Remus Shepherd asked: Could scientific research/teaching/politics/other endeavours benefit from a 'Bazaar' approach of distributed design?
Let's not flatter ourselves too much. The scientific community don't need to learn Open Source methods from us, because they taught them to us in the first place. Open Source is essentially the application to software of the open publication/peer review methods that have been so successful in the scientific community for centuries.
Much of ESR's writing has been about the internal dynamics of the OSS community. Well and good, but the world is wide and non-OSS influences tend to dominate. Howcome so few comments on the role that the anti-OSS camp have to play?
My own take is that the coopetitive nature of OSS makes their attacks (Mindcraft, Dvorak, etc.) just another form of peer review -- which is what makes OSS thrive. That said, I'd rather see ESR's analysis.
I wonder if they'll offer upgrades to people using the older (Brand M) version. Not just because it would be nice to get some benefit from the shelfware that we bought from them, but because the statistics will be VERY enlightening.
Face it -- paper is universal. Every time some clueless twit sends out a broadcast e-mail in Word2000 format we all just send it over to our AA, who faithfully kills a few trees so that we can read it and discover that it's an announcement that someone we've never heard of will be out of office next Tuesday.
It's not accidental that two of the top industries in the State of Washington are lumber and MICROS~1
TurkishGeek asked: Thanks for the useful information. If you don't mind, may I ask you what kind of work with memory manufacturers you are involved in? Based on some of your previous posts, I guess you are more involved in process technology rather..
I'm an I/O designer and signal-integrity engineer with a semiconductor company (thus the handle); I help out with some of the JEDEC memory standards work and the memory manufacturers there.
BTW, I had lunch today with an engineer from Intel working on this problem. He says it comes from crosstalk among the data lines and (although he was diplomatic and didn't say so directly) appears to be inherent in Rambus rather than being the Camino chip per se.
Red Herring wrote: I heard someone else mention DDR SDRAM... this may work in a server, but it requires a huge amount of power, which in turn means cooling, which in turn means cost. It's not terribly suitable for a general desktop. RDRAM manages power better.
Interesting observation, but I'm curious where you get the power comparison. RDRAM certainly has more power-management features than DDR but that doesn't mean it uses less power. The proof would be in actual power-usage comparisons.
FWIW SSTL-2 drivers run about 8 mA/bit during active signaling, for a total across 64 bit memory of about 512 mA. RDRAM burns 25mA per bit per node times 16 bits times 2-4 nodes for a total signaling current of 800-1600 mA.
WRT on-chip power the memory arrays, sense amps, etc are all standard in both cases. There is a difference in the DLL (DDR's DLL can be run without static power) and I/O circuits (Rambus runs current-mode signaling and therefore has some analog circuitry with attendant static currents in the RAC.)
Rambus is a parallel architecture running at serial-transfer data rates. All of the data lines have to arrive together on the same clock edge or they get confused with another word. Intel and the memory shops have specs on how much skew each part is allowed to introduce and how much each is required to tolerate, and what's left is the budget for the motherboard. The MB manufacturers build their motherboards to spec but now it turns out that the Camino needs a little more "eye opening" (data good window) than was in the budget. Since even empty slots cause some reduction in eye opening, the situation isn't improved by plugging a blank in the empties (Rambus is a ring topology, so you can't have totally empty slots.) In fact, a full slot might improve matters by resyncing everything.
A Rambus channel is 16 bits wide and transfers data on both edges of a clock, so a 400 MHz Rambus memory system (what Intel calls 800 MHz for marketing reasons) transfers 1600 MB/s, although with some delay for packet processing. The memory manufacturers couldn't get yield at that speed and wanted 600, but the system houses said they couldn't sell 600; Intel compromised on a hair over 700. Call it 1400 MB/s
For comparison, a 64-bit PC133 DIMM transfers 1066 MB/s with less latency. A PC200 DDR DIMM transfers 1600 MB/s with latency a hair less than the PC133, and a PC266 DDR DIMM transfers 2133 MB/s at about the same latency as the PC200.
