Re:Original P6 (PPro) was good but Windows wasn't
on
Intel Undercuts AMD
·
· Score: 1
The PPro's fatal flaw was that it could not execute 16-bit legacy code without being crippled. Intel fscked up by assuming that this would force *everyone*, including Microsoft, to develop and purchase software that no longer used that legacy code. It did not work out that way.
The flawed design was only purchased for the small segment of the market which was 32-bit, such as NT and *nix, so it became Intel's biggest flop since the iAPX432 (it was in desperation after that previous debacle that Intel as quickly as they could hacked together a strange, asymmetrical, segmented design out of their 8080/8085 CPU called the 8086).
So, Intel resorted to incremental changes while they fixed the problems with the PPro, the Pentium MMX processors. And AMD gained market share by using the opportunity. Microsoft traded off making Windows 95 compatable with legacy applications for what they might have gained by going all 32-bit and making Intel happy. They are not partners, each company has to respect what the market share of the other means to their own profits but it is each companies' duty to maximize their own for their stockholders.
So AMD managed to build market share and compete when Intel was knocked back, and all the other observations in the rest of your comment.:-)
Well, the Harvard authorities are reportedly returning the site. Presumably it will be restored, somewhere. Past experience says that the controversy will be resolved in and by the glare of publicity and concentrated attention.
One credible? The other? Both, despite specific disagreements and rivalries? Neither?
Oddly, I remember India taking a moral stand against such tests done on other animals, and prohibiting their export to the USA because of very similar experiments.
Perhaps now that they have their own nukes they've changed their mind some?;-)
Well, of course! But it would be morally bankrupt not to warn unwary potential consumers as well. Also currently at issue is @Home's resistance to sharing their outer network and allowing consumers to pick which ISP to connect directly to at the cable aggregation points - the quality of service provided, and treatment of IP customers, relates directly to that issue. Which in turn relates to the ability to cancel the service and go with someone else.
It limits free speech immensely - that is part of @Home's business model. The point is, a subscriber just can't *go* anywhere on the Internet except in the "audience" direction, if they use a network provider that bluntly forbids or cripples servers. The difference in capability is like the difference between a cheap transistor radio and a cellular phone.
Think about it - if Microsoft would, for example, by default enable their Personal Web Server during Win9x/20xx's install/OEM customer setup; give them the option to walk through one of those pick-a-color-and-layout-fill-in-the-blanks whizzards; and give them a few examples of what they can do - then @home and other ISPs that overcensor servers would be totally screwed. Tens, perhaps hundreds of thousands of angry customers, all demanding that this standard feature work, so at least a few people can see their vanity page; so their artistic little Jennifer's friends and the rest of the world can see the webcam on her special page - and the database of older images she's learned to build and web-enable - and later, so she can do that from the Linux server she's learned to run. That the WWW cache be turned on in both directions. And so on. Someone pointed out earlier that @Home's server prohibition does not extend to running NetMeeting. Can anyone or everyone guess why?:)
Unfortunately Microsoft probably won't do that, because 90 percent of the users would set their home page to their home page instead of MSN's "Start page" - they want to advertise as much as @Home does. Well, maybe not quite as much - from @Home's anti-open-access propaganda:
"If [...] cable companies were required to strip away the potential for generating advertising and commerce revenues by pursuing an access-only model, this would lead to an increase in the price of cable modem services. That isn't in the interest of the consumer."
Unless those consumers choose among the rest of the ISPs as well as @Home, and competition does its usual dance on prices.:)
So. Under @Home's short-sighted business model, everything that does not issue their advertising is abuse. Any of their own customers who issue their own data - yes, those paying for IP service - are treated as competitors! I suspect many ISPs who covet their customer's outbound bandwidth for the ISP's own WWW service do the same.
There aren't many businesses that could get away with that kind of crap treatment of their own clients, and there isn't any reason IP purchasers should have to put up with it.
I feel three-quarters ready to create a Slash engine site for small server questions, instructions, tips, announcements, and so on. Any comment on that idea? alone. Yup, including
With copper, was a nightmare due to ground loops. Ground one end, the other, both, neither, the current flow kept things from working.
I had to go to two AUI-to-fiber adapters I was lucky enough to find at a low but still painful price. With only optical fiber and no electrical connections things were fine, of course.
