WHOA! TCP/IP get's pwned!

Mobius · Post by **Mobius** » Tue Mar 16, 2004 12:55 am

http://www.ncsu.edu/news/press_releases/04_03/099.htm

The article specifically mentions Internet gamers.

That's gotta be great news.

JMEaT · Post by **JMEaT** » Tue Mar 16, 2004 1:18 am

Wolf on Air · Post by **Wolf on Air** » Tue Mar 16, 2004 1:25 am

I call bovine excrement. Besides, wasn't this posted some 6 months ago, too? I seem to recall reading this before.

Dr. Injong Rhee, associate professor of computer science, said BIC can achieve speeds roughly 6,000 times that of DSL and 150,000 times that of current modems.

Here they aren't talking about the protocol, but rather whatever gigabit fiber connection they tested it on. In other words, sensationalist writing. A protocol upgrade could possibly mean a 5-10% gain. Anything more is technically impossible. Since you seem to take pride in understanding the workings of the world, Moobie, maybe you should engage your brain before posting.

Still, I could always be wrong, but I'll believe it when I see it.

CDN_Merlin · Post by **CDN_Merlin** » Tue Mar 16, 2004 6:29 am

Wolf, they are talking about a PROTOCOL, not a fibre cable.

The protocol is named BIC-TCP, which stands for Binary Increase Congestion Transmission Control Protocol.

fliptw · Post by **fliptw** » Tue Mar 16, 2004 8:49 am

Umm.

In short, it claims to be a new protocol that manages network traffic to improve RTT times by doing something about dropped packets at tail routers.

Vertigo · Post by **Vertigo** » Tue Mar 16, 2004 9:01 am

Wolf on Air wrote:I call bovine excrement. Besides, wasn't this posted some 6 months ago, too? I seem to recall reading this before.

Dr. Injong Rhee, associate professor of computer science, said BIC can achieve speeds roughly 6,000 times that of DSL and 150,000 times that of current modems.
Here they aren't talking about the protocol, but rather whatever gigabit fiber connection they tested it on. In other words, sensationalist writing. A protocol upgrade could possibly mean a 5-10% gain. Anything more is technically impossible. Since you seem to take pride in understanding the workings of the world, Moobie, maybe you should engage your brain before posting.

Still, I could always be wrong, but I'll believe it when I see it.

I'll rephrase that...

They're talking about the hugely increased scaling possibilities, the new protocol can be scaled to WAY higher speeds than regular TCP/IP. Their wording is just poorly chosen, or as WoA said, it's just sensationalist writing.

Warlock · Post by **Warlock** » Tue Mar 16, 2004 9:25 am

IIRC NetBEUI is one of the fastest protocalls out there BUT its onley down fall is that it cant be routed.

but long time ago they talk about a differnt protocall that was ganna kill TCP/IP and i still havent see it.

i wonder when the new ip system comes out

Admiral LSD · Post by **Admiral LSD** » Tue Mar 16, 2004 9:37 am

IPv6? That's out and about now, I see people with v6 IPs joining the IRC channels I lurk in all the time. It requires some v6 over v4 tunnelling voodoo to work right atm but it's out there and people are using it.

Tricord · Post by **Tricord** » Tue Mar 16, 2004 10:44 am

You guys are scrambling everything.

There is a protocol stack, which can be seen here. The rightmost part of the diagram is the physical layer (actual cables), on which the data link layer operates (frame synchronisation, basic checksumming). This layer connects one physical network device to another (e.g. a modem to a modem bank, or a PC to a switch, etc.). After that comes the network protocol, in the case of internet IP (Internet Protocol). At this point, two versions can be used, IPv4 and IPv6. Note that the underlaying data link layer does not restrict the choice of IPv4 or IPv6. IP provides two transport layers, which are more known. UDP provides a connectionless, datagram-based delivery service, with no guarantees in terms of speed, integrity or delivery. TCP is a lot more complex, is connection-oriented, and provides among integrity and delivery guarantees also congestion control and flow control. It is this protocol they are talking about in the article.

It is no problem to define a third transport protocol next to UDP and TCP, but the problem is that no application uses it. Everything is built on TCP, so you can't switch to something else. You could discard TCP and replace it with the new protocol and make it so that the upper layers do not see any change, but this would require every host on the internet to upgrade at the same time.

