What do you consider large DBs? I've used a 4.51 server under NT4 with only 64MB (on an old 166Mhz box!), and only when the databases grew over 4GB to approx 8GB total, the machine then needed 128MB (plus a reboot every 60-90 days).
Ok, there were only 10 or so users, but it was fast enough for all email and database activity.
The only gripe I have with Domino R5 for Linux is that imap/pop3 serving is so slow. Much slower than a simple standard pop3d/imapd on a similar Linux box.
I have used 4.51 for years, even tried it with wine years ago and it worked nicely, and never seen a problem like that. Have since switched to R5 and will try it with Wine tomorrow (good link!).
Luckily there are many people per provider per geographical region. Between them, there is definite gain to be had. And at the whole stretches between the multicasting server and the exchange points of the providers, the gain is even more because the multicasting traffic for multiple providers is merged in a single stream.
Sure, multiple providers in a particular region without an exchange point in the region results in duplicate traffic. That's just a choice of the provider, setting up a local exchange point costs money too, and routing it a few hops more may be cheaper.
s/physical geography/internet topology.
True, the Internet does not give much about geography. When I said neigbourhood or 'two blocks away', please think in Internet topology. 'two hops away' as you wish.
I referred to reduction of traffic on the the backbones. With DSL and Cablemodems giving everybody a really nice endspeed, it all accumulates on the backbones and the routers just can't handle it. They are minimizing routing table length as it is (of course, IPv6 should alleviate that problem too). Without multicasting, they will never be able to support real multimedia to the end user. On a good day, an expensive 10Gbit OC192 stretch can only serve 5000 people with a unique 2mbit stream without multicasting. Think of only half a million people active behind a particular backbone stretch: half a million streams of 2 megabit, that translates into 100 OC192 links... No backbone will be able to carry that kind of traffic for a long time. So, if we want real multimedia content now, we have to use multicasting.
Every web medium/big web site gets a couple of hits per second at least. Hence, you're never watching it alone. It just feels that way.
You're right about multicast for html and images, that can only be proxied. I guess I'm also saying that web pages with html, images, javascipt, flash etc are not all that uebercool either. Sure, it can be extremely good, if the content is good (heck, I even like www.terraserver.com, even though it is from the Monopolist Intimidatorsoft(TM)), but as a multimedia experience it is still mediocre. I still get more of a jolt from watching a good action movie with surround sound on a large screen than from _ANY_ of the websites out there (and I do have large monitors).
Web servers are still missing something, it's still nothing more than just a faster-than-dialup-Internet experience. There is no Information Superhighway-feel to it. It still hasn't even come near to its potential.
In order to compete with watching a movie, I'm saying that web sites need more multimedia content
Maybe I should have been more clearer about that in my previous post.
Multicasts can start every minute too, you know. As soon as two people subscribe to it, it saves bandwidth. With multicasting, watching movies over the internet on request becomes a possiblity. Plus, when integrating such streaming content on a web site it not only increases the experience, it even upens up new possibilities for the income site like replacing those annoying flickering banner ads with some nicely flowing streams. The backbones won't be able to handle all that bandwidth unless multicasting is used.
Sure, we have caching for the web. to be honest, as a web user paying for my own clicks, I frankly don't give a rats ass about a web server log. A good cache does If-Modified-Since requests anyway (and send the original IP in its headers too), so the logs are fixable. A stale cache only happens when the web server has wrong "Expires:" headers or does not respond well to if-modified-since requests. Which too many sites out there do...
(FYI, a proxy really does wonders on site like slashdot, even although only the images are cached due to the dynamic content. All pages load at least twice as fast with proxy).
I really hate it when web designers that think that their server is soo important that they pragma:no-cache everything, even the images that change less than once a month.
Everybody uses a cache all the time for their personal browsing, builtin into netscape/iexplode.exe. Plus I use squid with 1GB of cache to speed up all those slow webpages with a gazillion images, resulting in the front page taking seconds to load even on a cablemodem. With my cache, all that slowness is history because the images almost never change, and if they do, it only a few will be changed so it will still be fast. yes my proxy sends IMS requests, so your logs show my hits. If people would proxy more, there would be a lot less 'Server too busy' pages from IIS-based junkservers.
"everyone wanting the same streaming video at the same time"
You're overdoing it. That is broadcasting, multicasting is a lot more selective. It doesn't have to be everybody. _Any_ number of people greater than one behind the same backbone (or from the server upstream) router will do. Even if there is a unique stream for two netizens in a particular corner of the net, then traffic to those two netizens is reduced by 50%.