I suppose it takes considerably less infrastructure than the providing of a half-million of the internet's connections from smaller companies, given that this involves also receiving payment from all those half-million people. The overall bandwidth cost per user would be lower, because statistics guarantee that the more users, the lower the bandwidth needed for each. Sharing infrastructure costs with cable television, software taking over tasks that would be done by humans at smaller companies, and standard economies of scale in purchasing, provisioning, and other areas should make up for much of the higher bandwidth that is briefly needed for those quick page updates.
The rate cap is a cheap, easy, and rotten solution. Even supposing it isn't intended to screw up not only the "warez monkey" on the block but the legitimate user as well, it couldn't have been better designed to.
I hope you haven't been taken in by @Home's fantasy that they can cache the internet, or otherwise, without actually buying significant peering bandwidth, provide enough content to be other than a "half-assed experience" compared to that available to those with full IP connections. The idea is absurd - ultimately there is going to be about as much sucked from one part of the Internet as another.
The next "killer app" that emerges to equal importance with email and the WWW could very well need as much upstream as downstream bandwidth. Interpersonal multimedia communications tend to be two way, for example. The choice between abandoning new technologies and adding more bandwidth should be a no-brainer.
It will be years, yes, until T1 speeds are available for that price. Continuously. But there isn't any more reason not to allow fast uploads briefly than there is not to allow downloads at those speeds for the same amount of time - even if just a few persons use the vast majority of the upstream bandwidth, there should be more than enough remaining for the relatively small upstream needs of the rest of the users.
As for whether the servers might or might not feature MP3 or warez, or pet pictures, or whatever a person might feel should somehow have less priority than their own oh-so-important data needs, I choose to sayeth and care not. Other people's data is theirs; what I read and serve is mine.
Well, he's right about that being "the type of user @Home is looking for". In fact, I gain the impression that they would prefer would:
Buy the cable modem subscription for status, or because they are told it is "good". Be in the wide range between too computer-naive to effectively use a computer's capabilities on the internet and not so naive that they cannot run a Windows 98 box. Not have had a previous decent internet connection to apply the advertised speed multipliers to. Become quickly bored with the Internet as @Home Would Have It Be and thus condescends to present it to them, but keep the account to read email with once a month.
As for the credibility of his sociology, I simply suggest the AC author of the previous response look at the Magellan Search Voyeur at http://voyeur.mckinley.com/cgi-bin/voyeur.cgi - unless their personal ethics forbid reading other than easily cachable static pages.
He has a T1. Max speed: 1544000 bits/s. His audio channels use "128k". at least 128000bits/s. 12 of these could optimally fit in the T1. He is serving the internet needs of up to 350 people. This is an amplification rate of 29 times the bandwidth from his net to the multicast servers! And you think *that* 29:1 ratio is unacceptable?
It's making me feel better about my 128K ISDN connection. If my dedicated connection is going to cost three times what a cable modem does, then I want it to work as well at least.:-)
Exactly! And for a connection going in the opposite direction, trading the server and client end, just the opposite condition prevails. So it all balances out over the whole Internet, get it?
5. Don't try to put in a set of water mains and promote water use beyond what your sewer system can handle, even if you can make more selling water than charging for processing sewage. And if you do make that mistake, don't bitch that your customers are overloading the "outgoing" channel.
In large, and despite fantasies of caching sheeplike subscribers into a desired few web pages, there is as much incoming John Doe as there is outgoing John Doe. The packets don't just appear on the net magically - where are those 10, 100, etc John Does going to get their data, otherwise - while probably *not* themselves using much outgoing bandwidth, and thus freeing their own net's outgoing bandwidth up?
@Home has X amount of incoming network bandwidth. You can bet your bottom stack location @Home doesn't use ADSL to its peers - it has X outgoing bandwidth wired up too.
Caching servers don't deal well with dynamically built pages (like the one you're reading), responses to POSTed data for forms, or pages not in the ten percent of URLs requested ninety percent of the time (or whatever percent rule web cache servers depend on). These have to come off of the Internet. Unless, of course, the page requested is already *on* an @Home subscriber's server.. oops, that's out.
Okay. The data is coming from somewhere. and going to somewhere. "Flash Crowds" are likely to occur, tying up networking to the locations doing the serving, and network-adjacent sites. This means that - gosh - those people aren't tying up other sites!
I'm not saying someone should be allowed to operate a site that continuously serves at megabit data rates to a cable modem. But @Home is the one that is "leeching" if it feeds off the Internet's data without providing a channel to feed as much data to the net, and trying to cleave the internet into an unnatural division between "subscriber" and "content provider". To the extent that pages cannot be served at decent rates even to a few users, despite @home having *more* outgoing than incoming bandwidth free - the proportion of pages cached is not going to be anywhere *near* the ratio of outgoing to incoming packet bytes from apps like web browsers.
