If I understand them right, they're claiming that integrated graphics and CPU/GPU hybrids are just a toy, and that you want discrete graphics if you're serious.
Ken Olsen famously said that "the PC is just a toy". When did you last use a "real" computer?
TCP is mostly controlled by round trip time measurement and window size. Response to packet loss is a backup mechanism.
That's not true. The currently deployed variants of TCP -- Reno, NewReno and SACK-TCP -- only use packet loss as a measure of congestion. ECN-enhanced variants only use packet loss and ECN marking.
There do exist experimental variants of TCP that use delay (TCP-Vegas comes to mind), but they are not widely deployed at this time.
Rather than taking the congestion as a given [...] I wonder if someone's done some research into why it exists and whether it's due to hot spots forming in the traffic flow.
Modern networking protocols are designed to use as much throughput as is available. Think of unused capacity as wasted capacity.
Let me make this clear on an example. Suppose that there's 10Mb/s free, and you're transferring a large file. Then you want your file transfer to go at a rate as close to the available 10Mb/s as possible. If it goes at 9.9Mb/s, then all is fine; but if it goes at 1Mb/s, then you're wasting 9Mb/s.
So while congestion is to be avoided, the network is designed to always stay at the very edge of congestion; a slight instability, and some routers will get congested.
To do anything based on flows, routers would have to keep track of all the active flows, which amounts to all open TCP connections going through that router.
Only if you want to be fair.
In practice, however, you only want to be approximately fair: to ensure that the grandmother can get her e-mail through even though there's a bunch of people on her network running Bittorrent. So in practice it is enough to keep track of just enough flow history to make sure that you're fair enough often enough, and no more.
A number of techniques have been developed to do that with very little memory; my favourite happens to be Stochastic Fair Blue.
Linux has a range of RED and BLUE implementions. Hold a contest at your local LUG or LAN Gamer's meets, to see who can set it up the best.
RED is tricky to set up, but neither Blue nor PI require much tuning, if any. (I'm running Blue on all of my 2.6 Linux routers, and RED on all the 2.4 ones.)
Flow management also includes ECN. Have you switched that on yet?
Yes, I have. On all of my hosts and routers. It's a big win for interactive connections, but doesn't matter that much for bulk throughput.
There's MTUs and window sizes to consider - default works fine most times, but do you understand those controls and when they should be used?
Unless you're running some fancy link technology, you don't get to tune your MTU. If, like most of us, you're running Ethernet and WiFi only, you're stuck with 1500 bytes.
As for window sizes, they're pretty much tuned automatically nowadays, at least if you're running a recent Linux or Windows Vista.
While approximately fair is not quite as good as fair, avoiding hard state makes for a cheaper and more reliable network, which allows you to more easily over-provision your link capacity. Unless, of course, you're in the business of selling routers that implement hard state...
My favorite lecturer quote, "Oh, I don't really do any coding at all"
That's not a big deal. Computer Science is not about coding in particular, but understanding the practices to design and implement solutions to a problem. Computer Science is more about applied math then writing in language X.
Computer Science is more than just applied maths. It's applied maths and how it relates to systems design and programming.
As the OP noted, a lot of CS academics have little programming experience, and this has some rather disturbing consequences. A lot of CS research is unrelated to real-world problems, since the researchers have not had the occasion to develop intuitions that would allow them to feel what is relevant. A lot of CS teaching is cold and abstract, and fails to interest the students, because the lecturers do not have enough programming culture to make their lectures live.
The best CS academics are people who manage to combine a solid theoretical background with a moderate amount of practical experience.
Patents are all about saying "only we can do this".
Says who?
Patents are about making sure that good inventions are disclosed before their inventor dies.
Patents as a business model -- the view that you take -- is a distortion of the patent system, and, as I mentioned before, the source of all of our troubles.
I think that any company that has patent and is not implementing them [...] should loose their patent
Patents should not be awarded to companies, only to individuals.
Historically, patent letters have been invented for a reason: to avoid a master craftsman carrying his trade secrets into the grave. The deal is simple: you get a monopoly on an invention until your death, and society as a whole benefits once you're no more. Simple, clean model.
Awarding patents to companies is a subversion of this model, and is the source of all of our troubles.
A lot of people think they need their ISP's help to get IPv6 connectivity. That's not the case. If you're running Windows Vista, or if you use an Apple Airport router, you should get connectivity to the IPv6 Internet out of the box. If you're running Linux, I've writtent a short HOWTO about IPv6 under Linux.
No, IPv5 is ST-II, the Internet Stream protocol, defined in RFC 1190. TUBA is one of the experimental proposals that led to IPv6, and has never been assigned an IP version number.
The real issue here is the morality of the fee. Those who are pirates download content worth significantly more than $5.
What you are raising is the issue of the morality of taxation. We pay taxes for education, whether we have children or not, because we believe that society as a whole benefits from schools.
a technologically impaired senior who wants to communicate with his children who live in another state/country will also have to pay.
Lanier is up to something, but he's missing the point. In the 1960s and 70s came up Tenex, Unix, the Smalltalk system, the Lisp Machine, the MESA system, full systems built from scratch.
Today, most of the software people, be it in Academia, in Industry or in the Free Software community, are designing building blocks -- pieces of software that are designed to fit within an existing system (Unix, Windows, the web, whatever).
The common theme is that it should be totally free and unregulated
The common theme is rather that we want more spectrum to be available to the general public without the restrictions of the amateur radio bands.
What Wifi has shown is that allowing parts of the spectrum to be used freely allows a lot of applications that would simply never happen on the licensed bands. Many of us would like to play with a smallish part of the 700MHz band.
spectrum ain't software so it's not like everyone can get a copy.