And with thousands of people hitting popular spots on the net simultaneously, there is a huge potential gain.
You may think, for example, that you're the only one browsing slashdot and watching live streaming lectures from your favorite scientist, but there may be somebody in your city doing the exact same thing. If there is only one, then mutlicasting helps.
Look at it from the other side: A web server is it's popular pages, images, or files at a high rate (N times per second). And that server has only one or two upstream providers. So the exact same data packets are sent upstream only split seconds after each other. Multicasting here would save a lot of bandwidth up to the routers where the streams must split, and all the way to the web reader if they are behind the same backbone router.
The web browsing user wouldn't even notice, and no server would ever have to send the same content more than a few times per second, no matter how many hits they get.
Ok, an example: You and somebody 2 blocks away are both browsing slashdot. All the other people are doing other unique stuff. So why send all those slashdot images twice? Now replace 'images' with any imaginable content.
Of course not. If IMC were to second guess the meaning, they could just as well assume that the IP number is correct and a typo was made when they looked up the owner of the netblock...
If the exact letter of the order is not what matters, but the perceived intent, then I can give another nice example how that would never work:
A receiver may honestly think that it is a practical joke being pulled on them, or that it is a forged letter and not an actual court order, or maybe a stunt from a disgruntled former employee, then they could dismiss it and it would be legal. However, in reality, that is not the case.
A local geek was requested for comments? An actual geek or just a knowledgeable free software person?
If he/she was an actual pure 100% geek, then it may be a good time to introduce some geektreats into the law:
- All future laws should be numbered with the prefix 'RFC'
- Tax deduction on geekwear
- Make Donations of CPU cycles tax deductable.
- Make PING->ACK and SYN->ACK->RST official governmental language to be used in government activities (courtroom, house sessions, etc).
- URL-compliant world: All street and postal addresses to be written in URL format: Where do you live? "address://mycity/mainstreet/1034". Plus clickable street signs.
- Social security includes money for a DSL connection for everybody.
- More hours of open source computer programming classes in schools. "Ok children, let's go to lesson two: now everybody type in './configure ; make'"
- Make the congressional library available on gnutella and freenet. Plus post all new items on usenet.
- All governmental sessions to be live MP3 streamed.
(most suggestions above not to be taken seriously of course)
I must say I was surprised at the amounts mentioned here. I can see it must be pretty hard even with $49.95 montly.
I'm not sure about the situation in the US, but in Europe often when houses are built, only one telco is allowed (by the government) to install the local loop (last mile), which gives the exising single telco in the country a virtual monopoly.
So, with a government-supplied monopoly, they would try to squeeze out competition? That's just bad, plain bad.
(I'm not sure if they overcharge that much in Europe though, and there is healthy competition between multiple DSL providers (although the telco decides in which regions DSL is available, and users often do pptp (vpn) on the telco's network and are then routed to the DSL provider of their choice, so you're packets are on the telco network for longer than just the last mile))
In many European countries, though the telephone networks where built when the telco's were government owned _AND_ subsidized. Henze, the public paid for the network, not the company that happened to be the result of privatization.
I don't think the last mile is 100% of the difficulty with DSL. The sheer volume of traffic on the backbones as a result of people being connected at megabit speeds gives a whole slew of new problems. Suddenly the best router you can buy simply can't handle the traffic anymore... what to do now?... It simply takes time to design a large high volume TCP/IP network.
I do wonder why it is then, that we don't see more multicasting of multimedia content. Multicasting saves a huge bundle on the backbones and keeps broadband users happy. Lack of innovation from the content providers? Broadband users are not waiting for more flash plugins or flickering ads on web pages, they want the real multimedia Internet experience that was promised in promotions of the 'information superhighway'. Why isn't there a big race going on between radio stations to open shop on the Internet with streaming MP3? And how about the same for TV stations (at 640kbit and up, video often looks pretty ok) Sure, there is some streaming going on, but it does not yet seem like a revolutionary thing for what we currently know as radio and TV...
The future is...... hmm, wait...... hmm, wait...... first a lawsuit, then - we promise - we will innovate.........
And how would he have been able to increase the security of all the Linux systems on the net by trying to make another, much more seldomly used and already pretty secure operating system even more secure?
I think his target was to improve Linux, and he had a clue.
