On Wed, 28 Mar 2001, Jamie Ostrowski wrote: > Actually I thought it could be set up like a cb radio. You would have > a certain number of "channels" or different frequencies running across the > same wire, like broadband has "channels" as well going over coax. OK, um, that's what I'm talking about too. I think the use of the term "broadband" may be what's misleading here. There is no difference between broadcast RF signals and things that you can push down a wire, except that it doesn't take as much energy to do so. A "channel" is simply those frequencies between some lower bound and some upper bound. That is to say, by definition, a channel has some finite bandwidth. The sum of the bandwidths of all the channels sort of *has to* be less than the bandwidth of the transmission medium. So, if you want to give people channels, you're describing divvying up the available bandwidth. Sometimes, you can do this in such a way that throughput is optimized, but (if I understood correctly) you are trying to create *more* bandwidth. Channels would not do this uniformly. The thing that you can't get around is that frequency and time are flip sides of the same coin. You can't send data quickly without high frequencies, and you can't send a lot at the same time without *many* frequencies. If you have many frequencies, some are destined to be high freqs. So, you end up needing high frequencies again, hence, bandwidth. > Each host "ethernet" card would have a certain number of channels they > communicate on. Sure, but (see above) channels are band limited. > If any other hosts on the network need to > communicate, their "ethernet" cards find the lowest open channel and > transmit on that channel. Anytime any host wants to communicate with > another, it simply picks the lowest channel available. That's a good concept, but unfortunately there's nothing *simple* about it! Shared frequency is a very difficult and interesting area of engineering. Thing is, you'd be talking about NICs that cost more than the computers they were in. > the netword agree on that protocol, it would be very simple and not > too complex or time consuming. I tried drawing a simple map of > possibilities in hosts communicating across the network, and couldn't > come up with a way there would be conflicts. I think it'd be a little more tough than you realize. How would the receiving NIC keep track of what it was listening for on what channel? If you did it by sending data identifiers on each channel, I think you'd have a lot of overhead and bring your data rate down. Would it be enough to matter? Dunno, but I still can't get around how to do a transmission medium that is better than just a faster one (wider pipe.) > Another nice thing about this is it would prevent ping attacks. If your > machine is being pinged to death, you could go to the multi channel card > and put a lock on the channel your card is on, and simply open another > lowest channel. You would have a lot more freedom. You might; but you'd be working so hard to pay for the hardware that you might not get to enjoy it! <g> I am not suggesting that your idea's a bad one. But, there's some laws of physics to get around. Every digital signal is transmitted ultimately as an analog one. Remember, as the stupid quote goes, "186,000 miles per second: not just a good idea -- it's the Law!" -- "To misattribute a quote is unforgivable." --Anonymous