The solution to the problem: An ISP should not be so greedy as to *combine* caching and selling off even the limited outgoing bandwidth they have. Not, IMO, by forbidding the serving of the same data that *drives* the growth of the internet, and *sells* those cable modems for @Home - that popular data that you state should be restricted.
Are the majority of @home subscribers are mainly concerned with downloading linux quickly? D'oh! If it wasn't for such items as the porn and even such troublesome things as the Star Wars trailers, there wouldn't be an @Home to subscribe to, like it or not. There probably wouldn't be V.90 modems yet. I'm not sure what extent MP3 will drive technology. But, like it or not and the RIAA doesn't, MP3 is a legal, legitimate music format, and via it or another format, music *is going to be distributed on the Internet* for MP3 players. And to quote Angelo Sotira (who runs the Dimension Music site) from the Wired article at http://www.wired.com/news/news/culture/story/20427 .html: "I'm 18... and no matter what advertising you run or how you promote to me, I will never buy a CD again." So much for keeping MP3 off the Internet.
Personally, I'm glad that "adult entertainment" and other hedonistic services have helped provide incentive for fast connections and backbones to be developed and deployed. And will continue to do so. Yes, providing bandwidth-intensive services, and so on.
And I suspect that in the year 2005 at the latest, I'll hear people complaining about the bandwidth used by teledildonics slowing Linux downloads on T3+ connections probably developed for that very industry. Let's just think a little about whom is leeching off whom.
What does whether or not a site has "Warez/mp3/porn(Insert Illegal stuff here)" have to do with have to do with the subject of the article? Data is data, after all - whether it contains software designed for network monitoring, some unknown artists' digital recordings or erotic imagery, or even something illegal. It takes the same time to send the same bits, after all (though I wouldn't advocate adding "illegal" to the three categories you specifically name). And after all, the bandwidth wouldn't be used unless someone wants and requests the data. Nope, totally off the subject.
It just occurred to me: @home is being awfully foolish to (apparently) implement this inside their network. If, as I gain the impression is the case, those packets are actually being sent over segments of their network before being dropped, then they aren't gaining much if any benefit for the other subscribers. Their client software in the subscriber's PC *should do traffic shaping itself*. If the problem only exists between the cable modem and the NIC (and the cable modem isn't just a modem+bridge), then the cable modem's own networking is flawed.
The only reason I can see for doing it further into @Home's network would be to *deliberately* screw up so-called "subscriber abuse" cases, who are actually trying to use their connections, and to condition those subscribers to use the internet in just those few ways desired, like reading the WWW and their @Home email box. Rather a rancid tactic, and the only reason they could have any objection to the Linux Traffic Shaper limiting to 127.9k or whatever.
If they should say otherwise, and that even less than 128K outgoing is "abuse", then the obvious question would be "Okay, so what bandwidth *am* I allowed?" -- 2nd quote from/usr/src/linux/Documentation/networking/shape.txt:
"The shaper shapes transmitted traffic. It's rather impossible to shape received traffic except at the end (or a router) transmitting it."
Unless, you don't mind fscking up the person sending the traffic.
From the help for the Linux Traffic Shaper option ("experimental"): CONFIG_SHAPER: x x x x The traffic shaper is a virtual network device that allows you to x x limit the rate of outgoing data flow over some other network x x device. See Documentation/networking/shaper.txt for more x x information. To set up and configure shaper devices, you need the x x shapecfg program, available via FTP (user: anonymous) from x x ftp://shadow.cabi.net/pub/Linux in the shaper package.
The shaper limits outgoing traffic. In the 2.2.xx documentation it lists the maximum as "about 256K, it will go above this but get a bit blocky."
If you try limiting to just under the 128K cap, then packets should not be dropped as readily. Assuming that the cap and not simply an overloaded network is responsible.
Nope, except for near-mythical multicast broadcasts, at least as much is read from a web or cache server onto the network as is sent to the user - usually more, due to dropped and resent packets - and the server receives at least as many bytes of acks as the client sends it. Perfectly symmetrical.
If, however, you are only concerned with the cable modem end rather than the Internet as a whole, then in the cases the article is concerned with you have it exactly backwards - the servers require as much or more bandwidth to the Internet as from.
