PhilFrisbie :
That would be true if TCP waited for each acknowledgment (ack) before sending the next message, but it does not. TCP will send a number of messages while waiting for acks, so it can saturate one direction of a full duplex link.
Yes I know that but this is not my question. I'll start again and incorporate your last answer. If I'm alone, downloading a large file on a high quality, 1/2 duplex link, where, say, 50 packets are shipped to one ACK, would you think it (the file) would come down at closer to 10 mbs, 5mbs, 0.5 mbs or at some other speed? As noted, I have 10 down and only 1.5 up so it seems to me that making the end-to-end service full duplex is a waste of money as I would expect that the 10/1.5 service offering pretty much reflects the kind of performance I would expect to get in a 1/2 duplex situation
As a supplemental, is it not the case that the line always transmits at 10 mbs (or at least at a constant speed) and it is more about the amount of time it can sustain the speed that dictates the throughput of the connection. It seems to me that it may be a like CPU busy - there is no such thing as 30% CPU - a CPU is always either 100% or 0% busy. It is just that when measured over a period of time, the CPU is found to be busy in 30% of the samples & not that the CPU goes at different speeds. So, if I get 3 mbs down does that not mean that 70% of the time the line is either idle or uploading and 30% of the time it is downloading at 10 meg?