TurkishGeek wrote: But in overall production costs, ignoring the price premiums tacked on the price by companies, RDRAM is more cost-effective. That is, RDRAM is actually cheaper to produce.This alone makes it attractive in the long run.
Dunno where you get this. I work with memory manufacturers and the lowest quoted die-size penalty for Rambus is 20% over DDR SDRAM. If you talk to the engineers instead of the spokesuits the number is more like 40%, and the yield is quite a bit lower.
Rambus advocates will argue over the amounts of the price premium and the performance changes relative to PC100 SDRAM. Based on whether you highball the performance and lowball the price adder or highball the price adder and lowball the performance you can either conclude that RDRAM is the inevitable next mainstream memory or that it's DOA, but nobody in the trade is claiming that RDRAM is cheaper to produce than SDRAM or DDR.
Copper atoms are heavier than aluminum atoms and thus get kicked around less by high current densities (electromigration). Seriously! On-chip current densities are so high that a serious failure mode is from the wires flowing downstream.
WRT resistance, in the last few years signals have been suffering quite a bit of slowing due to the resistive delay introduced by wiring resistance and capacitance. In GHz processors this actually becomes a limiting effect.
Transient supply drops are so serious that they either degrade logic timing or in extreme cases flip flops. Lower supply resistance helps a lot.
As for wanting gold interconnects, no you don't. For one thing, copper is a better conductor than gold. Besides that, gold is a disaster in silicon processing (it diffuses like lightning and scavenges carriers. Low transconductance and high leakage everywhere = slow and hot.)
Howzabout that? Linus gets quoted on the upcoming devalutation of software the same afternoon that Ballmer admits that Microsoft stock is overpriced. Market tumbles.
Just for grins, imagine that Judge Jackson finds against Microsoft and fines them $1.00 (that's one dollar US, no change)
Pause.
Savor for the moment the dilemma that that would put Bill in. He has, for all practical purposes, won. Certainly the headlines will read that way. And yet he'll have to appeal.
Have a look at OS sales and one thing should be very, very clear: end-users aren't driving this bus. Total retail OS sales are 'way down in the noise because the OS that runs on nearly every machine is the one loaded onto it at the factory. The market for operating systems is the OEM preload market. Now in that context, how much competition is there? Ask yourself what would happen if Microsoft had told (e.g.) Compaq in 1996 that they could only have Windows at full retail price. Could Compaq have substituted a "competitor's" operating system? If you look at the actual dynamics of the industry, end-users aren't the consumers. As with TV, they're the product. Microsoft offers Dell, Compaq, and the rest access to us the same way that NBC delivers consumers to General Motors.
Jburkholder wrote: Geez, too bad the NSA doesn't have anything better to do than spy on average computer hackers. Guess they got tired of interecpting everyone's email and following foriegn nationals around watching for them to rent u-hauls and buy fertilizer.
Paranoia aside, there are a number of reasons to be upset about the NSA installing backdoors.
They've already been caught reading supposedly-secure commercial mail and passing the fruits to commercial competitors. They may not be interested in your sex life, but do you want to have your business available to your competitors?
A large part of the security of encryption lies in the fact that any given key is of limited value -- you can only read one person's mail, etc. If that key is a backdoor to every WinNT box on the planet, though, the picture changes somewhat. In the end a security hole is a security hole.
Keep in mind that the NSA is first and foremost a political entity. RIght now they're trying to convince Congress to pour lots more money and power their way based on the supposed insecurity of computers in the USA. Their chances of success are obviously improved by reduced computer security.
Cptn Proton wrote: That's not the only grind we should have. When GATT was signed, it automatically extended the life of patents from 17 years to 20 years, thus amending the constitution in unconstitutional ways. Why has this not been addressed by the supreme court???
That would make so-called 1989 software patents available in six instead of eight years.