What ISPs are doing when they regulate by content (ala httpd servers) or data direction is not supplying "a connection to the internet", an IP channel that might be considered a common carrier or even pseudo-common carrier service; but instead, well, an online service - one that will be, frankly, virtually useless for many new internet applications in a very short time. In no way does it deserve to be awarded the title of "broadband" - I don't think it deserves to be called an "internet" connection. I won't repeat the reports and examples in previous comments of how throttling outbound connections is already harming inbound connections and preventing normal use of existing programs.
Basically, what we are seeing here is greed - @Home wants to reserve all its outbound bandwidth for its own web servers, or resell same bandwidth to someone with a similar use. They don't see their customer as anything more than a passive observer of sites on the real internet, and are effectively destroying the ability of their customers to communicate their ideas and contributions to others on the internet. "You must listen, but cannot speak." In other words, it is exactly backwards of what the Internet was designed and built as - a *two* way service which gathers its data from the whole net, as well as sends - this is a concept known as "interconnectivity", or a "free Internet".
It is also a concept that is inimical to @HOME's preference for serving their customers data from cache servers rather than the actual Internet. This works a *lot* better when the majority of the requests are for a few "portal" pages and those within a few links of them. Diversity is *not* appreciated, please - serve our customers dull, static (literally) pages! Hardly a surprise that @HOME would discourage the idea as much as possible - Note the #1 listed goal in the leaked document, above the "upstream enhancement":
"1.ONadvantage Proxy Evangelization - Improve speed by storing data as close to the subscribers as possible by using proxy or caching servers."
I don't consider this dishonest, really. That's what @HOME sells, and that's what their customers are buying from them. Sure, they would have to charge a little more to let their customers talk as well as listen dumbly. What would be dishonest would be pretending that they are selling a real IP connection. The danger I see is that this and similar policies are becoming the rule, rather than the exception, and *real* broadband internet connections may become unavailable - strangling in the cradle new inventions and innovations that could otherwise have as much promise as email or the WWW.
I see only two outcomes: either this kind of thing will go on until companies are calling anything but reading static web pages "subscriber abuse" as Mr. Wolfram puts it - people will be afraid to call the ISP when their videoconferencing fails lest they be told they are stealing bandwidth (and have their link summarily *further* castrated) - or, and I hate to say it, there must be rapid and comprehensive FCC and other regulation.
Because it costs more to provide Internet connections than those services @Home sells, there should be incentives, in the spirit of the FCC deregulation's goal of enhancing new communications technology.
First and foremost, it is essential that any communities considering creating or renewing franchise agreements, must be certain that those agreements allow the citizens of those communities to have a voice so their opinions may be heard; by guaranteeing the ISP will not hog all the outgoing bandwidth, so that those in the franchise area may use their connections to run legtitimate servers, or to read/send whatever other data they may choose to communicate. For personal pleasure, internet business, or otherwise. There should be strong attention to the nature of the equipment to be installed as well, to provide decent rates in both directions - DSLite must be considered simply unacceptable except as a first step towards broadband IP, or unless other options like HDSL are also available, priced proportional to bandwidth or better.
On the Federal and State level, there is no reason that online services who take advantage of and muzzle their users' outgoing channels, should be treated the same for tax or regulatory purposes as actual ISPs who respect freedom of speech and the Internet. Existing and proposed taxes on services or the DSL connections for such should be moved to reserved for those who would otherwise profit by providing an inferior "fast" connection deliberately damaged in such manners, and true common carrier status could be granted to those, by contrast, who are acting as true common carriers.
I tried a couple of these boards in an SMP WinNT configuration. Both consistently corrupted data on two systems with different hardware, and I was unable to get any real help from Promise Support ("Could you try moving the drives farther apart, then call me back?").
Unfortunately, it is not "all done in the hardware". It turns out to be a rather ordinary UDMA/33 card with a BIOS that simulates RAID hardware, and needs special drivers for OSes like Linux that don't use the BIOS. I have heard that works in a nonraid fashion with late Linux kernels, which would allow software RAID.
A linux driver would be nice. I didn't have much luck with the card in a multiprocessor NT 4.0 situation, though; two different versions of the FastTrak card, actually. Despite all efforts over about two weeks, the filesystems quickly deteriorated even as NT was being loaded. With two different motherboards and sets of disks, cables, et cetra. The problem disappeared when only a single processor was used.
Slashdot Mirror
The PPro's fatal flaw was that it could not execute 16-bit legacy code without being crippled. Intel fscked up by assuming that this would force *everyone*, including Microsoft, to develop and purchase software that no longer used that legacy code. It did not work out that way.