With spread-spectrum technologies, a lot of people can get a lot of fun out of a small amount of spectrum.
Slashdot Mirror
Look at it the other way: whenever you're not running a game, "the computer" takes some Video RAM and gives it back to the system...
If I understand them right, they're claiming that integrated graphics and CPU/GPU hybrids are just a toy, and that you want discrete graphics if you're serious. Ken Olsen famously said that "the PC is just a toy". When did you last use a "real" computer?
SCTP uses the same congestion control algorithm as TCP.
That's not true. The currently deployed variants of TCP -- Reno, NewReno and SACK-TCP -- only use packet loss as a measure of congestion. ECN-enhanced variants only use packet loss and ECN marking.
There do exist experimental variants of TCP that use delay (TCP-Vegas comes to mind), but they are not widely deployed at this time.
It's been done, and it's called Equal Cost Multi-Path.
Modern networking protocols are designed to use as much throughput as is available. Think of unused capacity as wasted capacity.
Let me make this clear on an example. Suppose that there's 10Mb/s free, and you're transferring a large file. Then you want your file transfer to go at a rate as close to the available 10Mb/s as possible. If it goes at 9.9Mb/s, then all is fine; but if it goes at 1Mb/s, then you're wasting 9Mb/s.
So while congestion is to be avoided, the network is designed to always stay at the very edge of congestion; a slight instability, and some routers will get congested.
Only if you want to be fair.
In practice, however, you only want to be approximately fair: to ensure that the grandmother can get her e-mail through even though there's a bunch of people on her network running Bittorrent. So in practice it is enough to keep track of just enough flow history to make sure that you're fair enough often enough, and no more.
A number of techniques have been developed to do that with very little memory; my favourite happens to be Stochastic Fair Blue.
RED is tricky to set up, but neither Blue nor PI require much tuning, if any. (I'm running Blue on all of my 2.6 Linux routers, and RED on all the 2.4 ones.)
Yes, I have. On all of my hosts and routers. It's a big win for interactive connections, but doesn't matter that much for bulk throughput.
Unless you're running some fancy link technology, you don't get to tune your MTU. If, like most of us, you're running Ethernet and WiFi only, you're stuck with 1500 bytes.
As for window sizes, they're pretty much tuned automatically nowadays, at least if you're running a recent Linux or Windows Vista.
While approximately fair is not quite as good as fair, avoiding hard state makes for a cheaper and more reliable network, which allows you to more easily over-provision your link capacity. Unless, of course, you're in the business of selling routers that implement hard state...
Computer Science is more than just applied maths. It's applied maths and how it relates to systems design and programming.
As the OP noted, a lot of CS academics have little programming experience, and this has some rather disturbing consequences. A lot of CS research is unrelated to real-world problems, since the researchers have not had the occasion to develop intuitions that would allow them to feel what is relevant. A lot of CS teaching is cold and abstract, and fails to interest the students, because the lecturers do not have enough programming culture to make their lectures live.
The best CS academics are people who manage to combine a solid theoretical background with a moderate amount of practical experience.
Says who?
Patents are about making sure that good inventions are disclosed before their inventor dies.
Patents as a business model -- the view that you take -- is a distortion of the patent system, and, as I mentioned before, the source of all of our troubles.
Patents should not be awarded to companies, only to individuals.
Historically, patent letters have been invented for a reason: to avoid a master craftsman carrying his trade secrets into the grave. The deal is simple: you get a monopoly on an invention until your death, and society as a whole benefits once you're no more. Simple, clean model.
Awarding patents to companies is a subversion of this model, and is the source of all of our troubles.
A lot of people think they need their ISP's help to get IPv6 connectivity. That's not the case. If you're running Windows Vista, or if you use an Apple Airport router, you should get connectivity to the IPv6 Internet out of the box. If you're running Linux, I've writtent a short HOWTO about IPv6 under Linux.
There are at least two implementation of Blue for Linux,
Blue
Stochastic Fair Blue
No, IPv5 is ST-II, the Internet Stream protocol, defined in RFC 1190. TUBA is one of the experimental proposals that led to IPv6, and has never been assigned an IP version number.
DSDV
OLSR
AODV
Babel.
How is this different from a tax on ADSL lines, then?
What you are raising is the issue of the morality of taxation. We pay taxes for education, whether we have children or not, because we believe that society as a whole benefits from schools.
That's why taxation is usually progressive.
How is this different from taxation, and subsidising culture from public funds?
You're late. RFC 4838 does just that.
Lanier is up to something, but he's missing the point. In the 1960s and 70s came up Tenex, Unix, the Smalltalk system, the Lisp Machine, the MESA system, full systems built from scratch.
Today, most of the software people, be it in Academia, in Industry or in the Free Software community, are designing building blocks -- pieces of software that are designed to fit within an existing system (Unix, Windows, the web, whatever).
Or, as Rob Pike put it, systems research is irrelevant (PDF).
The common theme is rather that we want more spectrum to be available to the general public without the restrictions of the amateur radio bands.
What Wifi has shown is that allowing parts of the spectrum to be used freely allows a lot of applications that would simply never happen on the licensed bands. Many of us would like to play with a smallish part of the 700MHz band.
With spread-spectrum technologies, a lot of people can get a lot of fun out of a small amount of spectrum.
So it still needs proprietary drivers, right?
Fair enough, I should have mentioned that a lot of Windows user-space software is still deficient in its IPv6 support.
As for the 10 minutes delay -- that's the first time I hear about it.
Microsoft have a perfectly functional IPv6 stack in XP SP2 and Vista.