A couple of misconceptions here. First, as others have noted, there is no Constitutional requirement that patents lapse in seventeen years. The only Constitutional mandate wrt the patent system is that there must be one. As for the other poster who questioned the treaty-vs-legislation mechanism, the point is moot since
treaties ratified by Congress have the force of law, and
Congress revised the statutory law to make it consistent with the Berne Convention (not GATT)
Secondly, the life of a patent has not been extended to twenty years from seventeen. What has happened is that an additional limitation has been added to cause patents to expire twenyt years after the date of filing if that is sooner than seventeen years after the date of issuance. This not only harmonizes the USA with the rest of the world, but also prevents the "submarine patent." (A recent example was the character who filed back in the 50s for a patent on optical storage of information. His lawyers kept it tied up in the USPTO for decades while they continually updated the claims, so that when it issued recently it covered CD media!)
Disclaimer: I am not a lawyer, don't play one on TV, nor on the Net. I do however hold several patents and retain patent attorneys who keep me up to date on details like this.
Preemptive publication. Large companies (e.g. IBM) routinely use this method to protect themselves from later infringement suits when they don't want to spend the time and money to patent the invention. The drawback is that it only works to protect the inventor once the lawsuit is filed, although a smart plaintiff will back off real fast due to the chance of a ticked-off judge imposing sanctions. This does not keep the plaintiff from pulling the same stunt with someone else the next day.
Publication is of absolutely no help in preventing patents. For all I know someone is filing a patent on Quicksort as we read this, and if they do the USPTO is quite likely to grant it. Once granted, it's almost impossible to overturn a patent, and as long as it stands those letters will keep going out.
The fundamental problem is that the examiners in the USPTO only check against the existing patent database when considering an application. Up until the Eighties they didn't allow software patents at all but then a bright lawyer found a way to file an application worded such that the algorithm was implemented in hardware or any equivalent and the USPTO got their noses rubbed in the fact that hardware and software are interchangeable. So they started accepting software patent applications -- a virgin field. With nothing in the prior art database that they could point to in turning down the application, they just granted them all.
Keep that point about the equivalence of hardware and software in mind. Scream as we will, we are not going to make software patents go away, because the USPTO didn't want them in the first place and was only convinced by the USSC and mathematics. Neither of which are particularly influenced by whining.
Our only hope for straightening out this mess is to get the Congress to insist on examiners with at least passing familiarity with the field. Which isn't going to be possible when an NCG makes more money in industry than an examiner can ever hope to, and in industry she won't have to live in the DC cesspit.
Goody suggested: Get someone like the people who make VMWare to get Win9* to run on Alpha and start packaging RedHat on Alpha machines.
VMWare? Why bother? Compaq now owns the original translate-on-the-fly Windows emulator, DEC's excellent FX!32. All others are pale imitations.
The trouble is that DEC had a sever case of craniorectal insertion (I know, I worked with them on two horribly mismanaged Alpha projects.) To this day Compaq tries to charge more for DEC Unix than a high-end workstation hardware costs. They just don't grasp the importance of GIVING AWAY the FX!32 software as the only way to bootstrap their hardware sales.
Of course, if the reverse-engineering is done in Australia (or for that matter in a State that neglects to follow UCITA on reverse-engineering) and the specs are posted to the Net, it becomes a First Amendment issue instead of contract law.
How sad.
Also, as a practical matter, it might be difficult to sue hundreds of anonymous contributors who just might have done the RE work even if they are in UCITA states, especially if the gatekeeper is in Australia (etc). Australian courts might be a tad reluctant to issue subpoenas against Australian citizens who don't want to cooperate with a US civil action over activities explicitly legal in Australia, and US courts don't have jurisdiction.
I suppose it's nice that there's a Notes server for Linux but it's not much comfort, since I'd have to replace the Linux and Solaris desktop machines with LoseDoze machines to use it.
Notes is a touchy subject right now since the employer is switching over to it and that means no more mail on the Solaris workstation. Somehow having all the engineers share an NT box for e-mail is seen as a productivity advance.