:-)
The flawed design was only purchased for the small segment of the market which was 32-bit, such as NT and *nix, so it became Intel's biggest flop since the iAPX432 (it was in desperation after that previous debacle that Intel as quickly as they could hacked together a strange, asymmetrical, segmented design out of their 8080/8085 CPU called the 8086).
So, Intel resorted to incremental changes while they fixed the problems with the PPro, the Pentium MMX processors. And AMD gained market share by using the opportunity. Microsoft traded off making Windows 95 compatable with legacy applications for what they might have gained by going all 32-bit and making Intel happy. They are not partners, each company has to respect what the market share of the other means to their own profits but it is each companies' duty to maximize their own for their stockholders.
So AMD managed to build market share and compete when Intel was knocked back, and all the other observations in the rest of your comment.
Democracy can be so inconvenient, can't it?
Well, the Harvard authorities are reportedly returning the site. Presumably it will be restored, somewhere. Past experience says that the controversy will be resolved in and by the glare of publicity and concentrated attention.
One credible? The other? Both, despite specific disagreements and rivalries? Neither?
Oddly, I remember India taking a moral stand against such tests done on other animals, and prohibiting their export to the USA because of very similar experiments.
;-)
Perhaps now that they have their own nukes they've changed their mind some?
Well, of course! But it would be morally bankrupt not to warn unwary potential consumers as well. Also currently at issue is @Home's resistance to sharing their outer network and allowing consumers to pick which ISP to connect directly to at the cable aggregation points - the quality of service provided, and treatment of IP customers, relates directly to that issue. Which in turn relates to the ability to cancel the service and go with someone else.
It limits free speech immensely - that is part of @Home's business model. The point is, a subscriber just can't *go* anywhere on the Internet except in the "audience" direction, if they use a network provider that bluntly forbids or cripples servers. The difference in capability is like the difference between a cheap transistor radio and a cellular phone.
:)
:)
Think about it - if Microsoft would, for example, by default enable their Personal Web Server during Win9x/20xx's install/OEM customer setup; give them the option to walk through one of those pick-a-color-and-layout-fill-in-the-blanks whizzards; and give them a few examples of what they can do - then @home and other ISPs that overcensor servers would be totally screwed. Tens, perhaps hundreds of thousands of angry customers, all demanding that this standard feature work, so at least a few people can see their vanity page; so their artistic little Jennifer's friends and the rest of the world can see the webcam on her special page - and the database of older images she's learned to build and web-enable - and later, so she can do that from the Linux server she's learned to run. That the WWW cache be turned on in both directions. And so on. Someone pointed out earlier that @Home's server prohibition does not extend to running NetMeeting. Can anyone or everyone guess why?
Unfortunately Microsoft probably won't do that, because 90 percent of the users would set their home page to their home page instead of MSN's "Start page" - they want to advertise as much as @Home does. Well, maybe not quite as much - from @Home's anti-open-access propaganda:
"If [...] cable companies were required to strip away the potential for generating advertising and commerce revenues by pursuing an access-only model, this would lead to an increase in the price of cable modem services. That isn't in the interest of the consumer."
Unless those consumers choose among the rest of the ISPs as well as @Home, and competition does its usual dance on prices.
So. Under @Home's short-sighted business model, everything that does not issue their advertising is abuse. Any of their own customers who issue their own data - yes, those paying for IP service - are treated as competitors! I suspect many ISPs who covet their customer's outbound bandwidth for the ISP's own WWW service do the same.
There aren't many businesses that could get away with that kind of crap treatment of their own clients, and there isn't any reason IP purchasers should have to put up with it.
I feel three-quarters ready to create a Slash engine site for small server questions, instructions, tips, announcements, and so on. Any comment on that idea?
alone. Yup, including
For one next-door neighbor and friend only.
With copper, was a nightmare due to ground loops. Ground one end, the other, both, neither, the current flow kept things from working.
I had to go to two AUI-to-fiber adapters I was lucky enough to find at a low but still painful price. With only optical fiber and no electrical connections things were fine, of course.
I suppose it takes considerably less infrastructure than the providing of a half-million of the internet's connections from smaller companies, given that this involves also receiving payment from all those half-million people. The overall bandwidth cost per user would be lower, because statistics guarantee that the more users, the lower the bandwidth needed for each. Sharing infrastructure costs with cable television, software taking over tasks that would be done by humans at smaller companies, and standard economies of scale in purchasing, provisioning, and other areas should make up for much of the higher bandwidth that is briefly needed for those quick page updates.