If they can't get root access, they can't change user. And with 32bit UIDs, which are available from 2.4 on, the browser can run in its lonely little dedicated UID space, without even having access to the user's files, just the browser cache and configuration.
There is your 'security model' for you. Much better than 'trust all officially microsoft approved activex applets and give only one prompt for all others'
You may be right on the speed of development part, except that 95% of the Linux users out there don't use the KDE or Gnome CVS to stay up-to-date with the latest features. Most wait until it's in their distrubution, so that they can 'rpm -i', or 'apt-get install' it.
Fast security fixes will only help if the distribution packagers are right on top of it with fast response of packaging the security-fixed versions. And even then, the user must known about it, or have an automated (cron) way of keeping up-to-date with just the security fixes.
I too had such problems in the past until I learned the following: There are a couple of things you can do to largely alleviate that problem. I hope these tips will help some of you guys out there:
- If possible, it's best to make all patches with 'apt-get source' followed by doing the patch and a 'dpkg --buildpackage' followed by installing the newly generated patched.deb... That keeps your database intact, plus allows you to easily put up the patches source and binary on your web site for others to enjoy.
- Alternatively, keep the original package installed (the patch might break other programs depending on the package), and install the patched binaries in/usr/local/packagename-version/*/ and modify the ld.so settings for the programs requiring the patched version. That's how I've been using Xfree 4.0 and 4.01 since long before it was aptable, and now still my database is 100% intact.
I think the speed increase continues to impress, at least for me and people like me.
We've seen definite performance increases on all of our uses of the processors: Simulations, for digital signal processing and ASIC chip design. Each time going from the (ancient) Intel P5/MMX to the good old K6 chip towards (dual) Celeron, towards the now current Athlon, we've been welcoming the speed increase each time. Thank the guiding master entity with religous background (for example: God) for the 1Ghz Athlons.
It shaves hours off the simulation time of large designs and a lot of time off the many large test database runs of the signal processing code.
I wonder how the new 1Ghz Thunderbirds arriving soon will compare with the older 800Mhz slot A athlon... Too bad the Thunderbirds don't fit on the Abit KA7, the only Athlon motherboard to work reliable with 1GB RAM.
... Immediately raising the question: "When can we start running the triathlons"?;-))
Everybody seems so pinned on specialty DDR SDRAMs and RDRAMs for speed, but when multiprocessing is discussed, nobody seems to mention that the processors will be sharing the memory bus, basically degrading the specialty RAMs to regular SDRAM... With biathlons, we _should_ be getting twice the DDR SDRAM speed in order to get all of the 200% speed that we hope to get... QDR SDRAM?
All right, similar overlooks have been made in the past... I agree with dual celerons beating single celerons, because the extra celeron costs so little compared to the whole system. But Atlhons are so much more expensive that without the additional RAM bandwidth, the dual socket A motherboards will probably best a better choice for dual Durons than they are for dual Athlons.
btw. the Athlon successor will have 'MAR' added to the name??? -> Marathon;-)
How long do you think it will take with such a method to etch out a large wafer?
I don't see how such a method can help continue moore's law without greatly sacrificing the other reason besides speed and power consumption why ASICs are so nice: cheap in volume.
Unique Document Locator / Directory Services ?
on
Gnutella Not Scaling?
·
· Score: 1
It sounds as if searching is the only thing to find information in a gnutella/freenet network.
Is anybody considering that we might want to use the protocol for hypertext? What I mean is such that we can type the unique document identifier in the browser location box, instead of http://www.blablablabla.com
Slashdot Mirror
What do you consider large DBs? I've used a 4.51 server under NT4 with only 64MB (on an old 166Mhz box!), and only when the databases grew over 4GB to approx 8GB total, the machine then needed 128MB (plus a reboot every 60-90 days).
Ok, there were only 10 or so users, but it was fast enough for all email and database activity.
The only gripe I have with Domino R5 for Linux is that imap/pop3 serving is so slow. Much slower than a simple standard pop3d/imapd on a similar Linux box.
I have used 4.51 for years, even tried it with wine years ago and it worked nicely, and never seen a problem like that. Have since switched to R5 and will try it with Wine tomorrow (good link!).
Luckily there are many people per provider per geographical region. Between them, there is definite gain to be had. And at the whole stretches between the multicasting server and the exchange points of the providers, the gain is even more because the multicasting traffic for multiple providers is merged in a single stream.
