> 100 baseT and fiber networks allow this to be done
Careful. One needs to pay attention to theoretical speed, practical speed, and comparative speed. The theoretical versus practical contrast is important because it can easily take an order of magnitude or more out of any given comparison. For example, most 10Mb/s LANs actually peak somewhere about 2Mb/s maximum throughput because the technique used for sharing the resource creates a lot of overhead with multiple users. You may get a burst through at 10Mb/s, but you won't sustain it. Likewize, ISDN at 64Kb/s for one B channel doesn't sound a lot higher than 56Kb/s on a modem (ignoring for the moment it is only 56Kb/s in one direction), but in practical terms the throughput is *much* higher because it is a digital signal so there are very, very few retries and such to lower the effective speed.
The comparative versus practical contast comes into play relative to expectations of what one proposes to replace, especially given that software folks seem to keep using every tiny bit of bandwidth and storage we give them (yes, I am one of them, but I don't work that way). So, 10Mb/s sounds pretty fast when you compare it to a 33Kb/s modem, but let's remember that we have gotten to the point where local disk speeds are running up into the 40MB/s range and without the competitive degredation mentioned above. So, lets say my word processor has this modest 2MB overlay to do some particular feature. On my local disk, this takes predictable small fraction of a second to load and I barely notice that it wasn't always there. Even on the cable modem, where I'll bet you will have trouble averaging 1Mb/s throughput, I have to wait 20 seconds. OK maybe for a download; not so OK if I am just doing some typing and happen to use a feature outside the base set. And when the net is slow? |