The rate cap is a cheap, easy, and rotten solution. Even supposing it isn't intended to screw up not only the "warez monkey" on the block but the legitimate user as well, it couldn't have been better designed to.
I hope you haven't been taken in by @Home's fantasy that they can cache the internet, or otherwise, without actually buying significant peering bandwidth, provide enough content to be other than a "half-assed experience" compared to that available to those with full IP connections. The idea is absurd - ultimately there is going to be about as much sucked from one part of the Internet as another.
The next "killer app" that emerges to equal importance with email and the WWW could very well need as much upstream as downstream bandwidth. Interpersonal multimedia communications tend to be two way, for example.
The choice between abandoning new technologies and adding more bandwidth should be a no-brainer.
It will be years, yes, until T1 speeds are available for that price. Continuously. But there isn't any more reason not to allow fast uploads briefly than there is not to allow downloads at those speeds for the same amount of time - even if just a few persons use the vast majority of the upstream bandwidth, there should be more than enough remaining for the relatively small upstream needs of the rest of the users.
As for whether the servers might or might not feature MP3 or warez, or pet pictures, or whatever a person might feel should somehow have less priority than their own oh-so-important data needs, I choose to sayeth and care not. Other people's data is theirs; what I read and serve is mine.
@Home probably loves the idea of all those customers of theirs, each watching their own RealVideo stream of that press conference. :-)
Well, he's right about that being "the type of user @Home is looking for". In fact, I gain the impression that they would prefer would:
Buy the cable modem subscription for status, or because they are told it is "good".
Be in the wide range between too computer-naive to effectively use a computer's capabilities on the internet and not so naive that they cannot run a Windows 98 box.
Not have had a previous decent internet connection to apply the advertised speed multipliers to.
Become quickly bored with the Internet as @Home Would Have It Be and thus condescends to present it to them, but keep the account to read email with once a month.
As for the credibility of his sociology, I simply suggest the AC author of the previous response look at the Magellan Search Voyeur at http://voyeur.mckinley.com/cgi-bin/voyeur.cgi - unless their personal ethics forbid reading other than easily cachable static pages.
What? do the math, buddy!
He has a T1. Max speed: 1544000 bits/s.
His audio channels use "128k". at least 128000bits/s. 12 of these could optimally fit in the T1.
He is serving the internet needs of up to 350 people. This is an amplification rate of 29 times the bandwidth from his net to the multicast servers!
And you think *that* 29:1 ratio is unacceptable?
It's making me feel better about my 128K ISDN connection. If my dedicated connection is going to cost three times what a cable modem does, then I want it to work as well at least. :-)
Exactly! And for a connection going in the opposite direction, trading the server and client end, just the opposite condition prevails. So it all balances out over the whole Internet, get it?
5. Don't try to put in a set of water mains and promote water use beyond what your sewer system can handle, even if you can make more selling water than charging for processing sewage. And if you do make that mistake, don't bitch that your customers are overloading the "outgoing" channel.
Umm, don't see it. Because:
7 .html: "I'm 18 ... and no matter what advertising you run or how you promote to me, I will never buy a CD again." So much for keeping MP3 off the Internet.
In large, and despite fantasies of caching sheeplike subscribers into a desired few web pages, there is as much incoming John Doe as there is outgoing John Doe.
The packets don't just appear on the net magically - where are those 10, 100, etc John Does going to get their data, otherwise - while probably *not* themselves using much outgoing bandwidth, and thus freeing their own net's outgoing bandwidth up?
@Home has X amount of incoming network bandwidth. You can bet your bottom stack location @Home doesn't use ADSL to its peers - it has X outgoing bandwidth wired up too.
Caching servers don't deal well with dynamically built pages (like the one you're reading), responses to POSTed data for forms, or pages not in the ten percent of URLs requested ninety percent of the time (or whatever percent rule web cache servers depend on). These have to come off of the Internet. Unless, of course, the page requested is already *on* an @Home subscriber's server.. oops, that's out.
Okay. The data is coming from somewhere. and going to somewhere. "Flash Crowds" are likely to occur, tying up networking to the locations doing the serving, and network-adjacent sites. This means that - gosh - those people aren't tying up other sites!