Sure, multiple providers in a particular region without an exchange point in the region results in duplicate traffic. That's just a choice of the provider, setting up a local exchange point costs money too, and routing it a few hops more may be cheaper.
s/physical geography/internet topology.
True, the Internet does not give much about geography. When I said neigbourhood or 'two blocks away', please think in Internet topology. 'two hops away' as you wish.
I referred to reduction of traffic on the the backbones. With DSL and Cablemodems giving everybody a really nice endspeed, it all accumulates on the backbones and the routers just can't handle it. They are minimizing routing table length as it is (of course, IPv6 should alleviate that problem too). Without multicasting, they will never be able to support real multimedia to the end user. On a good day, an expensive 10Gbit OC192 stretch can only serve 5000 people with a unique 2mbit stream without multicasting. Think of only half a million people active behind a particular backbone stretch: half a million streams of 2 megabit, that translates into 100 OC192 links... No backbone will be able to carry that kind of traffic for a long time. So, if we want real multimedia content now, we have to use multicasting.
Every web medium/big web site gets a couple of hits per second at least. Hence, you're never watching it alone. It just feels that way.
You're right about multicast for html and images, that can only be proxied. I guess I'm also saying that web pages with html, images, javascipt, flash etc are not all that uebercool either. Sure, it can be extremely good, if the content is good (heck, I even like www.terraserver.com, even though it is from the Monopolist Intimidatorsoft(TM)), but as a multimedia experience it is still mediocre. I still get more of a jolt from watching a good action movie with surround sound on a large screen than from _ANY_ of the websites out there (and I do have large monitors).
Web servers are still missing something, it's still nothing more than just a faster-than-dialup-Internet experience. There is no Information Superhighway-feel to it. It still hasn't even come near to its potential.
In order to compete with watching a movie, I'm saying that web sites need more multimedia content
Maybe I should have been more clearer about that in my previous post.
Multicasts can start every minute too, you know. As soon as two people subscribe to it, it saves bandwidth. With multicasting, watching movies over the internet on request becomes a possiblity. Plus, when integrating such streaming content on a web site it not only increases the experience, it even upens up new possibilities for the income site like replacing those annoying flickering banner ads with some nicely flowing streams. The backbones won't be able to handle all that bandwidth unless multicasting is used.
Sure, we have caching for the web. to be honest, as a web user paying for my own clicks, I frankly don't give a rats ass about a web server log. A good cache does If-Modified-Since requests anyway (and send the original IP in its headers too), so the logs are fixable. A stale cache only happens when the web server has wrong "Expires:" headers or does not respond well to if-modified-since requests. Which too many sites out there do...
(FYI, a proxy really does wonders on site like slashdot, even although only the images are cached due to the dynamic content. All pages load at least twice as fast with proxy).
I really hate it when web designers that think that their server is soo important that they pragma:no-cache everything, even the images that change less than once a month.
Everybody uses a cache all the time for their personal browsing, builtin into netscape/iexplode.exe. Plus I use squid with 1GB of cache to speed up all those slow webpages with a gazillion images, resulting in the front page taking seconds to load even on a cablemodem. With my cache, all that slowness is history because the images almost never change, and if they do, it only a few will be changed so it will still be fast. yes my proxy sends IMS requests, so your logs show my hits. If people would proxy more, there would be a lot less 'Server too busy' pages from IIS-based junkservers.
"bio-informatics."
Oh yes. So clear.
Please clarify.
"I'm sure these guys are having the time of their lives"
I'm not so sure about that. I saw a mention that at least one of these guys was involved in the Gemini and Apollo projects at NASA.
I'm sure that at least for that guy, this is just another fun project to toy with.
Working on a first in basically letting people sit on top of an explosion in an effort to leave the planet is hard to top.
"everyone wanting the same streaming video at the same time"
You're overdoing it. That is broadcasting, multicasting is a lot more selective. It doesn't have to be everybody. _Any_ number of people greater than one behind the same backbone (or from the server upstream) router will do. Even if there is a unique stream for two netizens in a particular corner of the net, then traffic to those two netizens is reduced by 50%.
And with thousands of people hitting popular spots on the net simultaneously, there is a huge potential gain.
You may think, for example, that you're the only one browsing slashdot and watching live streaming lectures from your favorite scientist, but there may be somebody in your city doing the exact same thing. If there is only one, then mutlicasting helps.