Many companies have found that one of the Net's biggest benefits is its ability to give them high bendwidth customer feedback (eg auto makers with model selector pages, which tell them what features people want far better than post-hoc sales analyses.) As usual, the political sector is lagging by a good bit, with most pols acting like the Net is another one-way pipe like television.
Examples: most Congressional offices either have no e-mail address at all or/dev/null it; wide experience tells us that your chance of having your mail read by a human is zero. Perhaps more frightening are the suggestions that the Y2K elections will be massively spamvertised.
After a trial spam or two from the RNC in 1996, several pols tried spamming in 1998 and ran into the (predictable) consequence: spam alienates several nines of the people it reaches. Which might be an acceptable consequence to the spammer if the worst that happens is that they don't send ten dollars to the top five names on the list, but doesn't look so good if the recipients vote.
So, with Bush and Gore setting up massmail pyramids, the RNC hiring spammers, McCain sneaking prospam legislation through at midnight before floor votes, and Congress extending the franking priveledge to spam accounts, the question is:
How do we get a clue-by-four to our Lords and Masters that the Net is much better for information gathering than propaganda pushing?
Slashdot Mirror
Remus Shepherd asked:
Could scientific research/teaching/politics/other endeavours benefit from a 'Bazaar' approach of distributed design?
Let's not flatter ourselves too much. The scientific community don't need to learn Open Source methods from us, because they taught them to us in the first place. Open Source is essentially the application to software of the open publication/peer review methods that have been so successful in the scientific community for centuries.
Much of ESR's writing has been about the internal dynamics of the OSS community. Well and good, but the world is wide and non-OSS influences tend to dominate. Howcome so few comments on the role that the anti-OSS camp have to play?
My own take is that the coopetitive nature of OSS makes their attacks (Mindcraft, Dvorak, etc.) just another form of peer review -- which is what makes OSS thrive. That said, I'd rather see ESR's analysis.
I wonder if they'll offer upgrades to people using the older (Brand M) version. Not just because it would be nice to get some benefit from the shelfware that we bought from them, but because the statistics will be VERY enlightening.
Face it -- paper is universal. Every time some clueless twit sends out a broadcast e-mail in Word2000 format we all just send it over to our AA, who faithfully kills a few trees so that we can read it and discover that it's an announcement that someone we've never heard of will be out of office next Tuesday.
It's not accidental that two of the top industries in the State of Washington are lumber and MICROS~1
TINLC
Dan Huffnal was right. It's all because of the Lumber Cartel
TurkishGeek asked:
Thanks for the useful information. If you don't mind, may I ask you what kind of work with memory manufacturers you are involved in? Based on some of your previous posts, I guess you are more involved in process technology rather..
I'm an I/O designer and signal-integrity engineer with a semiconductor company (thus the handle); I help out with some of the JEDEC memory standards work and the memory manufacturers there.
BTW, I had lunch today with an engineer from Intel working on this problem. He says it comes from crosstalk among the data lines and (although he was diplomatic and didn't say so directly) appears to be inherent in Rambus rather than being the Camino chip per se.
Red Herring wrote:
I heard someone else mention DDR SDRAM... this may work in a server, but it requires a huge amount of power, which in turn means cooling, which in turn means cost. It's not terribly suitable for a general desktop. RDRAM manages power better.
Interesting observation, but I'm curious where you get the power comparison. RDRAM certainly has more power-management features than DDR but that doesn't mean it uses less power. The proof would be in actual power-usage comparisons.
FWIW SSTL-2 drivers run about 8 mA/bit during active signaling, for a total across 64 bit memory of about 512 mA. RDRAM burns 25mA per bit per node times 16 bits times 2-4 nodes for a total signaling current of 800-1600 mA.
WRT on-chip power the memory arrays, sense amps, etc are all standard in both cases. There is a difference in the DLL (DDR's DLL can be run without static power) and I/O circuits (Rambus runs current-mode signaling and therefore has some analog circuitry with attendant static currents in the RAC.)