I'm not saying someone should be allowed to operate a site that continuously serves at megabit data rates to a cable modem. But @Home is the one that is "leeching" if it feeds off the Internet's data without providing a channel to feed as much data to the net, and trying to cleave the internet into an unnatural division between "subscriber" and "content provider". To the extent that pages cannot be served at decent rates even to a few users, despite @home having *more* outgoing than incoming bandwidth free - the proportion of pages cached is not going to be anywhere *near* the ratio of outgoing to incoming packet bytes from apps like web browsers.
The solution to the problem: An ISP should not be so greedy as to *combine* caching and selling off even the limited outgoing bandwidth they have. Not, IMO, by forbidding the serving of the same data that *drives* the growth of the internet, and *sells* those cable modems for @Home - that popular data that you state should be restricted.
Are the majority of @home subscribers are mainly concerned with downloading linux quickly? D'oh! If it wasn't for such items as the porn and even such troublesome things as the Star Wars trailers, there wouldn't be an @Home to subscribe to, like it or not. There probably wouldn't be V.90 modems yet. I'm not sure what extent MP3 will drive technology. But, like it or not and the RIAA doesn't, MP3 is a legal, legitimate music format, and via it or another format, music *is going to be distributed on the Internet* for MP3 players. And to quote Angelo Sotira (who runs the Dimension Music site) from the Wired article at http://www.wired.com/news/news/culture/story/2042
Personally, I'm glad that "adult entertainment" and other hedonistic services have helped provide incentive for fast connections and backbones to be developed and deployed. And will continue to do so. Yes, providing bandwidth-intensive services, and so on.
And I suspect that in the year 2005 at the latest, I'll hear people complaining about the bandwidth used by teledildonics slowing Linux downloads on T3+ connections probably developed for that very industry. Let's just think a little about whom is leeching off whom.
I guess you spend all your time on the @home home page, right? :-)
What does whether or not a site has "Warez/mp3/porn(Insert Illegal stuff here)" have to do with have to do with the subject of the article? Data is data, after all - whether it contains software designed for network monitoring, some unknown artists' digital recordings or erotic imagery, or even something illegal. It takes the same time to send the same bits, after all (though I wouldn't advocate adding "illegal" to the three categories you specifically name). And after all, the bandwidth wouldn't be used unless someone wants and requests the data. Nope, totally off the subject.
That's just it, @Home isn't being an internet node, just a leech online service sucking off it.
Fine, as long as they're willing to put the second part in writing. :-)
It just occurred to me: @home is being awfully foolish to (apparently) implement this inside their network. If, as I gain the impression is the case, those packets are actually being sent over segments of their network before being dropped, then they aren't gaining much if any benefit for the other subscribers. Their client software in the subscriber's PC *should do traffic shaping itself*. If the problem only exists between the cable modem and the NIC (and the cable modem isn't just a modem+bridge), then the cable modem's own networking is flawed.
/usr/src/linux/Documentation/networking/shape.txt:
The only reason I can see for doing it further into @Home's network would be to *deliberately* screw up so-called "subscriber abuse" cases, who are actually trying to use their connections, and to condition those subscribers to use the internet in just those few ways desired, like reading the WWW and their @Home email box. Rather a rancid tactic, and the only reason they could have any objection to the Linux Traffic Shaper limiting to 127.9k or whatever.
If they should say otherwise, and that even less than 128K outgoing is "abuse", then the obvious question would be "Okay, so what bandwidth *am* I allowed?"
--
2nd quote from
"The shaper shapes transmitted traffic. It's rather impossible to shape received traffic except at the end (or a router) transmitting it."
Unless, you don't mind fscking up the person sending the traffic.
From the help for the Linux Traffic Shaper option
("experimental"):
CONFIG_SHAPER: x
x x
x The traffic shaper is a virtual network device that allows you to x
x limit the rate of outgoing data flow over some other network x
x device. See Documentation/networking/shaper.txt for more x
x information. To set up and configure shaper devices, you need the x
x shapecfg program, available via FTP (user: anonymous) from x
x ftp://shadow.cabi.net/pub/Linux in the shaper package.
The shaper limits outgoing traffic. In the 2.2.xx documentation it lists the maximum as "about 256K, it will go above this but get a bit blocky."
If you try limiting to just under the 128K cap, then packets should not be dropped as readily. Assuming that the cap and not simply an overloaded network is responsible.
Nope, except for near-mythical multicast broadcasts, at least as much is read from a web or cache server onto the network as is sent to the user - usually more, due to dropped and resent packets - and the server receives at least as many bytes of acks as the client sends it. Perfectly symmetrical.