Look at it from the other side: A web server is it's popular pages, images, or files at a high rate (N times per second). And that server has only one or two upstream providers. So the exact same data packets are sent upstream only split seconds after each other. Multicasting here would save a lot of bandwidth up to the routers where the streams must split, and all the way to the web reader if they are behind the same backbone router.
The web browsing user wouldn't even notice, and no server would ever have to send the same content more than a few times per second, no matter how many hits they get.
Ok, an example: You and somebody 2 blocks away are both browsing slashdot. All the other people are doing other unique stuff. So why send all those slashdot images twice? Now replace 'images' with any imaginable content.
(Disclaimer: IANAL)
"It is not for IMC to second guess the order"
Of course not. If IMC were to second guess the meaning, they could just as well assume that the IP number is correct and a typo was made when they looked up the owner of the netblock...
If the exact letter of the order is not what matters, but the perceived intent, then I can give another nice example how that would never work:
A receiver may honestly think that it is a practical joke being pulled on them, or that it is a forged letter and not an actual court order, or maybe a stunt from a disgruntled former employee, then they could dismiss it and it would be legal. However, in reality, that is not the case.
A local geek was requested for comments? An actual geek or just a knowledgeable free software person?
If he/she was an actual pure 100% geek, then it may be a good time to introduce some geektreats into the law:
- All future laws should be numbered with the prefix 'RFC'
- Tax deduction on geekwear
- Make Donations of CPU cycles tax deductable.
- Make PING->ACK and SYN->ACK->RST official governmental language to be used in government activities (courtroom, house sessions, etc).
- URL-compliant world: All street and postal addresses to be written in URL format: Where do you live? "address://mycity/mainstreet/1034". Plus clickable street signs.
- Social security includes money for a DSL connection for everybody.
- More hours of open source computer programming classes in schools. "Ok children, let's go to lesson two: now everybody type in './configure ; make'"
- Make the congressional library available on gnutella and freenet. Plus post all new items on usenet.
- All governmental sessions to be live MP3 streamed.
(most suggestions above not to be taken seriously of course)
I must say I was surprised at the amounts mentioned here. I can see it must be pretty hard even with $49.95 montly.
I'm not sure about the situation in the US, but in Europe often when houses are built, only one telco is allowed (by the government) to install the local loop (last mile), which gives the exising single telco in the country a virtual monopoly.
So, with a government-supplied monopoly, they would try to squeeze out competition? That's just bad, plain bad.
(I'm not sure if they overcharge that much in Europe though, and there is healthy competition between multiple DSL providers (although the telco decides in which regions DSL is available, and users often do pptp (vpn) on the telco's network and are then routed to the DSL provider of their choice, so you're packets are on the telco network for longer than just the last mile))
s/henze/hence
cmdrtaco, can we have a spellcheker here?
In many European countries, though the telephone networks where built when the telco's were government owned _AND_ subsidized. Henze, the public paid for the network, not the company that happened to be the result of privatization.
... hmm, wait... ... hmm, wait... ... first a lawsuit, then - we promise - we will innovate... ... ...
;-(
I don't think the last mile is 100% of the difficulty with DSL. The sheer volume of traffic on the backbones as a result of people being connected at megabit speeds gives a whole slew of new problems. Suddenly the best router you can buy simply can't handle the traffic anymore... what to do now?... It simply takes time to design a large high volume TCP/IP network.
I do wonder why it is then, that we don't see more multicasting of multimedia content. Multicasting saves a huge bundle on the backbones and keeps broadband users happy. Lack of innovation from the content providers? Broadband users are not waiting for more flash plugins or flickering ads on web pages, they want the real multimedia Internet experience that was promised in promotions of the 'information superhighway'. Why isn't there a big race going on between radio stations to open shop on the Internet with streaming MP3? And how about the same for TV stations (at 640kbit and up, video often looks pretty ok) Sure, there is some streaming going on, but it does not yet seem like a revolutionary thing for what we currently know as radio and TV...
The future is...
sigh.
Either this patent is limited in scope, or even very common programs like tripwire are prior art...
And how would he have been able to increase the security of all the Linux systems on the net by trying to make another, much more seldomly used and already pretty secure operating system even more secure?
I think his target was to improve Linux, and he had a clue.
"Who cares if you can get root access?"
If they can't get root access, they can't change user. And with 32bit UIDs, which are available from 2.4 on, the browser can run in its lonely little dedicated UID space, without even having access to the user's files, just the browser cache and configuration.
