to the new hard drive and stash the old one somewhere.
I wouldn't hold my breath for it. Using Bluetooth will require some serious battery power (at least, compared to what is ordinarily put into watches). USB is nice because it can power the device from the host (and perhaps even recharge the watch).
I like this snippet:
Unbeknownst to Gorshkov and Ivanov, the agents had installed onto the "company's" computers a program that logged the young men's keystrokes as they were accessing the tech.net.ru systems in Russia. That allowed U.S. law enforcement to obtain the hackers' passwords.
0wned by FBI's keylogger, har har!
Problem 1: UDP and Congestion
Um, no. TCP's congestion control doesn't get used in single-hop connections such as between nodes on a LAN. If there is overlap in the coverage of the 802.11 access points, they will share the air by nature of the collision avoidance mechanism in the MAC layer. If everyone is blasting UDP at full speed, they should get equal shares of the medium (for some definition of "equal," har har).
They still won't fit into 802.11G, but not by the factor that you suggest.
Um, no. 802.11g has a raw bitrate of 54 Mbps, which should give around 30Mbps effective TCP throughput. Plenty of bandwidth even for MPEG2.
And I can already stream divx over 802.11b, it only takes 2 Mbps or so for the standard 700 MB movies.
Recall the first Matrix, where Agent Smith rants about the humans' deficiencies, and mentions the failure of an early beta version of the Matrix. It failed because they made the simulation too good, and people were missing all the pain and suffering.
So they put IPv4 back in.
Hey, neat idea. Let's see what they've got:
The tunnel is driven by six 40-foot diameter fans that are powered by six 22,500 horesepower motors.
22,500 horsepower... now that, my friends, is what I call a fan!