If, however, you are only concerned with the cable modem end rather than the Internet as a whole, then in the cases the article is concerned with you have it exactly backwards - the servers require as much or more bandwidth to the Internet as from.
What ISPs are doing when they regulate by content (ala httpd servers) or data direction is not supplying "a connection to the internet", an IP channel that might be considered a common carrier or even pseudo-common carrier service; but instead, well, an online service - one that will be, frankly, virtually useless for many new internet applications in a very short time. In no way does it deserve to be awarded the title of "broadband" - I don't think it deserves to be called an "internet" connection. I won't repeat the reports and examples in previous comments of how throttling outbound connections is already harming inbound connections and preventing normal use of existing programs.
Basically, what we are seeing here is greed - @Home wants to reserve all its outbound bandwidth for its own web servers, or resell same bandwidth to someone with a similar use. They don't see their customer as anything more than a passive observer of sites on the real internet, and are effectively destroying the ability of their customers to communicate their ideas and contributions to others on the internet. "You must listen, but cannot speak." In other words, it is exactly backwards of what the Internet was designed and built as - a *two* way service which gathers its data from the whole net, as well as sends - this is a concept known as "interconnectivity", or a "free Internet".
It is also a concept that is inimical to @HOME's preference for serving their customers data from cache servers rather than the actual Internet. This works a *lot* better when the majority of the requests are for a few "portal" pages and those within a few links of them. Diversity is *not* appreciated, please - serve our customers dull, static (literally) pages! Hardly a surprise that @HOME would discourage the idea as much as possible - Note the #1 listed goal in the leaked document, above the "upstream enhancement":
"1.ONadvantage Proxy Evangelization - Improve speed by storing data as close to the subscribers as possible by using proxy or caching servers."
I don't consider this dishonest, really. That's what @HOME sells, and that's what their customers are buying from them. Sure, they would have to charge a little more to let their customers talk as well as listen dumbly. What would be dishonest would be pretending that they are selling a real IP connection. The danger I see is that this and similar policies are becoming the rule, rather than the exception, and *real* broadband internet connections may become unavailable - strangling in the cradle new inventions and innovations that could otherwise have as much promise as email or the WWW.
I see only two outcomes: either this kind of thing will go on until companies are calling anything but reading static web pages "subscriber abuse" as Mr. Wolfram puts it - people will be afraid to call the ISP when their videoconferencing fails lest they be told they are stealing bandwidth (and have their link summarily *further* castrated) - or, and I hate to say it, there must be rapid and comprehensive FCC and other regulation.
Because it costs more to provide Internet connections than those services @Home sells, there should be incentives, in the spirit of the FCC deregulation's goal of enhancing new communications technology.
First and foremost, it is essential that any communities considering creating or renewing franchise agreements, must be certain that those agreements allow the citizens of those communities to have a voice so their opinions may be heard; by guaranteeing the ISP will not hog all the outgoing bandwidth, so that those in the franchise area may use their connections to run legtitimate servers, or to read/send whatever other data they may choose to communicate. For personal pleasure, internet business, or otherwise. There should be strong attention to the nature of the equipment to be installed as well, to provide decent rates in both directions - DSLite must be considered simply unacceptable except as a first step towards broadband IP, or unless other options like HDSL are also available, priced proportional to bandwidth or better.
On the Federal and State level, there is no reason that online services who take advantage of and muzzle their users' outgoing channels, should be treated the same for tax or regulatory purposes as actual ISPs who respect freedom of speech and the Internet. Existing and proposed taxes on services or the DSL connections for such should be moved to reserved for those who would otherwise profit by providing an inferior "fast" connection deliberately damaged in such manners, and true common carrier status could be granted to those, by contrast, who are acting as true common carriers.
I tried a couple of these boards in an SMP WinNT configuration. Both consistently corrupted data on two systems with different hardware, and I was unable to get any real help from Promise Support ("Could you try moving the drives farther apart, then call me back?").
Unfortunately, it is not "all done in the hardware". It turns out to be a rather ordinary UDMA/33 card with a BIOS that simulates RAID hardware, and needs special drivers for OSes like Linux that don't use the BIOS. I have heard that works in a nonraid fashion with late Linux kernels, which would allow software RAID.
A linux driver would be nice. I didn't have much luck with the card in a multiprocessor NT 4.0 situation, though; two different versions of the FastTrak card, actually. Despite all efforts over about two weeks, the filesystems quickly deteriorated even as NT was being loaded. With two different motherboards and sets of disks, cables, et cetra. The problem disappeared when only a single processor was used.