There is your 'security model' for you. Much better than 'trust all officially microsoft approved activex applets and give only one prompt for all others'
You may be right on the speed of development part, except that 95% of the Linux users out there don't use the KDE or Gnome CVS to stay up-to-date with the latest features. Most wait until it's in their distrubution, so that they can 'rpm -i', or 'apt-get install' it.
Fast security fixes will only help if the distribution packagers are right on top of it with fast response of packaging the security-fixed versions. And even then, the user must known about it, or have an automated (cron) way of keeping up-to-date with just the security fixes.
Currently, that is not the situation we're in...
Am I the only one who looks at the date?
April Fools...
From what I understand not all can play 256kbit though...
In Europe, the BBC has a weekly show called 'Robot Wars'. It's been running for years and is really popular. Really nice entertainment.
Does anybody know if any US network has bought the rights to re-broadcast it in the US?
"for products only available on other continents."
Hey, 90% of the spam I received was that way when I lived in Europe...
Btw, am I the only one who notices that there is a strong correlation between the amount of spam and SCHOOL HOLIDAYS?
Geesh, most of these 'get richt quick' or 'buy my CDROM' spams are from teenagers or clueless collegers on a day off...
If most of the spammers are kids, then spam control is mostly in the hands of parents and schools. Better education and moral values might reduce it.
Btw, I think it will never go away, but in weeks like these (spring break), it's just too much.
I too had such problems in the past until I learned the following: There are a couple of things you can do to largely alleviate that problem. I hope these tips will help some of you guys out there:
.deb... That keeps your database intact, plus allows you to easily put up the patches source and binary on your web site for others to enjoy.
/usr/local/packagename-version/*/ and modify the ld.so settings for the programs requiring the patched version. That's how I've been using Xfree 4.0 and 4.01 since long before it was aptable, and now still my database is 100% intact.
- If possible, it's best to make all patches with 'apt-get source' followed by doing the patch and a 'dpkg --buildpackage' followed by installing the newly generated patched
- Alternatively, keep the original package installed (the patch might break other programs depending on the package), and install the patched binaries in
I think the speed increase continues to impress, at least for me and people like me.
We've seen definite performance increases on all of our uses of the processors: Simulations, for digital signal processing and ASIC chip design. Each time going from the (ancient) Intel P5/MMX to the good old K6 chip towards (dual) Celeron, towards the now current Athlon, we've been welcoming the speed increase each time. Thank the guiding master entity with religous background (for example: God) for the 1Ghz Athlons.
It shaves hours off the simulation time of large designs and a lot of time off the many large test database runs of the signal processing code.
I wonder how the new 1Ghz Thunderbirds arriving soon will compare with the older 800Mhz slot A athlon... Too bad the Thunderbirds don't fit on the Abit KA7, the only Athlon motherboard to work reliable with 1GB RAM.
LOL
;-))
;-)
... Immediately raising the question: "When can we start running the triathlons"?
Everybody seems so pinned on specialty DDR SDRAMs and RDRAMs for speed, but when multiprocessing is discussed, nobody seems to mention that the processors will be sharing the memory bus, basically degrading the specialty RAMs to regular SDRAM... With biathlons, we _should_ be getting twice the DDR SDRAM speed in order to get all of the 200% speed that we hope to get... QDR SDRAM?
All right, similar overlooks have been made in the past... I agree with dual celerons beating single celerons, because the extra celeron costs so little compared to the whole system. But Atlhons are so much more expensive that without the additional RAM bandwidth, the dual socket A motherboards will probably best a better choice for dual Durons than they are for dual Athlons.
btw. the Athlon successor will have 'MAR' added to the name??? -> Marathon
How long do you think it will take with such a method to etch out a large wafer?
I don't see how such a method can help continue moore's law without greatly sacrificing the other reason besides speed and power consumption why ASICs are so nice: cheap in volume.
It sounds as if searching is the only thing to find information in a gnutella/freenet network.
Is anybody considering that we might want to use the protocol for hypertext? What I mean is such that we can type the unique document identifier in the browser location box, instead of http://www.blablablabla.com
gnutella://lightbulbscewring.for.dummies.xml
freenet://why.not.eat.that.yellow.snow.FAQ.html
When the URL always is a 'search request', then we can't be sure that we get a document each time, let alone the same document.
How about incorporating some sort of directory system and resource location and/or identification method?
That could well result in making the web obsolete, old technology. Yes we've seen the Web, it was nice: Time to move on?