Rambus is a parallel architecture running at serial-transfer data rates. All of the data lines have to arrive together on the same clock edge or they get confused with another word. Intel and the memory shops have specs on how much skew each part is allowed to introduce and how much each is required to tolerate, and what's left is the budget for the motherboard.
The MB manufacturers build their motherboards to spec but now it turns out that the Camino needs a little more "eye opening" (data good window) than was in the budget. Since even empty slots cause some reduction in eye opening, the situation isn't improved by plugging a blank in the empties (Rambus is a ring topology, so you can't have totally empty slots.) In fact, a full slot might improve matters by resyncing everything.
A Rambus channel is 16 bits wide and transfers data on both edges of a clock, so a 400 MHz Rambus memory system (what Intel calls 800 MHz for marketing reasons) transfers 1600 MB/s, although with some delay for packet processing. The memory manufacturers couldn't get yield at that speed and wanted 600, but the system houses said they couldn't sell 600; Intel compromised on a hair over 700. Call it 1400 MB/s
For comparison, a 64-bit PC133 DIMM transfers 1066 MB/s with less latency. A PC200 DDR DIMM transfers 1600 MB/s with latency a hair less than the PC133, and a PC266 DDR DIMM transfers 2133 MB/s at about the same latency as the PC200.
HTH. HAND.
TurkishGeek wrote:
But in overall production costs, ignoring the price premiums tacked on the price by companies, RDRAM is more cost-effective. That is, RDRAM is actually cheaper to produce.This alone makes it attractive in the long run.
Dunno where you get this. I work with memory manufacturers and the lowest quoted die-size penalty for Rambus is 20% over DDR SDRAM. If you talk to the engineers instead of the spokesuits the number is more like 40%, and the yield is quite a bit lower.
Rambus advocates will argue over the amounts of the price premium and the performance changes relative to PC100 SDRAM. Based on whether you highball the performance and lowball the price adder or highball the price adder and lowball the performance you can either conclude that RDRAM is the inevitable next mainstream memory or that it's DOA, but nobody in the trade is claiming that RDRAM is cheaper to produce than SDRAM or DDR.
As for wanting gold interconnects, no you don't. For one thing, copper is a better conductor than gold. Besides that, gold is a disaster in silicon processing (it diffuses like lightning and scavenges carriers. Low transconductance and high leakage everywhere = slow and hot.)
Howzabout that? Linus gets quoted on the upcoming devalutation of software the same afternoon that Ballmer admits that Microsoft stock is overpriced. Market tumbles.
Just for grins, imagine that Judge Jackson finds against Microsoft and fines them $1.00 (that's one dollar US, no change)
Pause.
Savor for the moment the dilemma that that would put Bill in. He has, for all practical purposes, won. Certainly the headlines will read that way. And yet he'll have to appeal.
Have a look at OS sales and one thing should be very, very clear: end-users aren't driving this bus. Total retail OS sales are 'way down in the noise because the OS that runs on nearly every machine is the one loaded onto it at the factory.
The market for operating systems is the OEM preload market.
Now in that context, how much competition is there? Ask yourself what would happen if Microsoft had told (e.g.) Compaq in 1996 that they could only have Windows at full retail price. Could Compaq have substituted a "competitor's" operating system?
If you look at the actual dynamics of the industry, end-users aren't the consumers. As with TV, they're the product. Microsoft offers Dell, Compaq, and the rest access to us the same way that NBC delivers consumers to General Motors.
Geez, too bad the NSA doesn't have anything better to do than spy on average computer hackers. Guess they got tired of interecpting everyone's email and following foriegn nationals around watching for them to rent u-hauls and buy fertilizer.
Paranoia aside, there are a number of reasons to be upset about the NSA installing backdoors.
That's not the only grind we should have. When GATT was signed, it automatically extended the life of patents from 17 years to 20 years, thus amending the constitution in unconstitutional ways. Why has this not been addressed by the supreme court???
That would make so-called 1989 software patents available in six instead of eight years.
