Bandwith loss of distance - am I being fed a line of bull or do I have research to do?

Question

I just had a strange conversation with a man who was trying to explain to me that it is impossible for two healthy networks to communicate at each-other over the ocean without significant bandwidth loss.

For example - if you have a machine connected at 100Mb/sec here http://www.hetzner.de/en/hosting/unternehmen/rechenzentrum attempt to communicate to a machine in the US with exactly the same setup you'd only achieve a fraction of the original connection speed. This would be true no matter how you distributed the load - the total loss over distance would be the same. "Full capacity" between the US and Germany would be less than half of what it would be to a data center a mile from the originator with the same setup.

If this is true that means my entire understanding of how packets work is wrong. I mean, if there's no packet loss why would there be any issue other than latency? I'm trying to understand his argument but am at a loss. He seems intelligent and 100% sure of his information. It was very difficult to understand because he explained data like a river and I was thinking of it as a series of packets.

Can someone explain to me what I'm missing, or am i just dealing with a madman in a position of authority and confidence?

Answer 1

He could be referring to the number of packets you would be able to have 'in flight' at any one time.

Take a look at Wikipedia's entry on Bandwidth Delay Product for some more information on this: http://en.wikipedia.org/wiki/Bandwidth-delay_product

That said, depending on the link you have between those two places, then I don't think latency would be that much of an issue to cause problems with this (assuming a fibre connection, not satellite).

He could also be referring to the fact that there would be a number of round trips to setup a TCP connection so the apparent speed to an end user who might be setting up lots of small connections (web browsing) might be less.

-Matt