WoA is right. TCP is not maximizing bandwidth throughput, but in exchange it provides flow control (if the PC on receiving end is on a slow link (e.g. dialup) the server will automatically adjust it's transmission speed to match the client) and it also provides congestion control (if TCP detects packet loss, it will ask for a retransmit at half the speed of the lost packet). This has as side effect that TCP is fair (two TCP connections over the same physical link will roughly get the same bandwidth, it's not that one connection will get everything and the other one no bandwidth at all).

TCP is actually far more complex than most of you imagine. Also, TCP/IP is senseless since you are mixing two layers. You could as well use TCP over another network layer, such as IPv6 or AppleTalk. Or, you could use UDP over IP, but we don't say UDP/IP do we?

CDN_Merlin · Post by **CDN_Merlin** » Tue Mar 16, 2004 10:48 am

IPv6 is also a headache to work with. Using a Hex numbering system for subnets is retarded. Even IPv4 and VLSM is retarded.

They should of reworked this ages ago and thought of a IP version that is easier to use and friendlier to subnetting or wasting of IP addresses.

fliptw · Post by **fliptw** » Tue Mar 16, 2004 12:00 pm

umm...

There isn't a really easier way of representing 32-bit and 128-bit numbers than what IPv4 and IPv6 are already using.

Topher · Post by **Topher** » Tue Mar 16, 2004 1:24 pm

fliptw wrote:umm...

There isn't a really easier way of representing 32-bit and 128-bit numbers than what IPv4 and IPv6 are already using.

Of course there is! Do it in whitespace, idiot.

1. More secure - No one can see what your address is!
2. More fast - Nothing is faster to transmit than empty space!
3. More economical - Something from nothing is always good.

Yah, what Tricord said.

Grendel · Post by **Grendel** » Tue Mar 16, 2004 2:31 pm

"Researchers in North Carolina State Universityâ??s Department of Computer Science have developed a new data transfer protocol for the Internet that makes todayâ??s high-speed Digital Subscriber Line (DSL) connections seem lethargic."

"The protocol is named BIC-TCP, which stands for Binary Increase Congestion Transmission Control Protocol."

"Dr. Injong Rhee, [..], said BIC can achieve speeds roughly 6,000 times that of DSL and 150,000 times that of current modems."

That doesn't make sense to me -- is it a new hardware or transprort layer ? Also 6000 times the speed of DSL ? I seriousely doubt one could do that w/ phone lines.

Krom · Post by **Krom** » Tue Mar 16, 2004 8:05 pm

At least not phone lines over 10 feet long anyway

Admiral LSD · Post by **Admiral LSD** » Tue Mar 16, 2004 8:44 pm

Of course, people believed it was impossible to get DSL speeds over regular phone lines once too

Sickone · Post by **Sickone** » Sat Mar 20, 2004 12:22 am

Grendelm - don't confuse what is said

The article didn't say it WOULD speed up DSL 6000x

It said given a PERFECT MEDIA, it could move 6000 times more data than a DSL connection on TCP/IP

Frankly - it is not very amazing to me that better protocols are available or could be created. BIG FRICKEN DEAL

The article is so poorly wrirren it makes it sound like if you could/did switch to the protocol you DSL would get 6000 times faster - pure BS

DSL running TCP/IP moving data (not game packets) has about 10% overhead worst case. So a magic protocol could only buy you 10%

the BIC-TCP is a concept to handle more data in bigger pipes

Avder · Post by **Avder** » Sat Mar 20, 2004 4:09 am

If you were going to eliminate overhead youd have to completely rework the MAC layer too. Presently layers 2-7 all add some kind of overhead in some way. Mostly just "where to send" data and CRC's and stuff. The way data travels over the network it would be impossible to make a routeable network protocol without some overhead. Even non-routeable protocols like NetBEUI have some overhead. And speed advertized by anyone or any device is always layer one speed, the actual number of bits that can be transmitted per second. Many of those bits arent actual data but the overhead that is needed to get your data from point A to point B.

Only on a 2 computer network with a perfectly loss-less, distroetion-less medium would you be able to use a system with no overhead. But then youd only be able to use exactly one application across the network because app differentiation would add overhead! Oh no!

As for this "TCP/IP gets pwned" stuff,

I'll believe it when I see it. TCP/IP works quite nicely for the global network that is the internet.

Vindicator · Post by **Vindicator** » Sat Mar 20, 2004 4:16 am

Youre gettin quite a bit of use out of that plate there...

Avder · Post by **Avder** » Sat Mar 20, 2004 8:54 am

I could make a few more if youre dying for variety.

Its the break in period. I need to get use out of them and see how they fit.

So far so good.