A couple of misconceptions here. First, as others have noted, there is no Constitutional requirement that patents lapse in seventeen years. The only Constitutional mandate wrt the patent system is that there must be one. As for the other poster who questioned the treaty-vs-legislation mechanism, the point is moot since
Secondly, the life of a patent has not been extended to twenty years from seventeen. What has happened is that an additional limitation has been added to cause patents to expire twenyt years after the date of filing if that is sooner than seventeen years after the date of issuance. This not only harmonizes the USA with the rest of the world, but also prevents the "submarine patent." (A recent example was the character who filed back in the 50s for a patent on optical storage of information. His lawyers kept it tied up in the USPTO for decades while they continually updated the claims, so that when it issued recently it covered CD media!)
Disclaimer: I am not a lawyer, don't play one on TV, nor on the Net. I do however hold several patents and retain patent attorneys who keep me up to date on details like this.
Goody suggested:
Get someone like the people who make VMWare to get Win9* to run on Alpha and start packaging RedHat on Alpha machines.
VMWare? Why bother? Compaq now owns the original translate-on-the-fly Windows emulator, DEC's excellent FX!32. All others are pale imitations.
The trouble is that DEC had a sever case of craniorectal insertion (I know, I worked with them on two horribly mismanaged Alpha projects.) To this day Compaq tries to charge more for DEC Unix than a high-end workstation hardware costs. They just don't grasp the importance of GIVING AWAY the FX!32 software as the only way to bootstrap their hardware sales.
Tragedy. In the Greek sense.
Of course they did. Read the news out of Judge Jackson's courtroom.
My Little Pony wrote:
Any what do you do if your traps go off? Take the server off line? Can you say "Denial of Service"?
No, but any IP caught scanning closed ports goes into the deny table for a while. ipchains is SOOOO handy that way.
Of course, if the reverse-engineering is done in Australia (or for that matter in a State that neglects to follow UCITA on reverse-engineering) and the specs are posted to the Net, it becomes a First Amendment issue instead of contract law.
How sad.
Also, as a practical matter, it might be difficult to sue hundreds of anonymous contributors who just might have done the RE work even if they are in UCITA states, especially if the gatekeeper is in Australia (etc). Australian courts might be a tad reluctant to issue subpoenas against Australian citizens who don't want to cooperate with a US civil action over activities explicitly legal in Australia, and US courts don't have jurisdiction.
Dang! Try harder next time, Bill.
Don't you think you laid the parody on just a little too thick? A lighter touch would have made for a better troll.
I kinda doubt it's evolution, but something is amyth in Kansas.
I suppose it's nice that there's a Notes server for Linux but it's not much comfort, since I'd have to replace the Linux and Solaris desktop machines with LoseDoze machines to use it.
Notes is a touchy subject right now since the employer is switching over to it and that means no more mail on the Solaris workstation. Somehow having all the engineers share an NT box for e-mail is seen as a productivity advance.
Many companies have found that one of the Net's biggest benefits is its ability to give them high bendwidth customer feedback (eg auto makers with model selector pages, which tell them what features people want far better than post-hoc sales analyses.) As usual, the political sector is lagging by a good bit, with most pols acting like the Net is another one-way pipe like television.
/dev/null it; wide experience tells us that your chance of having your mail read by a human is zero. Perhaps more frightening are the suggestions that the Y2K elections will be massively spamvertised.
Examples: most Congressional offices either have no e-mail address at all or
After a trial spam or two from the RNC in 1996, several pols tried spamming in 1998 and ran into the (predictable) consequence: spam alienates several nines of the people it reaches. Which might be an acceptable consequence to the spammer if the worst that happens is that they don't send ten dollars to the top five names on the list, but doesn't look so good if the recipients vote.
So, with Bush and Gore setting up massmail pyramids, the RNC hiring spammers, McCain sneaking prospam legislation through at midnight before floor votes, and Congress extending the franking priveledge to spam accounts, the question is:
How do we get a clue-by-four to our Lords and Masters that the Net is much better for information gathering than propaganda pushing?